{"id": "2508.05762", "title": "Evaluating Universal Machine Learning Force Fields Against Experimental Measurements", "abstract": "Universal machine learning force fields (UMLFFs) promise to revolutionize materials science by enabling rapid atomistic simulations across the periodic table. However, their evaluation has been limited to computational benchmarks that may not reflect real-world performance. We introduce UniFFBench, a comprehensive evaluation framework featuring the MinX dataset -- a diverse collection of 1,500+ mineral systems spanning 85 elements, extreme thermodynamic conditions (0--5000 K, 0--1000 GPa), and structural complexity, including partial occupancy and disorder. This diversity, combined with experimental reference values for validation, enables assessment of UMLFF generalization across chemical space and conditions substantially beyond typical training scenarios. Our systematic evaluation of six state-of-the-art UMLFFs reveals a substantial ``reality gap'': models achieving impressive performance on computational benchmarks often fail when confronted with experimental complexity. Even the best-performing models exhibit higher density prediction error than the threshold required for practical applications. We observe disconnects between simulation stability and mechanical property accuracy, with prediction errors correlating with training data representation rather than the modeling method.", "date": "2025-08-07", "categories": []}
{"id": "2606.19375", "title": "Physics-Informed Discovery of Yield Functions in Plasticity via Convex Neural Representations", "abstract": "Identifying anisotropic yield functions remains challenging since yielding is not directly observed in full-field mechanical measurements, directional calibration can require many loading directions, and selecting an appropriate analytical form is nontrivial. This study proposes a physics-informed framework for discovering yield functions from full-field displacement data and reaction force data, without stress observations, plastic strain measurements, direct yield surface data, or a prescribed parametric yield function. The framework identifies the yield function as a mechanically constrained constitutive component inside elastoplastic stress integration, rather than through direct stress-space supervision. The yield function is represented by a convex neural network that enforces convexity and positive homogeneity of degree one while imposing the assumed tension-compression symmetry, and this neural yield function is trained with a differentiable stress update and a physics-informed force equilibrium loss across multiple loading cases. The proposed framework is validated using finite element (FE) benchmark studies with von Mises, Hill 1948, and Yld2000-2d yield functions, assessing yield contour agreement, displacement-noise sensitivity, identifiability through plastically active stress states, epistemic uncertainty, and polynomial-surrogate deployment. This study provides a mechanics-constrained pathway for discovering anisotropic yield functions from displacement and force data while keeping the identified component within the structure of elastoplastic stress integration.", "date": "2026-06-12", "categories": []}
{"id": "2603.28707", "title": "A Convex Route to Thermoelasticity: Learning Internal Energy and Dissipation", "abstract": "We present a physics-based neural network framework for the discovery of constitutive models in fully coupled thermomechanics. In contrast to classical formulations based on the Helmholtz energy, we adopt the internal energy and a dissipation potential as primary constitutive functions, expressed in terms of deformation and entropy. This choice avoids the need to enforce mixed convexity--concavity conditions and facilitates a consistent incorporation of thermodynamic principles. In this contribution, we focus on materials without preferred directions or internal variables. While the formulation is posed in terms of entropy, the temperature is treated as the independent observable, and the entropy is inferred internally through the constitutive relation, enabling thermodynamically consistent modeling without requiring entropy data. Thermodynamic admissibility of the networks is guaranteed by construction. The internal energy and dissipation potential are represented by input convex neural networks, ensuring convexity and compliance with the second law. Objectivity, material symmetry, and normalization are embedded directly into the architecture through invariant-based representations and zero-anchored formulations. We demonstrate the performance of the proposed framework on synthetic and experimental datasets, including purely thermal problems and fully coupled thermomechanical responses of soft tissues and filled rubbers. The results show that the learned models accurately capture the underlying constitutive behavior. All code, data, and trained models are made publicly available via https://doi.org/10.5281/zenodo.19248596.", "date": "2026-03-30", "categories": []}
{"id": "2606.18691", "title": "Robust and Interpretable Adaptation of Equivariant Materials Foundation Models via Sparsity-promoting Fine-tuning", "abstract": "Pre-trained materials foundation models, or machine learning interatomic potentials, leverage general physicochemical knowledge to effectively approximate potential energy surfaces. However, they often require domain-specific calibration due to physicochemical diversity as well as mismatches between practical computational settings and those used in constructing the pre-training data. To address this, we propose a sparsity-promoting fine-tuning method that selectively updates model parameters by exploiting the structural properties of E(3)-equivariant materials foundation models. On energy and force prediction tasks across molecular and crystalline benchmarks, our method matches or surpasses full fine-tuning and equivariant low-rank adaptation while updating only $\\sim$3~\\% of parameters, and in some cases as little as $\\sim$0.5~\\%. Beyond energy and force calibration, we further demonstrate task generalizability by applying our method to magnetic moment prediction and magnetism-aware total energy modeling. Finally, analysis of sparsity patterns reveals physically interpretable signatures, such as enhanced $d$-orbital contributions in transition metal systems. Overall, our results establish sparsity-promoting fine-tuning as a flexible and interpretable method for domain specialization of equivariant materials foundation models.", "date": "2026-06-17", "categories": []}
{"id": "2606.19133", "title": "Equivariant Graph Neural Networks Improve Optical Spectra Prediction for Materials Screening", "abstract": "Scalable prediction of optical spectra is a critical component of high-throughput materials screening for optoelectronic applications such as solar cells. Existing surrogate models are trained on spectra computed from lower levels of theory or rely on rotation-invariant scalar features, limiting their geometric expressiveness. We explore the use of equivariant graph neural networks for optical spectra prediction, adapting GotenNet to this task and evaluating it on multiple datasets including a recently published collection of 10,533 structures with spectra computed at the level of the random phase approximation (RPA). The proposed model outperforms the current state of the art, with the largest gains in the 0-8 eV range and on predicting the static real permittivity, both of particular relevance for thin-film optics.", "date": "2026-06-17", "categories": []}
{"id": "2410.08562", "title": "Adaptable Method for Crystal Design across Diverse Constraints and Objectives with Pretrained Property Predictors", "abstract": "Advanced crystal design can accelerate materials discovery across applications from photovoltaics to spintronics. Practical design must satisfy multiple properties and physical constraints, yet existing machine-learning-based approaches to such design often depend on large datasets, retraining, or task-specific generators. Here, we show that direct predictor-guided gradient optimization enables data-efficient, constraint-rich crystal design by combining off-the-shelf predictors with site-wise element masks, template initialization, and task-specific losses. In perovskites, it outperformed generative and Bayesian baselines under three targets -- band gap, formation energy, and tolerance factor -- and two hard constraints. DFT assessment further showed band-gap targeting competitive with a leading generative model despite using predictors trained on roughly one-tenth of the data. By flexibly combining pretrained predictors with application-oriented masks and custom losses, the same framework supported half-metal design. Such modularity could help researchers and engineers translate diverse application requirements directly into optimized candidate crystals with minimal computational cost.", "date": "2024-10-11", "categories": []}
{"id": "2605.29179", "title": "Sustainable Metal-Organic Framework Water Harvesters in the Artificial Intelligence Era", "abstract": "Metal-organic frameworks (MOFs) are excellent candidates for water harvesting due to their tunable pore environments, which can be precisely engineered to capture and release water in arid conditions. Integrating artificial intelligence (AI) into MOF discovery can further accelerate the design of high-performance sorbents by identifying structural features that enhance atmospheric water harvesting (AWH), stability, and cycling efficiency. In this Perspective, we examine key MOF design principles, including cooperative adsorption, operational relative humidity (RH), uptake capacity, hysteresis, and scalability. We highlight recent design advancements such as multivariate strategies and long-arm linker extension, and examine how these principles tune pore capacity and hydrophilicity, while preserving stability and crystallinity. Furthermore, we discuss how AI, large language models (LLMs), and data mining can accelerate the discovery process through predictive synthesis, inverse design, and elucidating synthesis-structure-property relationships for the next generation of MOF water harvesters.", "date": "2026-05-27", "categories": []}
{"id": "2606.17445", "title": "Toward Controllable Catalyst Inverse Design via Large-Scale Autoregressive Pretraining", "abstract": "Inverse design of heterogeneous catalysts remains challenging because catalyst surfaces exhibit substantial structural complexity with coupled surface-adsorbate interactions across a vast chemical space that is difficult to explore efficiently through conventional screening alone. Although machine learning-based high-throughput screening has accelerated catalyst discovery, its efficiency inevitably declines as the search space grows, motivating the development of generative models that can directly construct catalysts with target properties. Here, we present a conditional catalyst generative model based on the Generative Pretrained Transformer architecture with a numerical embedding layer that enables the generation of catalyst structures conditioned on both categorical and continuous properties within a single autoregressive framework. The model was pretrained on 133 million catalyst structures and subsequently fine-tuned on approximately 460,000 optimized structures with associated categorical properties and binding energies for conditional generation. The resulting model achieved 98% structural validity, 95% optimization validity, and high categorical condition fidelity, with a 93 % joint match rate for adsorbate type and composition. For binding energy conditioning, the match rate of approximately 20% represents a four-fold improvement over the baseline training distribution, and the generated distributions shift systematically toward the target values, enabling a 1.5 to 4-fold improvement in screening efficiency for reaction-targeted catalyst discovery without additional fine-tuning. These results show that large-scale autoregressive pre-training, combined with explicit property conditioning, provides a practical route toward controllable catalyst generation and accelerated catalysts discovery.", "date": "2026-06-16", "categories": []}
{"id": "2606.17668", "title": "ASTEROID: A Spatiotemporal Information Transformer for Forecasting Multi-Step Time Series of Molecular Dynamics", "abstract": "Molecular dynamics (MD) simulation is computationally demanding, particularly for large-scale systems requiring long-term analysis. Accurate forecast of the outcomes of a MD simulation is not only an attractive scientific challenge but also has substantial practical value. In this work, we developed a data-driven framework, termed ASTEROID (Advanced Spatiotemporal TransformER fOr Inferring Dynamics), that can directly predict multi-step atomic coordinates, avoiding conventional iterative integration. For this purpose, our ASTEROID reformulates MD trajectories as high-dimensional spatiotemporal sequences and integrates the Spatiotemporal Information (STI) Transformation equation into a Transformer architecture. The core innovation of ASTEROID lies in its ability to model multiscale spatiotemporal dependencies. In particular, for spatial dependencies, a local-global self-attention mechanism captures both short- and long-range interactions. For temporal dependencies, an encoder-decoder structure integrates global context with autoregressive forecasting. ASTEROID was evaluated on several quantum-mechanics derived molecular datasets. Our results indicate that ASTEROID achieved not only a higher level of accuracy in multi-step prediction than existing methods on various benchmarks, but also significantly reduced computational cost of conventional MD simulation. Moreover, the model supports iterative multi-step forecasting over an extended time scale. This work establishes a robust and generalizable data-driven paradigm for accelerating MD simulations.", "date": "2026-06-16", "categories": []}
{"id": "2601.21527", "title": "Sustainable Materials Discovery in the Era of Artificial Intelligence", "abstract": "Artificial intelligence (AI) has transformed materials discovery, enabling rapid exploration of chemical space through generative models and surrogate screening. Yet current generative AI models for materials discovery, which now drive exploration of vast chemical and structural spaces, optimize candidates exclusively for structural stability and functional properties, with no integration of environmental assessment at any stage of the design loop. Prospective and ex-ante life cycle assessment methods exist and have been applied to emerging technologies, but they operate as standalone downstream analyses, not as active constraints within generative or active-learning pipelines. The result is that environmental feedback, even when produced, arrives after design decisions have been made rather than informing them. The disconnect between atomic-scale design and lifecycle assessment (LCA) reflects fundamental challenges: (i) data scarcity across heterogeneous sources, (ii) scale gaps from atoms to industrial systems, (iii) uncertainty in synthesis pathways, and (iv) the absence of frameworks that co-optimize performance with environmental impact. In this Perspective, we propose integrating upstream ML-assisted materials discovery with downstream LCA into the ML-LCA framework, comprising five components: information extraction for building materials-environment knowledge bases, harmonized databases linking properties to sustainability metrics, multi-scale models bridging atomic properties to lifecycle impacts, ensemble prediction of manufacturing pathways with uncertainty quantification, and uncertainty-aware optimization enabling simultaneous performance-sustainability navigation. Case studies spanning polymers, glass, photoresists, and cement demonstrate both necessity and feasibility while identifying material-specific integration challenges.", "date": "2026-01-29", "categories": []}
{"id": "2606.15001", "title": "Distilling latent electrostatics from foundation machine learning interatomic potentials", "abstract": "Foundation machine learning interatomic potentials (MLIPs) have enabled atomistic simulations across broad regions of chemical and materials space, but many remain computationally expensive and lack explicit electrostatics, limiting their use for systems governed by long-range interactions and electrical response. Previously, we introduced Latent Ewald Summation (LES), which learns latent atomic charges and long-range electrostatics from density functional theory (DFT) energy and force labels alone. Here, we use LES to extract electrostatics that are latent in foundation models: energies and forces predicted by a teacher model are used to train a lightweight LES-augmented student MLIP, with optional fine-tuning on additional DFT data. The resulting models reduce computational cost while providing access to Born effective charge tensors, and infrared spectra. We benchmark student models distilled from a broad set of foundation MLIPs, including UMA, MACE, Orb, eSEN, GemNet-OC, PET, and EquiformerV2-based models, against experimental infrared spectra for liquid water, concentrated hydrochloric acid, and the anatase TiO2(101)-water interface. Across these systems, electrostatic response can be extracted from most foundation MLIPs. The benchmark further shows that the underlying DFT level and dataset used to train the teacher model play a larger role than architecture in determining electrostatic and spectroscopic accuracy. For the TiO2-water interface, fine-tuning with a modest amount of higher-level DFT data improves structural and infrared predictions. LES-based distillation therefore provides a practical route for converting foundation MLIPs into efficient, electrically responsive models, while also testing the physical fidelity encoded in foundation models.", "date": "2026-06-12", "categories": []}
{"id": "2606.15892", "title": "Scalar-pathway fidelity improves physical accuracy in short-range equivariant interatomic potentials", "abstract": "Accurate interatomic potentials enable molecular dynamics of materials, molecules, and interfaces beyond density-functional-theory length and time scales. Equivariant neural network potentials have improved the representation of local geometry. However, their deployable energy surfaces ultimately manifest through invariant scalar channels, whose aggregation and spectral resolution remain comparatively underexamined. Here we use Physics-Aware Neighborhood (PAN) pooling and Physics-Guided Spectral (PGS) mixers as controlled scalar-pathway probes: lightweight, symmetry-preserving modifications that act only on \\(\\ell=0\\) channels while leaving the equivariant tensor backbone unchanged. Using MACE as a high-body-order mechanistic scaffold, PAN adds coordination-sensitive amplitude modulation, whereas PGS augments edge and readout scalar features with radial and tapered spectral bases. Across metallic Ag, covalent Si, a short-range ionic LiF/Li--F subset, and MD17/rMD17 molecules, this scalar-pathway correction reduces MACE force errors by 22--27\\% and energy errors by 19--22\\%; on systems with stress labels, stress errors decrease by 27--28\\%, at approximately 5\\% additional inference-FLOPs cost. Directionally consistent gains in Allegro and NequIP further indicate that the correction is portable across distinct short-range equivariant backbones, although effect sizes remain architecture-dependent. These results identify scalar-pathway fidelity as a practical design dimension for short-range equivariant interatomic potentials.", "date": "2026-06-14", "categories": []}
{"id": "2606.16133", "title": "InvDesMobility: a reliability-gated first-principles feedback framework for closed-loop materials discovery", "abstract": "Inverse materials design starts from target functionality and searches for structures that can realize it. Its value in closed-loop discovery depends not only on prediction performance, but also on whether expensive first-principles results are independently validated, provenance-recorded, and admitted as feedback only when evidence is sufficient. This is especially important for composite properties such as carrier mobility, where a final scalar value hides intermediate quantities, fit quality, convergence history, and workflow assumptions. Here we present InvDesMobility, a reliability-gated first-principles feedback framework that integrates multi-agent automated DFT, evidence stratification, generative structure proposal, acquisition ranking, and auditable release. Using 516 2DMatPedia-derived candidates, the workflow produced 280 QC-passed materials and 573 retained carrier-direction seed channels after channel-level reliability gating. These records were split into two feedback objects: relaxed structures updated the generative model, while retained mobility channels trained the acquisition model and set validation priority. Over multiple iterations, InvDesMobility screened 2.4 x 10^6 structures, submitted 102 candidates for DFT validation, and retained 86 reliability-gated generated channels across 41 formulas. Overall, the main contribution is not a fixed list of high-mobility materials, but a transferable feedback contract that makes closed-loop inverse design both useful and auditable when learning from expensive calculated properties. All source data, retained feedback records, and workflows are available at https://github.com/DreamLufei/invDesMobility, with an accompanying evidence website at https://dreamlufei.github.io/invDesMobility/.", "date": "2026-06-15", "categories": []}
{"id": "2606.14003", "title": "XRDiff: Crystal Structure Prediction from Powder X-Ray Diffraction Data Using Diffusion Models", "abstract": "Determining the crystal structure of a material from its powder X-ray diffraction (PXRD) pattern is a central challenge in materials science. PXRD is an accessible and widely used characterization technique, yet recovering the atomic structure from diffraction data requires solving an underdetermined inverse problem due to the loss of phase information. Generative modeling can provide a prior over atomic structure and learn the mapping from PXRD patterns to crystal structures via simulated structure-spectrum pairs. We present XRDiff, a diffusion model that recovers crystal structures from PXRD given either the stoichiometry or, in a more challenging setting, the elemental constituents and total number of atoms in the unit cell. We evaluate on datasets where each stoichiometry has multiple polymorphs and all polymorphs of a given composition are held out together, ensuring that high performance reflects genuine use of the diffraction signal. XRDiff achieves strong structure recovery rates on simulated benchmarks, indicating that the model learns a spectrum-to-structure mapping precise enough to differentiate between polymorphs. To address generalization to experimental data, we compare a full-spectrum encoding against an encoding based on peak descriptors. The peak-based encoding generalizes substantially better, outperforming even a model trained on full spectra with augmentations fitted to the experimental noise distribution. These results demonstrate that representations robust to the noise and artifacts present in real-world PXRD offer a practical and scalable path toward closing the simulation-to-experiment gap, enabling zero-shot crystal structure solution from experimental PXRD with full or partial chemical composition input.", "date": "2026-06-12", "categories": []}
{"id": "2606.14215", "title": "LapidaryEngine: Fully Conversational Crystal Generation", "abstract": "The emergence of Large Language Models (LLMs) has inspired the vision of generating bespoke crystal materials directly from natural-language instructions, enabling users to design materials through intuitive, conversational interaction. Existing text-to-crystal generative models represent important early steps toward this goal, but they suffer from two critical limitations: (i) restricted input formats that require highly structured descriptions (e.g., chemical formulas), and (ii) one-directional generation, where models can map text to crystal but cannot perform the inverse. These limitations prevent fully conversational workflows and hinder alignment with users' inherently ambiguous and evolving desiderata. We address these challenges with LapidaryEngine, the first model to support fully conversational crystal generation. LapidaryEngine accepts free-form natural-language requests and performs iterative refinement and editing in a dialogue-like manner. The key innovation is a pivot representation, a third, intermediate form that enables bidirectional translation between text and crystal structures despite the absence of direct paired datasets. Leveraging this pivot allows robust interpretation of user feedback and precise structural control. We demonstrate LapidaryEngine across diverse tasks, including insulator discovery, stability optimization, compositional modification, and structural editing, showcasing its ability to align generated materials with user intent in an interactive manner.", "date": "2026-06-12", "categories": []}
{"id": "2606.14565", "title": "CANN-EUCLID: unsupervised constitutive artificial neural network model discovery from full-field data", "abstract": "Constitutive artificial neural networks (CANNs) provide interpretable material model discovery, but have so far been used in stress-supervised settings based on apparent stress-strain data from homogeneous tests. Because each test samples only a narrow loading path and provides homogenized rather than local stress information, robust discovery typically requires multiple loading modes to constrain the multidimensional response. This is challenging for soft biological tissues, where repeated testing, damage, and sample variability limit reliable information from a single specimen. Here, we combine CANNs with the stress-unsupervised full-field discovery framework EUCLID to identify sparse hyperelastic laws directly from displacement fields and reaction forces in one heterogeneity-inducing loading case. CANN-EUCLID minimizes equilibrium imbalance with sparsity-promoting regularization selecting compact active terms, without local stress measurements or a prescribed law. We evaluate the approach on isotropic and anisotropic benchmarks with prescribed ground-truth laws. When the ground truth is representable by the chosen CANN basis, our method recovers the correct terms with near-exact accuracy, including exponential terms with embedded parameters. When it is not contained in the basis, the method retains shared terms and approximates missing contributions using available basis functions. Generalization depends strongly on sampled deformation states: exponential strain-stiffening terms can be recovered accurately when sufficiently probed, but can produce large extrapolation errors when the stiffening regime lies outside the sampled domain. Forward FE validation simulations show that the discovered behavior accurately replicates the ground truth. These results establish stress-unsupervised CANN discovery as a promising framework for interpretable full-field constitutive model identification.", "date": "2026-06-12", "categories": []}
{"id": "2412.13012", "title": "Deep Learning Based Superconductivity: Prediction and Experimental Tests", "abstract": "The discovery of novel superconducting materials is a longstanding challenge in materials science, with a wealth of potential for applications in energy, transportation, and computing. Recent advances in artificial intelligence (AI) have enabled expediting the search for new materials by efficiently utilizing vast materials databases. In this study, we developed an approach based on deep learning (DL) to predict new superconducting materials. We have synthesized a compound derived from our DL network and confirmed its superconducting properties in agreement with our prediction. Our approach is also compared to previous work based on random forests (RFs). In particular, RFs require knowledge of the chemical properties of the compound, while our neural net inputs depend solely on the chemical composition. With the help of hints from our network, we discover a new ternary compound $\\textrm{Mo}_{20} \\textrm{Re}_{6} \\textrm{Si}_{4}$, which becomes superconducting below 5.4 K. We further discuss the existing limitations and challenges associated with using AI to predict and, along with potential future research directions.", "date": "2024-12-17", "categories": []}
{"id": "2509.07150", "title": "PLaID++: A Preference Aligned Language Model for Targeted Inorganic Materials Design", "abstract": "Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as a promising approach to improve correctness in LLMs, however, in many scientific problems, the objective is not necessarily to produce the correct answer, but instead to produce a diverse array of candidates which satisfy a set of constraints. We study this challenge in the context of materials generation. To this end, we introduce PLaID++, an LLM post-trained for stable and property-guided crystal generation. We find that performance hinges on our crystallographic representation and reward formulation. First, we introduce a compact, symmetry-informed Wyckoff text representation which improves computational efficiency and encourages generalization from physical priors. Second, we demonstrate that temperature scaling acts as an entropy regularizer which counteracts mode collapse and encourages exploration. By encoding symmetry constraints directly into text and guiding model outputs towards desirable chemical space, PLaID++ generates structures that are thermodynamically stable, unique, and novel at a $\\sim$50\\% greater rate than prior methods and conditionally generates structures with desired space group properties. Our work demonstrates the potential of adapting post-training techniques from natural language processing to materials design, paving the way for targeted and efficient discovery of novel materials.", "date": "2025-09-08", "categories": []}
{"id": "2512.21227", "title": "PhononBench:A Large-Scale Phonon-Based Benchmark for Dynamical Stability in Crystal Generation", "abstract": "In recent years, generative artificial intelligence has made significant advances in the design of crystalline materials, giving rise to approaches based on graph neural networks, diffusion models, and large language models. Existing evaluations commonly follow the stability-uniqueness-novelty (S.U.N.) framework, where stability is primarily assessed using thermodynamic criteria, which do not fully capture the dynamical stability essential for a material's practical existence. Dynamical stability is a key determinant of whether a material can be synthesized and persist, with phonon spectrum calculations serving as the standard for its evaluation. However, the high computational cost of such calculations has prevented large-scale assessment of dynamical stability in generated crystals. In this work, we introduce PhononBench, the first large-scale benchmark for dynamical stability in AI-generated crystals. Leveraging the recently developed MatterSim interatomic potential, which achieves density-functional-theory (DFT)-level accuracy in phonon predictions across more than 10,000 materials, PhononBench enables efficient phonon calculations and dynamical-stability analysis for 133,838 crystal structures generated by 7 leading crystal generation models. PhononBench reveals a widespread limitation of current generative models: unless otherwise specified, all reported dynamical-stability metrics are evaluated at a phonon-frequency threshold of -0.1 THz, with the average dynamical-stability rate across all generated structures being only 32.15%, and the top-performing model, MatterGen, reaching just 45.05%.In addition, we identify 32,995 crystal structures that are phonon-stable across the entire Brillouin zone under a strict threshold of -0.001 THz. In addition, a web-based service is accessible at http://phononbench.cn/, enabling minute-level ultra-fast phonon predictions.", "date": "2025-12-24", "categories": []}
{"id": "2602.04075", "title": "Thermodynamic assessment of machine learning models for solid-state synthesis prediction", "abstract": "Machine learning models have recently emerged to predict whether hypothetical solid-state materials can be synthesized. These models aim to circumvent direct first-principles modeling of solid-state phase transformations, instead learning from large databases of successfully synthesized materials. Here, we assess the alignment of several recently introduced synthesis prediction models with material and reaction thermodynamics, quantified by the energy with respect to the convex hull and a metric accounting for thermodynamic selectivity of enumerated synthesis reactions. A dataset of successful synthesis recipes was used to determine the likely bounds on both quantities beyond which materials can be deemed unlikely to be synthesized. With these bounds as context, thermodynamic quantities were computed using the CHGNet foundation potential for thousands of new hypothetical materials generated using the Chemeleon generative model. Four recently published machine learning models for synthesizability prediction were applied to this same dataset, and the resultant predictions were considered against computed thermodynamics. We find these models generally overpredict the likelihood of synthesis, but some model scores do trend with thermodynamic heuristics, assigning lower scores to materials that are less stable or do not have an available synthesis recipe that is calculated to be thermodynamically selective. In total, this work identifies existing gaps in machine learning models for materials synthesis and introduces a new approach to assess their quality in the absence of extensive negative examples (failed syntheses).", "date": "2026-02-03", "categories": []}
{"id": "2602.00424", "title": "Open Materials Generation with Inference-Time Reinforcement Learning", "abstract": "Continuous-time generative models for crystalline materials enable inverse materials design by learning to predict stable crystal structures, but incorporating explicit target properties into the generative process remains challenging. Policy-gradient reinforcement learning (RL) provides a principled mechanism for aligning generative models with downstream objectives but typically requires access to the score, which has prevented its application to flow-based models that learn only velocity fields. We introduce Open Materials Generation with Inference-time Reinforcement Learning (OMatG-IRL), a policy-gradient RL framework that operates directly on the learned velocity fields and eliminates the need for the explicit computation of the score. OMatG-IRL leverages stochastic perturbations of the underlying generation dynamics preserving the baseline performance of the pretrained generative model while enabling exploration and policy-gradient estimation at inference time. Using OMatG-IRL, we present the first application of RL to crystal structure prediction (CSP). Our method enables effective reinforcement of an energy-based objective while preserving diversity through composition conditioning, and it achieves performance competitive with score-based RL approaches. Finally, we show that OMatG-IRL can learn time-dependent velocity-annealing schedules, enabling accurate CSP with order-of-magnitude improvements in sampling efficiency and, correspondingly, reduction in generation time. The OMatG-IRL code is included in a new release of the Open Materials Generation (OMatG) framework available at https://github.com/FERMat-ML/OMatG.", "date": "2026-01-31", "categories": []}
{"id": "2602.10392", "title": "Tensor Methods: A Unified and Interpretable Approach for Material Design", "abstract": "When designing new materials, it is often necessary to tailor the material design to have some desired properties. As the set of design parameters grow, the search space grows exponentially, making the actual synthesis and evaluation of all material combinations virtually impossible. Even using traditional computational methods such as Finite Element Analysis becomes too computationally heavy to search the design space. Recent methods use machine learning (ML) surrogate models to more efficiently determine optimal material designs; unfortunately, these methods often (i) are notoriously difficult to interpret and (ii) under perform when the training data comes from a non-uniform sampling of the design space. We suggest the use of tensor completion methods as an all-in-one approach for interpretability and predictions. We observe classical tensor methods are able to compete with traditional ML in predictions, with the added benefit of their interpretable tensor factors (which are given completely for free, as a result of the prediction). In our experiments, we are able to rediscover physical phenomena via the tensor factors, indicating that our predictions are aligned with the true underlying physics of the problem. This also means these tensor factors could be used by experimentalists to identify potentially novel patterns, given we are able to rediscover existing ones. We also study the effects of both types of surrogate models when we encounter training data from a non-uniform sampling of the design space. We observe more specialized tensor methods that can give better generalization in these non-uniforms sampling scenarios. We find the best generalization comes from a tensor model, which is able to improve upon the baseline ML methods by up to 5% on aggregate $R^2$, and halve the error in some out of distribution regions.", "date": "2026-02-11", "categories": []}
{"id": "2606.11870", "title": "Modelling magnetic material properties with uncertainty-aware neural networks", "abstract": "Machine learning is increasingly applied to accelerate the discovery of novel materials by exploring large compositional and structural design spaces. Yet, the scarcity of high-quality data and the frequent need for out-of-distribution prediction introduce substantial uncertainty, making the assessment of model reliability essential. In this work, we investigate uncertainty quantification as a means to evaluate model confidence in the context of permanent magnet research. In a first study, we benchmark classical and modern machine learning models for predicting intrinsic magnetic properties, focusing on the quality of their uncertainty estimates. We apply Gaussian negative log-likelihood loss and dropout-based Bayesian approximation as practical strategies for estimating predictive uncertainty. In a second study, we transfer these architectural features for uncertainty estimation to a more complex task: predicting coercivity from microstructural information using a graph neural network. Together, these studies demonstrate that uncertainty quantification not only enhances the trustworthiness of predictions but is also transferable across different modeling tasks.", "date": "2026-06-10", "categories": []}
{"id": "2601.02424", "title": "A large-scale nanocrystal database with aligned synthesis and properties enabling generative inverse design", "abstract": "The synthesis of nanocrystals has been highly dependent on trial-and-error, due to the complex correlation between synthesis parameters and physicochemical properties. Although deep learning offers a potential methodology to achieve generative inverse design, it is still hindered by the scarcity of high-quality datasets that align nanocrystal synthesis routes with their properties. Here, we present the construction of a large-scale, aligned Nanocrystal Synthesis-Property (NSP) database and demonstrate its capability for generative inverse design. To extract structured synthesis routes and their corresponding product properties from literature, we develop NanoExtractor, a large language model (LLM) enhanced by well-designed augmentation strategies. NanoExtractor is validated against human experts, achieving a weighted average score of 88% on the test set, significantly outperforming chemistry-specialized (3%) and general-purpose LLMs (38%). The resulting NSP database contains nearly 160,000 aligned entries and serves as training data for our NanoDesigner, an LLM for inverse synthesis design. The generative capability of NanoDesigner is validated through the successful design of viable synthesis routes for both well-established PbSe nanocrystals and rarely reported MgF2 nanocrystals. Notably, the model recommends a counter-intuitive, non-stoichiometric precursor ratio (1:1) for MgF2 nanocrystals, which is experimentally confirmed as critical for suppressing byproducts. Our work bridges the gap between unstructured literature and data-driven synthesis, and also establishes a powerful human-AI collaborative paradigm for accelerating nanocrystal discovery.", "date": "2026-01-04", "categories": []}
{"id": "2601.09285", "title": "Enhancing Spatial Reasoning in Large Language Models for Metal-Organic Frameworks Structure Prediction", "abstract": "Metal-organic frameworks (MOFs) are porous crystalline materials with broad applications such as carbon capture and drug delivery, yet accurately predicting their 3D structures remains a significant challenge. While Large Language Models (LLMs) have shown promise in generating crystal structures, their application to MOFs is hindered by MOFs' high structural complexity arising from the large number of atoms in unit cell. Inspired by the success of block-wise paradigms in deep generative models for MOFs, we pioneer the application of LLMs in this domain by introducing MOF-LLM, the first LLM framework specifically adapted for block-level MOF structure prediction. To effectively harness LLMs for this 3D modular assembly task, our training paradigm integrates spatial-aware continual pre-training (CPT), structural supervised fine-tuning (SFT), and matching-driven reinforcement learning (RL). By incorporating explicit spatial priors and optimizing structural stability via Soft Adaptive Policy Optimization (SAPO), our approach substantially enhances the spatial reasoning in a Qwen-3 8B model for MOF structure prediction. Comprehensive experiments demonstrate that MOF-LLM achieves state-of-the-art performance with a match rate of 35.78% while exhibiting superior sampling efficiency of 0.04 seconds per structure.", "date": "2026-01-14", "categories": []}
{"id": "2602.14975", "title": "Faster Molecular Dynamics with Neural Network Potentials via Distilled Multiple Time-Stepping and Non-Conservative Forces", "abstract": "Following our previous work (J. Phys. Chem. Lett., 2026, 17, 5, 1288-1295), we propose the DMTS-NC approach, a distilled multi-time-step (DMTS) strategy using non-conservative (NC) forces to further accelerate atomistic molecular dynamics simulations using foundation neural network models such as FeNNix-Bio1. There, a dual-level reversible reference system propagator algorithm (RESPA) formalism couples a target accurate conservative potential to a simplified distilled representation optimized for the production of non-conservative forces. Despite being non-conservative, the distilled architecture is designed to enforce key physical priors, such as equivariance under rotation and cancellation of atomic force components. These choices facilitate the distillation process and therefore improve drastically the robustness of simulation, significantly limiting abnormal discrepancies between the two models, thus achieving excellent agreement with the forces data. Overall, the DMTS-NC scheme is found to be more stable and efficient than its conservative counterpart with additional speedups reaching 15-30% over DMTS. Requiring no fine-tuning steps, it is easier to implement and can be pushed to the limit of the systems physical resonances to maintain accuracy while providing maximum efficiency. We obtain additional speedup by combining hydrogen mass repartitioning (HMR), High Hydrogen Friction (HHF) to further extended the largest timestep up to 10fs of our schemes while conserving stability and accuracy. As for DMTS, DMTS-NC is applicable to any neural network potential and can be applied to approaches that are computationally heavier than FeNNix-Bio1. We show a proof of principle applying the approach to the distillation of MACE-OFF23 with consequent speedups ranging from 3.66 to 5.64 compared to single timestep.", "date": "2026-02-16", "categories": []}
{"id": "2606.07712", "title": "MatMind: A Structure-Activity Knowledge-Driven Generative Foundation Model for Materials Science", "abstract": "Progress in AI-driven crystal materials science has so far been carried by narrow architectures purpose-built for individual tasks -- graph neural networks for property prediction, diffusion and flow-matching models for crystal generation -- each excelling within its niche yet unable to act as a shared backbone across the full spectrum of materials problems. Generative large language models offer a fundamentally different paradigm, in which structural representation, quantitative prediction, and structure-activity reasoning can be unified within one model, but the materials community has yet to see this paradigm realized at a level competitive with established narrow specialists. Here we present MatMind, a generative foundation model purpose-built for crystal materials science under this paradigm, developed through the coordinated activation of structure-activity knowledge and physics-informed feedback within a progressive training framework -- combining structure-activity knowledge injection, a dual-head architecture that jointly trains language reasoning and numerical regression in a shared representation space, and multi-objective physics-informed reinforcement learning over stability, novelty, and structural diversity. Across three task families, MatMind attains the lowest mean absolute error on energy above hull, bulk modulus, and band gap -- surpassing graph neural network predictors purpose-built for these tasks -- reaches an S.U.N. rate of 65.3% on unconditional crystal generation, and achieves a comparable multiplicative improvement on magnetization-density-conditioned generation, where only 21 positive samples exist within over 600000 training entries. By matching or surpassing narrow specialists on their own ground while operating within a single unified model, MatMind shows that the LLM-based paradigm can serve as a viable backbone for crystal materials science going forward.", "date": "2026-06-05", "categories": []}
{"id": "2606.07836", "title": "Agentic multi-fidelity learning of quasiparticle and excitonic properties", "abstract": "Many-body GW-Bethe-Salpeter equation calculations are essential for accurate simulations of electronic structure and optical properties in modern low-dimensional nanomaterials. However, these methods are computationally demanding and can exhibit localized numerical instabilities or convergence failures that are difficult to detect within high-throughput workflows. We introduce an agent-guided multi-fidelity framework for correcting GW-Bethe-Salpeter excited-state landscapes in strained MoS2-WS2 bilayers. Across stacking registries, strain branches and reciprocal-space samplings, the workflow identifies spike-like excursions, near-zero-gap collapse and cross-fidelity inconsistencies associated with fragile long-wavelength dielectric screening. A structural agent evaluates calculations by assigning confidence weights and selectively using a small number of high-accuracy reference points. Machine learning models then transfer information across related systems and apply Gaussian process corrections to recover improved quasiparticle gaps and exciton binding energies, with calibrated uncertainty estimates. The approach corrects numerically induced artifacts without erasing physical strain dependence and substantially improves agreement with higher-fidelity references relative to a no-agent baseline. These results show that reliable surrogate learning for excited-state materials requires explicit diagnosis of numerical fragility, not direct interpolation of raw first-principles data points. The proposed framework is readily transferable to other optoelectronic nanomaterials characterized by strong quantum confinement, such as quantum dots, nanoribbons, layered two-dimensional semiconductors, and hybrid perovskite nanostructures.", "date": "2026-06-05", "categories": []}
{"id": "2606.08148", "title": "Inverse design of bespoke interatomic potentials via active learning by information-matching", "abstract": "Interatomic potentials (IPs) enable large-scale atomistic simulations beyond the reach of first-principles methods, but their predictive reliability depends critically on the selection of training data, quantified uncertainty, and model expressiveness. Active learning (AL) provides a principled framework for constructing efficient and accurate IPs, yet most strategies reduce parameter uncertainty without explicitly accounting for the specific material properties being predicted. The information-matching (IM) approach addresses this limitation by requiring that the selected training data provide at least as much parameter space information as needed to achieve prescribed uncertainty targets for selected quantities of interest (QoIs). Here, we apply IM to develop bespoke IPs specifically tailored for predicting plastic strength in metals. Due to the high computational cost of simulating plastic strength, we employ an indirect IM strategy that targets inexpensive intermediate QoIs that correlate with strength. The IM method enables precise parameter constraints with minimal training data, yielding precise predictions for both the intermediate QoIs and plastic strength. Yet, model error remains a key limitation, and a post hoc uncertainty inflation correction provides a viable means to mitigate this limitation. These findings illustrate both the promise and limits of uncertainty-aware AL for predicting complex material properties.", "date": "2026-06-06", "categories": []}
{"id": "2606.08150", "title": "Property-Informed Diffusion-Based Text-to-Microstructure Generation", "abstract": "Designing 3D metamaterial microstructures that meet the intended functions remains a major challenge, as it typically requires domain expertise, iterative simulations, and extensive manual tuning. Existing work on inverse design that automatically generates microstructures based on desired target properties often suffers from limited design diversity and faces challenges in ensuring the physical feasibility of the generated structures. To address this issue, a property-informed diffusion-based network is proposed that enables the generation of 3D microstructures directly from textual descriptions. Unlike traditional property conditioning methods, our approach leverages rich guidance in terms of semantics and physical properties in the text input to support diverse structure synthesis. To enforce consistency between the generated structures and the target textual prompts, a dual alignment strategy is adopted, including contrastive text-structure alignment and test-time reward-guided alignment. Experimental results show that the model is capable of generating semantically meaningful and physically plausible structures across a wide range of material categories. Our approach has good potential for interactive microstructure design and opens up new directions for combining language-based interfaces with inverse material discovery. Code is available at: https://github.com/hongsong-wang/PropDiff-TMG", "date": "2026-06-06", "categories": []}
{"id": "2606.09266", "title": "Physics-Guided Sequence-Based Generative Framework for Acoustic Metamaterial Inverse Design", "abstract": "Acoustic metamaterial (AMM) inverse design is particularly challenging for broadband target responses due to acoustic dispersion: a structure that matches the desired response at one frequency may deviate at others, and modifying geometry to improve one sub-band often perturbs neighboring sub-bands. Yet existing broadband inverse-design approaches are either constrained by predefined templates, or rely on image representations that fail to preserve the geometric precision and structural connectivity required by acoustic structures. We present MetaSeq, a physics-guided, sequence-based generative framework for acoustic metamaterial inverse design. At its core, MetaSeq introduces a language that represents each AMM as a structured sequence, rather than as a pixel grid or fixed template. This representation preserves precise geometry, explicitly encodes connectivity, and casts inverse design as a sequence-to-sequence task from target response to structure sequence. MetaSeq further constructs a balanced, high-fidelity dataset with efficient calibration and complexity-based sampling. To address the one-to-many nature of inverse design, MetaSeq combines supervised pretraining with reinforcement learning fine-tuning guided by a physics-based solver and validity checker. Extensive evaluations against COMSOL and five baselines show that MetaSeq reduces response error by 45% over the best baseline.", "date": "2026-06-08", "categories": []}
{"id": "2510.16023", "title": "A Conformation-Centric Generative Foundation Model for Linear Polymer Modeling and Design", "abstract": "Linear polymers, macromolecules formed from monomers covalently bonded into continuous chains, underpin countless technologies and are indispensable to modern life. While deep learning is advancing polymer science, existing methods typically represent the whole linear polymer solely through monomer-level descriptors, overlooking the global structural information inherent in polymer conformations, which ultimately limits their practical performance. Moreover, this important field still lacks a dedicated foundation model that can effectively support diverse downstream tasks, thereby severely constraining progress. To address these challenges, we introduce PolyConFM, a foundation model tailored for modeling and designing linear polymers through conformation-centric generative pretraining. Recognizing that each linear polymer is essentially a continuous chain whose conformation can be naturally decomposed into a sequence of local conformations (i.e., those of its repeating units), we pretrain PolyConFM under the conditional generation paradigm, reconstructing these local conformations via masked autoregressive (MAR) modeling and further generating their orientation transformations to recover the corresponding polymer conformation. Meanwhile, we construct a linear polymer conformation dataset via molecular dynamics simulations to mitigate data sparsity, thereby enabling conformation-centric pretraining. Experiments demonstrate that PolyConFM consistently outperforms representative task-specific methods across diverse downstream tasks, thereby equipping polymer science with a powerful tool targeting linear polymers.", "date": "2025-10-15", "categories": []}
{"id": "2602.16908", "title": "Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials", "abstract": "Allegro is a machine learning interatomic potential model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason, we apply multi-objective hyperparameter optimization to both objectives. Additionally, we experiment with modified architectures by constructing variants of Allegro: one extended with additional classical layers and one incorporating quantum-classical hybrid layers. We evaluate all models on QM9, rMD17-aspirin, rMD17-benzene, and a self-generated dataset of copper-lithium structures. As results, both variants surpass Allegro in force prediction accuracy across multiple datasets. The classical variant consistently improves over the baseline, while the quantum-classical hybrid variant achieves the best overall force prediction accuracy on the Cu-Li dataset, where it was fully optimized, outperforming the classical variant by approximately 13%. Notably, the hybrid variant also achieves competitive results on the remaining datasets despite using hyperparameters transferred from Cu-Li without dataset-specific optimization, suggesting that quantum-classical hybridization is a promising direction for enhancing MLIP architectures.", "date": "2026-02-18", "categories": []}
{"id": "2606.05050", "title": "Autonomous heterogeneous catalyst discovery with a self-evolving multi-agent digital twin", "abstract": "Theoretical heterogeneous catalysis promises rapid catalyst discovery, yet computational and machine-learning predictions often deviate from experiment and stay confined to narrow material families, for want of a faithful, condition-aware catalytic simulator. We present CatDT (Catalysis Digital Twin), a self-evolving multi-agent system that builds an autonomous digital twin of a working catalyst, unifying gas-solid and liquid-solid modeling. From only a bulk crystal and a natural-language reaction description, eight specialized agents and 27 scientific tools predict stable facets, reconstruct working surfaces, enumerate and rank reaction pathways, locate transition states, and compute kinetics in 5-30 min on a single GPU. Two innovations address the hardest steps: UniMech finds dominant pathways for novel materials at over $10^3\\times$ lower cost than exhaustive enumeration by fusing agent-guided proposals with energy-cached graph search, and a memory-augmented reinforcement loop raises barrier-calculation success from 41\\% to 84\\% across 600 catalytic surfaces. Across seven gas-solid benchmarks -- stepped metals, single-atom catalysts, ordered intermetallics, vacancy-rich 2D sulfides and carbides, and a strong-metal--support-interaction (SMSI) interface -- every CatDT prediction lies within 0.5-2 times experiment over four orders of magnitude. For propane dehydrogenation, CatDT independently discovers non-precious candidates rivaling the Pt-based industrial benchmark, with a proposed Ni@ZrO$_2$ SMSI overlayer reaching a simulated TOF of $1.63~\\text{s}^{-1}$ at $\\sim$100\\% selectivity. More broadly, the decisive factor for a faithful catalyst digital twin -- or any multi-stage scientific simulator -- is not raw LLM capability but the engineered harness around it: deterministic tools, persistent memory, and verified self-improvement that compound across models, tools, and runs.", "date": "2026-06-03", "categories": []}
{"id": "2606.06765", "title": "Reactivity-Informed Machine Learning for Performance Prediction and Design Space Exploration of Alkali-Activated Slag", "abstract": "Establishing quantitative relationships among mix design, raw material properties, curing conditions, and performance remains a long-standing challenge in cementitious materials, particularly for alkali-activated materials with variable precursor and activator chemistry. Here, we curated the largest literature-derived alkali-activated slag (AAS) dataset to date, comprising over 3100 compressive strength records, 155 chemically distinct ground granulated blast-furnace slags (GGBSs), and 24 attributes incorporating precursor chemistry, fineness, and reactivity. Multiple machine learning (ML) algorithms were benchmarked across progressively enriched feature scenarios, demonstrating that integrating GGBS compositions, fineness, curing conditions, and specimen geometry improves predictive performance. The average metal oxide dissociation energy (AMODE), a physically interpretable representation of precursor reactivity, provides a compact alternative descriptor to explicit oxide compositions while enabling comparable predictive performance. Model interpretation revealed physically consistent trends from heterogeneous data, including non-monotonic effects of Na2O dosage and silicate modulus, reduced predicted strength at higher water content and larger specimen size, and coupled oxide-level effects more coherently represented by AMODE than by individual oxide contents. Statistically constrained design space exploration reveals reactivity-dependent trade-offs among strength, embodied CO2 emissions, and cost. The design maps identify high-strength regions with substantially lower CO2 emissions than OPC-based references at similar cost. Overall, this work demonstrates how reactivity-informed ML can extract physically meaningful trends from heterogeneous AAS data and guide source-dependent binder design. The curated dataset is publicly accessible to support advances in cement and concrete research.", "date": "2026-06-04", "categories": []}
{"id": "2605.26179", "title": "AutoDFT: A Closed-Loop Multi-Agent Framework for Autonomous DFT Calculations", "abstract": "Density functional theory (DFT) serves as the basis for computational discovery in materials science and chemistry, yet each calculation demands extensive human effort: adjusting algorithms when convergence stalls, revising plans when unexpected physics emerges, and inserting steps as intermediate results reshape the problem. Existing LLM-based agents automate only the initial planning stage, producing a full execution plan upfront and leaving all subsequent adaptation to hand-crafted rules. As a result, these workflows remain fragile, do not generalize well beyond pre-planned scenarios, and often require expert intervention when failures or unexpected intermediate results require changes to the calculation path. Here, we introduce AutoDFT, a closed-loop multi-agent framework that embeds LLM reasoning into every stage of the DFT lifecycle, where a strategic planner produces a skeletal plan of step objectives; a step planner generates numerical parameters just in time from preceding results; and a monitor-recover-reflect cycle diagnoses failures, repairs them, and revises the plan when the evidence justifies it. We demonstrate both breadth and depth: breadth on VASPBench, a purpose-built benchmark spanning 34 tasks and 9 DFT calculation types, where AutoDFT achieves 94.1% task-level success with GPT-5.2; and depth on established materials databases, where AutoDFT produces quantitatively reliable property predictions across electronic, magnetic, and energetic properties. By closing the loop between planning and execution, AutoDFT enables experimentalists without deep computational expertise to obtain reliable first-principles results.", "date": "2026-05-25", "categories": []}
{"id": "2606.05199", "title": "Finite Element-Based Material Learning via Automatic Differentiation: Learning constitutive neural network models from full-field deformation data", "abstract": "The identification of constitutive neural network models from heterogeneous full-field deformation data provides a robust alternative to traditional calibration methods based on homogeneous stress-strain experiments, particularly given the high dimensionality of trainable parameters. Existing approaches must balance generality, robustness, and computational efficiency: Conventional finite element model updating is broadly applicable but computationally demanding; weak-form methods offer efficiency but are sensitive to noise and data scarcity; neural operator models are highly expressive but require extensive training datasets. This work presents FE-MAD (Finite Element-Based Material learning via Automatic Differentiation), an end-to-end differentiable framework that integrates a constitutive neural network model within a JAX-FEM nonlinear solver and identifies its parameters through gradient-based minimization of a measurement-mismatch loss. Newton tangent stiffness and loss gradients are computed automatically using forward- and reverse-mode automatic differentiation throughout the entire pipeline, thereby removing the need for analytic adjoints or offline surrogate models. FE-MAD is demonstrated for two architectures: a grey-box Constitutive Artificial Neural Network (CANN), a polyconvex, fully connected model with high flexibility, and a white-box CANN, an expert-system network with phenomenologically interpretable strain-energy terms. Focusing on incompressible isotropic hyperelasticity, FE-MAD is evaluated on three open experimental datasets: (1) full digital image correlation (DIC) of a perforated tensile specimen, (2) a reduced-data scenario with a one-dimensional stretch profile and global force-displacement curve, and (3) a heterogeneous matrix-inclusion system in which both phases constitutive laws are identified and generalized to twenty-two previously unseen samples.", "date": "2026-05-22", "categories": []}
{"id": "2606.04000", "title": "SPLIT-PINN: Separable Probability Learning Technique via Physics-Informed Neural Networks for High-Dimensional Probabilistic Modeling", "abstract": "We present a probabilistic modeling framework for incorporating small-scale spatial heterogeneity into macroscopic descriptions of material behavior for polycrystalline metallic materials. Spatially heterogeneous material state fields are represented using probability density functions (PDFs), providing a principled statistical description of microstructural variability and state evolution across different computational polycrystalline realizations. The framework is built on the inverse identification of a probabilistic transport model, formulated as a Liouville equation with an unknown drift term. To enable accurate, stable, and interpretable inference of this drift field in high-dimensional, transport-dominated settings, we develop a Separable Probability Learning Technique via Physics-Informed Neural Networks (SPLIT-PINN). This method incorporates a marginal-correction drift decomposition, orthogonality constraints, and residual-based adaptive training to enhance well-posedness, numerical stability, and physical consistency without imposing restrictive parametric assumptions. Using SPLIT-PINN, the drift field governing the temporal evolution of joint state PDFs is inferred directly from data. After benchmark validation, the framework is applied to physical computational datasets describing the evolution of polycrystalline microstructural states, including von Mises stress, dislocation density, and equivalent plastic strain rate. The learned Liouville model, trained on a single dataset, is subsequently used in forward predictions of the temporal evolution of joint and marginal PDFs for multiple unseen polycrystal realizations. Quantitative comparisons with reference PDFs demonstrate that the proposed framework yields accurate and robust probabilistic predictions and generalizes effectively across datasets.", "date": "2026-05-23", "categories": []}
{"id": "2606.04100", "title": "Stein Kernelized Molecular Dynamics for Active Learning of Interatomic Potentials", "abstract": "Machine learning interatomic potentials (MLIPs) enable efficient and accurate atomistic simulations but depend critically on the quality and diversity of the training data. We introduce Stein kernelized molecular dynamics (SKMD), an enhanced sampling method that uses interacting particle dynamics to acquire informative training configurations for the active learning and fine-tuning of MLIPs. SKMD corresponds to a stochastic variant of Stein variational gradient descent that is adapted for molecular dynamics by incorporating asynchronous particle updates and a kernel of global atomic descriptors, which provides a symmetry-aware measure of configurational similarity. Unlike other enhanced samplers used in molecular dynamics, SKMD preserves the Boltzmann distribution as the asymptotic distribution of the dynamics. This property enforces a balance between the exploration of diverse configurations and attraction toward high-probability regions of the energy landscape. We further propose an approach to efficient online data acquisition using an adaptive stopping criterion that selects non-redundant training data over the course of simulation. We demonstrate SKMD for the active learning of a neural network model of the M\\\"uller-Brown potential and the fine-tuning of a MACE interatomic potential for alanine dipeptide. Compared to active learning baselines, our method achieves higher model accuracy in fewer training iterations with the same number of acquired training samples.", "date": "2026-06-02", "categories": []}
{"id": "2512.16882", "title": "A Cartesian-3j Framework for Machine Learning Interatomic Potentials", "abstract": "Machine learning interatomic potentials (MLIPs) have brought substantial gains in the extrapolation capability in computational chemistry. However, most equivariant models are typically built with spherical tensors (STs), while Cartesian tensor formulations remain less developed despite their natural alignment with atomic coordinates and tensorial targets. In this work, we develop a Cartesian framework for irreducible Cartesian tensors (ICTs) by introduce the \\texttt{Cartesian-3j} symbol and Cartesian Generalized Clebsch-Gordan Coefficients, which serve as direct analogues of the \\texttt{Wigner-3j} symbol and Generalized Clebsch-Gordan coefficients defined for ST coupling. We extend the \\texttt{e3nn} library to support ICT product, and use this framework to build Cartesian counterparts of \\texttt{MACE}, \\texttt{NequIP}, and \\texttt{Allegro}, allowing the first controlled comparison where architectures are held fixed and only the tensor basis is changed. Our experiments show that irreducible Cartesian models can achieve accuracy comparable to spherical counterparts, but direct Cartesianization incurs unfavorable compute and memory scaling, motivating dedicated Cartesian architectural choices. Leveraging ICTs and our framework, we introduce \\texttt{TACE-v1-OAM-M} and demonstrate that it achieves competitive performance on Matbench Discovery compared to state-of-the-art ST models.", "date": "2025-12-18", "categories": []}
{"id": "2604.13354", "title": "Finetuning-Free Diffusion Model with Adaptive Constraint Guidance for Inorganic Crystal Structure Generation", "abstract": "The discovery of inorganic crystal structures with targeted properties is a significant challenge in materials science. Generative models, especially state-of-the-art diffusion models, offer the promise of modeling complex data distributions and proposing novel, realistic samples. However, current generative AI models still struggle to produce diverse, original, and reliable structures of experimentally achievable materials suitable for high-stakes applications.\n  In this work, we propose a generative machine learning framework based on diffusion models with adaptive constraint guidance, which enables the incorporation of user-defined physical and chemical constraints during the generation process. This approach is designed to be practical and interpretable for human experts, allowing transparent decision-making and expert-driven exploration. To ensure the robustness and validity of the generated candidates, we introduce a multi-step validation pipeline that combines graph neural network estimators trained to achieve DFT-level accuracy and convex hull analysis for assessing thermodynamic stability. Our approach has been tested and validated on several classical examples of inorganic families of compounds, as case studies. As a consequence, these preliminary results demonstrate our framework's ability to generate thermodynamically plausible crystal structures that satisfy targeted geometric constraints across diverse inorganic chemical systems.", "date": "2026-04-14", "categories": []}
{"id": "2606.03199", "title": "Fast Organic Crystal Structure Prediction with Unit Cell Flow Matching", "abstract": "Organic crystal structure prediction (CSP) is a requirement for computational modelling of organic solids, but traditionally costs several CPU-years per molecule. Generative models such as OXtal dramatically reduce this cost by sampling stable organic crystal structures directly. However, OXtal forgoes explicit lattice parametrization in favour of modelling large crops of the bulk material with expensive triangle layers, which can incur a computational cost of minutes per molecule. In this paper, we reduce this to seconds with Clari, a large-scale flow matching model that generates redundancy-free unit cells and replaces triangle layers with pure pair-bias attention. Clari requires only atom types and bonds as input and does not need an RDKit-sanitizable input molecule, which expands its applicability to challenging chemistries such as fullerenes, metal complexes, and atom clusters. We further ablate key design choices such as auxiliary losses, timestep distributions, noise priors, and self-conditioning. On OXtal's test sets, we surpass OXtal's solve rate while obtaining a speedup of $15$-$30\\times$. Because Clari also models explicit hydrogens, it supports inference-time scaling via direct energy ranking, without any decoration or relaxation step. When generating 150 crystals and selecting the top-30 by energy, we further improve solve rate while maintaining a speedup of $5$-$8\\times$. We also introduce the CSD Teaching Subset as a new test split of diverse and complex molecules for future benchmarking. Our contributions enable CSP within seconds, making large-scale virtual screening of organic solids practical. Code is available at https://github.com/aspuru-guzik-group/clari.", "date": "2026-06-02", "categories": []}
{"id": "2601.07742", "title": "PFT: Phonon Fine-tuning for Machine Learned Interatomic Potentials", "abstract": "Many materials properties depend on higher-order derivatives of the potential energy surface, yet machine learned interatomic potentials (MLIPs) trained with a standard loss on energy, force, and stress errors can exhibit error in curvature, degrading the prediction of vibrational properties. We introduce phonon fine-tuning (PFT), which directly supervises second-order force constants of materials by matching MLIP energy Hessians to DFT-computed force constants from finite displacement phonon calculations. To scale to large supercells, PFT stochastically samples Hessian columns and computes the loss with a single Hessian-vector product. We also use a simple co-training scheme to incorporate upstream data to mitigate catastrophic forgetting. On the MDR Phonon benchmark, PFT improves Nequix MP by 55% on average across phonon thermodynamic properties and achieves state-of-the-art accuracy among models trained on Materials Project trajectories. PFT also generalizes to improve properties beyond second-derivatives, improving thermal conductivity predictions that rely on third-order derivatives of the potential energy.", "date": "2026-01-12", "categories": []}
{"id": "2602.04861", "title": "From Evaluation to Design: Using Potential Energy Surface Smoothness Metrics to Guide Machine Learning Interatomic Potential Architectures", "abstract": "Machine Learning Interatomic Potentials (MLIPs) sometimes fail to reproduce the physical smoothness of the quantum potential energy surface (PES), leading to erroneous behavior in downstream simulations that standard energy and force regression evaluations can miss. Existing evaluations, such as microcanonical molecular dynamics (MD), are computationally expensive and primarily probe near-equilibrium states. To improve evaluation metrics for MLIPs, we introduce the Bond Smoothness Characterization Test (BSCT). This efficient benchmark probes the PES via controlled bond deformations and detects non-smoothness, including discontinuities, artificial minima, and spurious forces, both near and far from equilibrium. We show that BSCT correlates strongly with MD stability while requiring a fraction of the cost of MD. To demonstrate how BSCT can guide iterative model design, we utilize an unconstrained Transformer backbone as a testbed, illustrating how refinements such as a new differentiable $k$-nearest neighbors algorithm and temperature-controlled attention reduce artifacts identified by our metric. By optimizing model design systematically based on BSCT, the resulting MLIP simultaneously achieves a low conventional E/F regression error, stable MD simulations, and robust atomistic property predictions. Our results establish BSCT as both a validation metric for practitioners to assess MLIP utility and as an \"in-the-loop\" model design proxy that alerts MLIP developers to physical challenges that cannot be efficiently evaluated by current MLIP benchmarks. The BSCT dataset and evaluation are available on https://github.com/ryanliu30/bsct.git", "date": "2026-02-04", "categories": []}
{"id": "2605.30743", "title": "A Padding Method for Enhanced Encoding of Inorganic Structures with Varying Chemical Compositions", "abstract": "Designing novel inorganic materials through generative models remains an important challenge for material science, driven by the complexity and diversity of inorganic structures across expansive chemical compositions and structural landscape. The vast combinatorial space of inorganic compounds demands innovative, AI-driven approaches to overcome limitations in generative accuracy and efficiency. To address this, we introduce a novel method that redefines the encoding and generation of inorganic materials by utilizing domain-specific symmetry-aware representation. Our approach not only refines the representation of intricate inorganic structures but also contributes to the field of material discovery by enhancing the precision and stability of generated candidates. Central to our methodology is a novel padding technique that exploits crystal symmetry information to enhance the encoding process. By integrating Wyckoff position length-aware padding into an encoder architecture, we achieve a more robust informed representation of inorganic materials. This symmetry-driven enhancement improves deep learning models to generate stable, previously unexplored inorganic structures with superior accuracy and computational efficiency. Furthermore, we introduce an end-to-end system that leverages the machine learning potential models to seamlessly generate novel, even those unseen in the training data, and stable inorganic materials from initial data to validated output. This pipeline integrates advanced generative models with stability analysis, marking a significant leap forward in the automated exploration and design of next-generation inorganic materials. Our method improved reconstruction accuracy 5.3% in proton conductor data, and generated 63.5% more novel stable inorganic material to baseline model on the perov-5 dataset.", "date": "2026-05-29", "categories": []}
{"id": "2606.00187", "title": "AI-Guided Design and Optimization of Graphite-Based Anodes via Iterative Experimental Feedback", "abstract": "This study presents an iterative AI-guided workflow that accelerates graphite-based anode development by improving both formulation feasibility and process robustness. Sequential learning via AI/ML-guided multiobjective inverse design for anode optimization was implemented using the Citrine Platform. Starting from a noisy, incomplete dataset, the Citrine Platform was used to generate early surrogate models, which despite low predictive certainty highlighted missing process constraints. By iteratively adding feasibility labels and boundary condition failures, the workflow rapidly converged toward manufacturable, higher-performing formulations. Fabrication reliability improved from frequent process failures to 100% successful cell production, while the fraction of cells delivering $\\geq$ 350 mAh g$^{-1}$ increased from 28.4% to 84.8%, with capacity retention rising from 42.1% to 97.3%. These results demonstrate that structured, feedback-driven AI workflows can transform imperfect industrial data into actionable guidance, enabling faster, more reproducible optimization of battery electrode manufacturing.", "date": "2026-05-29", "categories": []}
{"id": "2606.00315", "title": "Coupling Language Models with Physics-based Simulation for Synthesis of Inorganic Materials", "abstract": "Modern generative machine learning (ML) models can propose novel inorganic crystalline materials with targeted properties; however, synthesis planning of these materials remains difficult due to the complexity of the associated physical processes and limited availability of computational tools. We introduce a novel hybrid framework to evaluate Large Language Models (LLMs) in inorganic synthesis planning by combining thermodynamic databases with simplified kinetics models to approximate realistic synthesis conditions. As a case study, we focus on the niobium-oxygen system, which features multiple industrially relevant oxide phases with well-characterized data. In computational simulations, we compare LLM-generated synthesis routes with classical path-planning algorithms, showing that the implicit priors in LLMs can yield more viable strategies. In our evaluation setting, classical search methods serve primarily as a foil rather than a direct competitor. This illustrates the relative complexity of the problem and highlights where the LLM's implicit priors add value.", "date": "2026-05-29", "categories": []}
{"id": "2606.00938", "title": "Machine Learning Surrogate Modeling for Homogenization of Hyperelastic Materials with Boolean Microstructures", "abstract": "Data-driven surrogate models are an alternative to numerical homogenization of heterogeneous materials. In this contribution, a supervised learning approach is presented for predicting effective Lam\\'e parameters of hyperelastic composites from low-dimensional microstructural descriptors. The data set is based on previously published numerical homogenization results for ensembles of two-phase stochastic microstructures generated by planar Boolean models, covering variations of inclusion shape, phase contrast, and area fraction; see Br\\\"andel, Brands, Maike, Rheinbach, Schr\\\"oder, Schwarz and Stoyan (2022). A neural network is trained on combinations of scalar and curve-valued statistical descriptors, including the area fraction, a derived scalar shape descriptor $\\tau$, the two-point correlation function $S_2(r)$, and the lineal-path function $\\ell(z)$. Additional data representing limiting cases of the parameter space are incorporated to stabilize training and improve extrapolation behavior. The surrogate is evaluated by leave-one-grain-type-out cross-validation in order to assess generalization to unseen grain geometries. Numerical results demonstrate that additional descriptors can reduce relative errors. A predictor trained with $\\tau$ and $S_2(r)$ provides a compact representation with good quantitative accuracy and regular dense response behavior. Adding the lineal-path function $\\ell(z)$ further reduces the error at the available data points, indicating that it is a promising additional descriptor; however, dense post-training response evaluations show that improved pointwise accuracy does not automatically guarantee physically admissible behavior between sampled parameter values. This motivates future work on physically constrained surrogate models, loss formulations, bounded output parametrizations, and a more systematic representation of curve-valued geometric descriptors.", "date": "2026-05-31", "categories": []}
{"id": "2606.01012", "title": "Property Prediction of Stacked Bilayer Materials: A Multimodal Learning Approach", "abstract": "AI for materials science is a critical topic within AI for science, aiming to accelerate materials discovery and produce accurate property predictions. Bilayer 2D material stacking is essential for exploring new materials with novel functions and inherent phenomena, enabling the creation of new 2D bilayers for diverse real-world applications. Research on bilayer vdWs materials has made significant progress from experimental and computational perspectives. Various bilayer materials have been successfully synthe sized experimentally and the increasing utilization of high-throughput computing technology has con structed several computational two-dimensional materials databases. However, the use of AI to model bilayer stacking and predict new properties remains underexplored, necessitating further research studies. In this work, we propose a novel multimodal learning approach to study the interfaces between dissimilar materials that jointly enable new or multiple functions, and to predict new properties arising from the vertical integration (stacking) of different functional material layers under given configurations. Comprehensive experiments demonstrate the effectiveness and efficiency of our approach compared to baseline methods. Our code is available at https://github.com/AnVuong123/bimat ml.", "date": "2026-05-31", "categories": []}
{"id": "2606.02455", "title": "Speculative Sampling For Faster Molecular Dynamics", "abstract": "Molecular dynamics (MD) is a key tool for simulating the dynamical behavior of atomic systems. However, MD is inherently serial, which makes it difficult to increase single-system throughput with concurrent compute. To address this, we introduce Langevin Speculative Dynamics (LSD), a distributed and model-agnostic speculative sampler for accelerating MD without adding relative error. Inspired by speculative methods in language and diffusion modeling, LSD uses a draft model to propose fast simulation steps and verifies them in parallel with a slower target model, applying a transport map from the draft to the target distribution. We extend speculative sampling to second-order Langevin dynamics, derive the achievable speedup as a function of physical parameters, show that LSD generalizes across different systems and draft-target combinations with a 3-9x speedup, and confirm theoretically and empirically that LSD samples trajectories from its target model distribution.", "date": "2026-06-01", "categories": []}
{"id": "2606.02507", "title": "Towards Automated Discovery: A Review of Generative Models, Multimodal Learning and Closed-Loop Workflows in Inverse Materials Design", "abstract": "Inverse materials design is shifting materials discovery from forward prediction to targeted proposal of candidates that satisfy objectives under physical constraints. Here, we review recent advances in generative crystal structure modeling, multimodal learning, and closed-loop design pipelines for crystalline solids. We survey how modern generators learn chemical-structural priors from large databases to enable controllable sampling of periodic structures, and compare leading model classes including variational autoencoders, normalizing flows, autoregressive formulations, and diffusion models. Particular attention is given to how feasibility constraints and physical priors are enforced across the workflow, through representation choices, training objectives, sampling-time guidance, and post-generation screening and relaxation. We also discuss how multimodal learning fuses diverse materials modalities, including crystal structures, thermodynamic, electronic information, microscopy, spectroscopy, processing context, and scientific text, to construct a more universal, transferable representation of chemical space. In addition, diverse inverse-design strategies are examined, particularly those that integrate conditional generation with latent optimization, Bayesian optimization, reinforcement learning, and active learning. Finally, we highlight recurring failure modes, such as surrogate exploitation, diversity collapse, distribution shift, and the stability-synthesizability gap, and outline discovery-grade evaluation practices based on staged reporting of validity, novelty, uniqueness, stability, and cost.", "date": "2026-06-01", "categories": []}
{"id": "2601.19966", "title": "Global Plane Waves From Local Gaussians: Periodic Charge Densities in a Blink", "abstract": "We introduce ELECTRAFI, a fast, end-to-end differentiable model for predicting periodic charge densities in crystalline materials. ELECTRAFI constructs anisotropic Gaussians in real space and exploits their closed-form Fourier transforms to analytically evaluate plane-wave coefficients via the Poisson summation formula. This formulation delegates non-local and periodic behavior to analytic transforms, enabling reconstruction of the full periodic charge density with a single inverse FFT. By avoiding explicit real-space grid probing, periodic image summation, and spherical harmonic expansions, ELECTRAFI matches or exceeds state-of-the-art accuracy across periodic benchmarks while being up to $633 \\times$ faster than the strongest competing method, reconstructing crystal charge densities in a fraction of a second. When used to initialize DFT calculations, ELECTRAFI reduces total DFT compute cost by up to ~20%, whereas slower charge density models negate savings due to high inference times. Our results show that accuracy and inference cost jointly determine end-to-end DFT speedups, and motivate our focus on efficiency.", "date": "2026-01-27", "categories": []}
{"id": "2602.09309", "title": "How Far Can You Grow? Characterizing the Extrapolation Frontier of Graph Generative Models for Materials Science", "abstract": "Every generative model for crystalline materials harbors a critical structure size beyond which its outputs become unreliable; we call this the extrapolation frontier. Despite its consequences for nanomaterial design, this frontier has never been systematically measured. We introduce RADII, a radius-resolved benchmark of ~75,000 crystal-derived nanoparticle structures (33-11,298 atoms) that treats radius as a continuous scaling knob, tracing generation quality from in- to out-of-distribution under leakage-free splits. Each model is conditioned on target composition and atom count, isolating geometric extrapolation as the evaluation variable. RADII provides frontier-specific diagnostics: per-radius error profiles pinpoint each architecture's scaling ceiling, surface-interior decomposition separates boundary from bulk failures, and cross-metric sequencing reveals which aspect of structural fidelity breaks first. Benchmarking five state-of-the-art architectures, we find that: (i) well-behaved models degrade by ~13% in global positional error beyond training radii, while divergent models show poor fidelity across scales, with local bond fidelity ranging from negligible degradation to over 2x error growth; (ii) no two architectures share a failure sequence, revealing the frontier as a multi-dimensional surface shaped by model family; and (iii) well-behaved models follow the expected geometric scaling exponent alpha ~ 1/3, whose in-distribution fit predicts out-of-distribution error, making frontiers forecastable. Scaling MatterGen to its published parameter count stabilizes sampling but does not close the frontier, while DiffCSP remains unstable at published scale. These findings establish output scale as a first-class evaluation axis for geometric generative models. Code and data: https://github.com/KurbanIntelligenceLab/RADII.", "date": "2026-02-10", "categories": []}
{"id": "2605.31532", "title": "Discovering Thermodynamically Admissible Dissipation Potentials via Grammar-Based Symbolic Regression", "abstract": "Constitutive laws for inelastic materials must satisfy strict thermodynamic admissibility requirements, yet current data-driven approaches sacrifice interpretability, even when formal guarantees are provided by physics-encoded architectures. We propose a symbolic regression framework for the data-driven discovery of dissipation potentials governing the evolution of internal variables within the Generalized Standard Materials (GSM) formalism. Starting from the Clausius--Duhem inequality, we enforce the thermodynamic requirements, convexity and non-negativity, that the dual dissipation potential must satisfy to guarantee non-negative mechanical dissipation. These requirements are formulated in the general subdifferential setting, encompassing rate-dependent (viscoelastic) and viscoplastic dissipative mechanisms, including potentials with genuine elastic domains, within a unified framework. Candidate potentials are generated by a composition-extended convexity-preserving grammar that guarantees thermodynamic admissibility \\emph{by construction}. The framework is validated on synthetic datasets spanning Newtonian, power-law, and Bingham viscoplastic ground truths under process and measurement noise, and on experimental oscillatory shear measurements of a synthetic elastomer across multiple strain amplitudes and frequencies, where the discovered potentials reproduce the amplitude-dependent softening of the dynamic moduli and outperform a calibrated linear Zener baseline.", "date": "2026-05-29", "categories": []}
{"id": "2509.17208", "title": "Active Learning for Machine Learning Driven Molecular Dynamics", "abstract": "Machine-learned coarse-grained (CG) potentials are fast, but degrade over time when simulations reach under-sampled bio-molecular conformations, and generating widespread all-atom (AA) data to combat this is computationally infeasible. We propose a novel active learning (AL) framework for CG neural network potentials in molecular dynamics (MD). Building on the CGSchNet model, our method employs root mean squared deviation (RMSD)-based frame selection from MD simulations in order to generate data on-the-fly by querying an oracle during the training of a neural network potential. This framework preserves CG-level efficiency while correcting the model at precise, RMSD-identified coverage gaps. By training CGSchNet, a coarse-grained neural network potential, we empirically show that our framework explores previously unseen configurations and trains the model on unexplored regions of conformational space. Our active learning framework enables a CGSchNet model trained on the Chignolin protein to achieve a 33.05\\% improvement in the Wasserstein-1 (W1) metric in Time-lagged Independent Component Analysis (TICA) space on an in-house benchmark suite.", "date": "2025-09-21", "categories": []}
{"id": "2511.14426", "title": "MiAD: Mirage Atom Diffusion for De Novo Crystal Generation", "abstract": "In recent years, diffusion-based models have demonstrated exceptional performance in searching for simultaneously stable, unique, and novel (S.U.N.) crystalline materials. However, most of these models don't have the ability to change the number of atoms in the crystal during the generation process, which limits the variability of model sampling trajectories. In this paper, we demonstrate the severity of this restriction and introduce a simple yet powerful technique, mirage infusion, which enables diffusion models to change the state of the atoms that make up the crystal from existent to non-existent (mirage) and vice versa. We show that this technique improves model quality by up to x2.5 compared to the same model without this modification. The resulting model, Mirage Atom Diffusion (MiAD), is an equivariant joint diffusion model for de novo crystal generation that is capable of altering the number of atoms during the generation process. MiAD achieves an 8.2% S.U.N. rate on the MP-20 dataset, which substantially exceeds existing state-of-the-art approaches. Code: https://github.com/andrey-okhotin/miad.git", "date": "2025-11-18", "categories": []}
{"id": "2602.10637", "title": "Coarse-Grained Boltzmann Generators", "abstract": "Sampling equilibrium molecular configurations from the Boltzmann distribution is a longstanding challenge. Boltzmann Generators (BGs) address this by combining exact-likelihood generative models with importance sampling, but practical scalability is limited. Meanwhile, coarse-grained surrogates enable the modeling of larger systems by reducing effective dimensionality, yet often lack a reweighting procedure required to ensure asymptotically correct statistics. In this work, we propose Coarse-Grained Boltzmann Generators (CG-BGs), a framework for reduced-order generative modeling with importance sampling in coarse-grained coordinate space. CG-BGs generate samples using a flow-based model and reweight them using a learned potential of mean force (PMF). We show that the PMF can be learned from rapidly converged trajectories via enhanced sampling force matching. Experiments demonstrate that CG-BGs capture solvent-mediated interactions in highly reduced representations while substantially reducing computational cost relative to atomistic BGs, providing a practical route toward equilibrium sampling of larger molecular systems.", "date": "2026-02-11", "categories": []}
{"id": "2603.01311", "title": "Catalyst-Agent: Autonomous heterogeneous catalyst screening with an LLM Agent", "abstract": "The discovery of novel catalysts tailored for particular applications is a major challenge for the twenty-first century. Traditional methods for this include time-consuming and expensive experimental trial-and-error approaches in labs based on chemical theory or heavily computational first-principles approaches based on density functional theory. Recent studies show that deep learning models like graph neural networks (GNNs) can significantly speed up the screening of catalyst materials by many orders of magnitude, with very high accuracy and fidelity. In this work, we introduce Catalyst-Agent, a Model Context Protocol (MCP) server-based, LLM-powered AI agent. It can explore vast material databases using the OPTIMADE API, make structural modifications, calculate adsorption energies using Meta FAIRchem's UMA (GNN) model via FAIRchem's AdsorbML workflow and slab construction, and make useful material suggestions to the researcher in a closed-loop manner, including structural modifications to refine near-miss candidates. It is tested on three pivotal reactions: the oxygen reduction reaction (ORR), the nitrogen reduction reaction (NRR), and the CO2 reduction reaction (CO2RR). Catalyst-Agent achieves a success rate of 33-41% among all the materials it chooses and evaluates, and manages to converge in 1-4 trials per successful material on average. This work demonstrates the potential of AI agents to exercise their planning capabilities and tool use for autonomous catalyst screening workflows.", "date": "2026-03-01", "categories": []}
{"id": "2605.28368", "title": "LEIA: Learned Environment for Interactive Architected Materials", "abstract": "World models have enabled interactive exploration of game environments and robotic manipulation, but physical engineering remains beyond their reach: real materials exhibit nonlinear constitutive laws, carry history-dependent internal state, undergo inertial dynamics, and may possess hierarchical structures spanning multiple length scales. We present LEIA (Learned Environment for Interactive Architected materials), a world model that lets engineers apply boundary conditions step by step and observe the resulting deformation and stress fields in real time. LEIA handles large three-dimensional unstructured meshes and generates autoregressive responses to user-specified loading. We introduce MicroPlate, a benchmark of architected plates spanning two regimes of microstructure modeling: architected lattices that resolve microstructure explicitly through three-dimensional geometry, and a homogeneous plate where microstructural change is modeled implicitly through internal degrees of freedom. MicroPlate is used to assess LEIA alongside four baseline methods across both regimes. Finally, we demonstrate that LEIA enables efficient candidate generation and ranking for fast surrogate-guided search for de novo designs of architected materials, with stress-accurate candidate ranking validated by finite element ground truth.", "date": "2026-05-27", "categories": []}
{"id": "2605.29555", "title": "From Blind Guess to Informed Judgment: Teaching LLMs to Evaluate Materials by Building Knowledge-Augmented Preference Signals", "abstract": "As candidate generation and high-throughput experimentation advance, the primary bottleneck in materials discovery is shifting from property prediction to making reliable evaluations among massive candidate sets. We propose a Knowledge-Augmented Preference Signals Framework, MaterEval, that automatically produces, for the same candidate, two evaluations: an informed judgment that follows expert rules and provides supporting evidence, and a rule-removed blind guess. By pairing the two evaluations as preference data, we guide general-purpose large language models (LLMs), originally lacking materials-specific criteria, from intuitive judgment toward reliable evaluation supported by explicit evidence. To balance throughput, cost, and reliability, we further introduce a fast-slow reasoning scheme that decouples large-scale rapid screening from in-depth review on a small subset. Using high-entropy alloy (HEA) assessment as a case study, we show that, without external retrieval and relying solely on internalized capabilities, small open-source LLMs achieve substantial gains in accuracy, conclusion consistency, and evidence discrimination, approaching the performance of rule-based closed-source LLMs. These results demonstrate that expert rules can be systematically transformed into learnable preference signals, enabling a low-cost and deployable evaluation module for autonomous materials discovery loops.", "date": "2026-05-28", "categories": []}
{"id": "2605.28287", "title": "AtomComposer: Discovering Chemical Space from First Principles with Reinforcement Learning", "abstract": "Discovering novel stable molecules without training data remains a grand scientific challenge. Current molecular generative models are trained on large, pre-curated datasets, which introduce biases and limit exploration of novel chemistry. In contrast, we propose a new paradigm: autonomous, generalized agents capable of mapping vast, unknown chemical spaces without any pretraining. For the first time, we present AtomComposer, a self-guided agent that autonomously constructs valid 3D isomers under stoichiometric constraints and is trained exclusively online using reinforcement learning. Unlike existing approaches that generally overfit to a specific chemical formula, we establish a multi-composition training scheme that enables a broad generalization across diverse chemistry, guided by energy- and validity-based rewards. Our agent can discover up to an order of magnitude more valid isomers on unseen test formulas than existing single-composition reinforcement-learning baselines trained with per-step energy rewards. These results fulfill the promise of online reinforcement learning as a powerful paradigm for scalable, from-scratch exploration of chemical configuration space.", "date": "2026-05-27", "categories": []}
{"id": "2605.28487", "title": "ProvMind: Provenance-grounded reasoning for materials synthesis", "abstract": "Materials process optimization requires reasoning over routes, conditions, tools and causal dependencies, yet most computational formulations flatten synthesis procedures into text or ordered steps. We introduce MatProcBench, a provenance-grounded benchmark constructed from literature-mined MatPROV graphs, to evaluate seven process-reasoning tasks spanning route continuity, step-level variable inference and global causal consistency under both same-split and shift-aware evaluation, including a strict dual-OOD split that combines temporal and material-class shift. We further introduce ProvMind, a process-memory reasoning framework that retrieves analogous training processes, converts them into provenance-aware option-level compatibility scores, and uses a language model for constrained final decision making. ProvMind achieves 52.84\\% accuracy on the dual-OOD split, outperforming prompting, retrieval-augmented and supervised fine-tuning baselines.", "date": "2026-05-27", "categories": []}
{"id": "2605.28594", "title": "Thermodynamic properties of chemically disordered compounds via AI-driven estimation of partition function with the PULSE method", "abstract": "In this article, we present an improved version of the PULSE method (Partition function Unsupervised Learning Sampling and Evaluation) for estimating the thermodynamic properties of chemically disordered compounds. The aim is to reduce the computational cost of Monte Carlo approaches for this type of material and to demonstrate that this generative tool can estimate thermodynamic properties by sampling and estimating the partition function of the system. To validate this innovative approach, we use the 2D Ising model as a benchmark. We demonstrate that our method accurately reproduces average properties with high precision and efficiency compared to traditional Monte Carlo sampling methods. Our results highlight the efficiency and adaptability of the PULSE method, making it a valuable tool for studying materials for which conventional methods are too inefficient to compute properties affected by chemical disorder at low cost.", "date": "2026-05-27", "categories": []}
{"id": "2512.16702", "title": "How accurate are foundational machine learning interatomic potentials for heterogeneous catalysis?", "abstract": "Foundational machine learning interatomic potentials (MLIPs) are being developed at a rapid pace, promising closer and closer approximation to ab initio accuracy. This unlocks the possibility to simulate much larger length and time scales. However, benchmarks for these MLIPs are usually limited to ordered, crystalline and bulk materials. Hence, reported performance does not necessarily accurately reflect MLIP performance in real applications such as heterogeneous catalysis. Here, we systematically analyze zero-shot performance of 80 different MLIPs, evaluating tasks typical for heterogeneous catalysis across a range of different data sets, including adsorption and reaction on surfaces of alloyed metals, oxides, and metal-oxide interfacial systems. We demonstrate that current-generation foundational MLIPs can already perform at high accuracy for applications such as predicting vacancy formation energies of perovskite oxides or zero-point energies of supported nanoclusters. However, limitations also exist. We find that many MLIPs catastrophically fail when applied to magnetic materials, and structure relaxation in the MLIP generally increases the energy prediction error compared to single-point evaluation of a previously optimized structure. Comparing low-cost task-specific models to foundational MLIPs, we highlight some core differences between these model approaches and show that -- if considering only accuracy -- these models can compete with the current generation of best-performing MLIPs. Furthermore, we show that no single MLIP universally performs best, requiring users to investigate MLIP suitability for their desired application.", "date": "2025-12-18", "categories": []}
{"id": "2605.26540", "title": "DGLD: Domain-Gated Latent Diffusion for the Discovery of Novel Energetic Materials", "abstract": "Energetic-materials performance gains translate directly into reduced propellant mass, smaller warheads, and more efficient civilian gas-generators, yet no new HMX-class compound has been disclosed in fifteen years. Designing one is a sparse-label problem: of ~66 k labelled CHNO molecules only ~3 k carry experimental or DFT-quality measurements, and naive generative models trained on the full mixture either memorise the high-performance tail or extrapolate without calibration. We introduce Domain-Gated Latent Diffusion (DGLD): a label-quality gate at training time, multi-task score-model guidance at sample time, and a four-stage chemistry-validation funnel ending in first-principles DFT audit. The result is 12 DFT-confirmed novel leads. The headline compound, 3,4,5-trinitro-1,2-isoxazole (L1), reaches \\r{ho}_\"cal\" =2.09 g/cm3 and D_\"K-J,cal\" =8.25 km/s and is structurally dissimilar from all 65 980 training molecules (nearest-neighbour Tanimoto 0.27). A co-headline lead, E1 (4-nitro-1,2,3,5-oxatriazole), exceeds L1 on calibrated detonation velocity (D_\"K-J,cal\" =9.00 km/s) from a chemotype family disjoint from L1's. DGLD is the only method to land in the productive quadrant (simultaneously novel and on-target) at DFT level. SMILES-LSTM memorises 18.3% of its outputs exactly; SELFIES-GA's best novel candidate loses 3.5 km/s under DFT audit; REINVENT 4 generates novel high-N heterocycles but peaks at D=9.02 km/s. Code, checkpoints, and 918 mined hard negatives are released on Zenodo (DOI 10.5281/zenodo.19821953); the next compound to enter the HMX-class band can be discovered, validated, and recommended for synthesis at the cost of a few GPU-days.", "date": "2026-05-26", "categories": []}
{"id": "2605.26543", "title": "PolyFusionAgent: A Multimodal Foundation Model and Autonomous AI Assistant for Polymer Property Prediction and Inverse Design", "abstract": "Polymer discovery is central to fields ranging from energy storage to biomedicine, but it is hindered by an astronomically large chemical design space and fragmented representations of structure, properties, and prior knowledge. This fragmentation leaves many AI models disconnected from physical and experimental reality, restricting their ability to support directly actionable design decisions. Here we introduce PolyFusionAgent, an interactive framework coupling a multimodal polymer foundation model (PolyFusion) with a tool-augmented, literature-grounded design agent (PolyAgent). PolyFusion aligns complementary polymer views including sequence, topology, 3D geometry, and fingerprints across millions of polymers to learn a shared latent space transferable across chemistries and data regimes, improving thermophysical property prediction and enabling property-conditioned generation of chemically valid, structurally novel polymers beyond the reference design space. PolyAgent closes the design loop by linking prediction and inverse design with evidence retrieval from the polymer literature, proposing, evaluating, and contextualizing hypotheses with explicit precedent in one workflow. Together, PolyFusionAgent enables interactive, evidence-linked polymer discovery combining large-scale representation learning, multimodal chemical knowledge, and verifiable scientific reasoning.", "date": "2026-05-26", "categories": []}
{"id": "2605.26741", "title": "MatFormBench: A Benchmarking Evaluation Framework for Target-Driven Materials Formulation", "abstract": "Inverse design of materials has significantly advanced target-driven formulation optimization, yet existing materials machine learning benchmarks remain limited to forward property prediction, failing to systematically evaluate inverse optimization and generation algorithms, a critical gap that hinders the progress of target-driven materials design. To address this limitation, we propose MatFormBench, a novel benchmarking ecosystem tailored to evaluate and guide generative strategies for target-driven formulation. MatFormBench integrates a physics-driven formulation generation scheme to generate synthetic samples that faithfully emulate realistic materials structure-property response relationships, complemented by five escalating difficulty levels to quantify the complexity of these relationships. To rigorously assess algorithm performance, we further propose MatFormScore, a multi-dimensional metric that comprehensively quantifies performance across five critical axes: target success, search efficiency, exploratory capacity, robustness, and stability. We validate MatFormBench by evaluating 39 diverse inverse design algorithms, covering classical surrogate-assisted black-box search, state-of-the-art deep generative models, and increasingly popular Large Language Model (LLM)-based recommendation strategies. Across 1170 standardized algorithm-task evaluations, diffusion-based models demonstrate the strongest overall performance, while Variational Autoencoder (VAE)-based and Genetic Algorithm (GA)-based methods exhibit distinct advantages in specific scenarios. By establishing a unified evaluation standard for target-driven materials formulation, MatFormBench enables reproducible benchmarking, principled algorithm comparison, and diagnostic analysis of inverse design strategies, providing a foundational tool for advancing materials inverse design.", "date": "2026-05-26", "categories": []}
{"id": "2602.20210", "title": "Multimodal Crystal Flow: Any-to-Any Modality Generation for Unified Crystal Modeling", "abstract": "Crystal modeling spans a family of conditional and unconditional generation tasks, including crystal structure prediction (CSP) and de novo generation (DNG). While recent deep generative models have shown promising performance, they remain largely task-specific, lacking a unified framework that shares crystal representations across tasks. To address this limitation, we propose Multimodal Crystal Flow (MCFlow), a unified multimodal flow model that realizes multiple crystal generation tasks as distinct inference trajectories via independent time variables for atom types and crystal structures. To enable multimodal flow in a standard transformer model, we introduce a composition- and symmetry-aware atom ordering with hierarchical permutation augmentation, injecting compositional and crystallographic priors without explicit structural templates. Experiments on the MP-20 and MPTS-52 benchmarks show that a single MCFlow model is competitive with task-specific baselines across CSP, DNG, and structure-conditioned atom type generation.", "date": "2026-02-23", "categories": []}
{"id": "2604.27636", "title": "Generative structure search for efficient and diverse discovery of molecular and crystal structures", "abstract": "Predicting stable and metastable structures is central to molecular and materials discovery, but remains limited by the cost of searching high-dimensional energy landscapes. Deep generative models offer efficient structure sampling, yet their outputs remain shaped by training data and can underexplore minima that are rare but physically relevant. We introduce generative structure search (GSS), a unified framework that formulates diffusion-based generation and random structure search (RSS) as limiting regimes of a common sampling process driven by learned score fields and physical forces. Coupling these drivers lets GSS use data priors to accelerate sampling while retaining energy-guided exploration of local minima. Across molecular and crystalline systems, GSS recovers diverse metastable structures with more than tenfold lower sampling cost than RSS for broad coverage and remains effective for compositions outside the training distribution. The results establish a physically grounded generative search strategy for discovering structures beyond the reach of data-driven sampling alone.", "date": "2026-04-30", "categories": []}
{"id": "2605.14759", "title": "Crys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinement", "abstract": "De novo crystal generation seeks to discover materials that are not merely realistic, but also stable and novel. However, most existing generative models are trained to maximize the likelihood of observed crystals, which encourages samples to stay close to known materials yet not necessarily align with the criteria that matter in discovery. Our empirical analysis shows that current crystal generative models exhibit a clear conflict between stability and novelty: samples near the observed distribution tend to retain stability but offer limited novelty, whereas samples farther from it often lose stability rapidly. This suggests that the useful region for discovering crystals that are both stable and novel is extremely narrow. To move beyond this limitation, we introduce Crys-JEPA, a joint embedding predictive architecture for crystals that learns an energy-aware latent space preserving formation-energy differences. In this space, stability assessment can be reformulated as an embedding-based comparison against accessible training crystals, reducing the reliance on expensive energy evaluation and task-specific external references. Building on Crys-JEPA, we further develop a screening-and-refinement pipeline that identifies promising generated crystals and reintroduces them to refine the generative model. On MP-20 and Alex-MP-20 datasets, we achieve improvements over baselines up to 53.8% and 72.7% on V.S.U.N. metric, respectively.", "date": "2026-05-14", "categories": []}
{"id": "2605.14769", "title": "Composable Crystals: Controllable Materials Discovery via Concept Learning", "abstract": "De novo crystal generation, a central task in materials discovery, aims to generate crystals that are simultaneously valid, stable, unique, and novel. Existing methods mainly rely on black-box stochastic sampling, providing limited control over how generated structures move beyond the observed distribution. In this paper, we introduce a concept-based compositional framework for crystal generation. We train a vector-quantized variational autoencoder to automatically discover a shared set of reusable crystal concepts, which serve as building blocks for guided generation. These learned concepts naturally exhibit interpretability from both local atomic environments and global symmetry patterns, and generalize to crystals from different distributions. By recombining such concepts, our framework enables controllable exploration of novel crystals beyond the training distribution, rather than relying solely on unconstrained random sampling. To further improve composition efficiency, we introduce a composition generator and iteratively refine it using high-quality samples generated by the model itself. The resulting concept compositions are then used to condition downstream crystal generation. Numerical experiments on MP-20 and Alex-MP-20 show that compositing concepts separately increase base model up to 53.2% and 51.7% on V.S.U.N metric, with particular gains in novelty.", "date": "2026-05-14", "categories": []}
{"id": "2605.24073", "title": "Multitask learning with semiempirical orbital charges enables sample-efficient MLIPs", "abstract": "Machine learning interatomic potentials (MLIPs) require generating computationally expensive, large-scale training datasets to accurately simulate materials and molecules. Incorporating electronic structure information using multitask learning improves sample efficiency, however, training on full Hamiltonian matrices, which scale quadratically with the number of atoms, is intractable for large datasets. In this work, we show that multitask learning utilizing orbitally resolved semiempirical charges significantly improves sample efficiency and accuracy in MLIPs. To efficiently predict orbital charges, we implement a specialized equivariant model, reducing charge prediction error compared to an invariant baseline. By augmenting training with computationally inexpensive GFN1-xTB orbital charges, which scale linearly with the number of atoms, our model achieves a 46\\% reduction in energy mean absolute error and requires five times less data to match the performance of energy-only models. Furthermore, our approach outperforms models trained on expensive density functional theory (DFT) atomic charges, capturing orbitally resolved electronic complexity and forcing the network to learn a physically accurate latent space that spontaneously clusters metals by shared chemical properties. Because orbital charges are only required during training, this approach preserves inference efficiency, providing a scalable recipe for developing accurate, data-efficient foundation models for complex chemical systems.", "date": "2026-05-22", "categories": []}
{"id": "2605.25866", "title": "UNATE: UNsupervised ATomic Embedding for crystal structures property prediction", "abstract": "Accurately predicting crystal properties is critical for accelerating materials discovery, but it is often limited by scarce labeled data and costly theoretical calculations. To alleviate this, we propose UNATE (Unsupervised Atomic Embedding), a framework that leverages structural information extracted from unlabeled crystal structures. UNATE integrates an unsupervised denoising autoencoder with self-supervised contrastive learning to learn robust atomic representations, which are then used as input features for downstream property prediction. Experimental results show that replacing raw atomic numbers with UNATE-pretrained node embeddings yields a 2.7\\% improvement over the full-data baseline. Notably, the benefits become more pronounced in scenarios with limited labeled data, reaching improvements of up to 10\\% when only 25\\% of the labeled data is used.", "date": "2026-05-25", "categories": []}
{"id": "2512.06404", "title": "GENIUS: An Agentic AI Framework for Autonomous Design and Execution of Simulation Protocols", "abstract": "Predictive atomistic simulations have propelled materials discovery, yet routine setup and debugging still demand computer specialists. This know-how gap limits Integrated Computational Materials Engineering (ICME), where state-of-the-art codes exist but remain cumbersome for non-experts. We address this bottleneck with GENIUS, an AI-agentic workflow that fuses a smart Quantum ESPRESSO knowledge graph with a tiered hierarchy of large language models supervised by a finite-state error-recovery machine. Here we show that GENIUS translates free-form human-generated prompts into validated input files that run to completion on $\\approx$80% of 295 diverse benchmarks, where 76% are autonomously repaired, with success decaying exponentially to a 7% baseline. Compared with LLM-only baselines, GENIUS halves inference costs and virtually eliminates hallucinations. The framework democratizes electronic-structure DFT simulations by intelligently automating protocol generation, validation, and repair, opening large-scale screening and accelerating ICME design loops across academia and industry worldwide.", "date": "2025-12-06", "categories": []}
{"id": "2410.12771", "title": "Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models", "abstract": "The ability to discover new materials with desirable properties is critical for numerous applications from helping mitigate climate change to advances in next generation computing hardware. AI has the potential to accelerate materials discovery and design by more effectively exploring the chemical space compared to other computational methods or by trial-and-error. While substantial progress has been made on AI for materials data, benchmarks, and models, a barrier that has emerged is the lack of publicly available training data and open pre-trained models. To address this, we present a Meta FAIR release of the Open Materials 2024 (OMat24) large-scale open dataset and an accompanying set of pre-trained models. OMat24 contains over 110 million density functional theory (DFT) calculations focused on structural and compositional diversity. Our EquiformerV2 models achieve state-of-the-art performance on the Matbench Discovery leaderboard and are capable of predicting ground-state stability and formation energies to an F1 score above 0.9 and an accuracy of 20 meV/atom, respectively. We explore the impact of model size, auxiliary denoising objectives, and fine-tuning on performance across a range of datasets including OMat24, MPtraj, and Alexandria. The open release of the OMat24 dataset and models enables the research community to build upon our efforts and drive further advancements in AI-assisted materials science.", "date": "2024-10-16", "categories": []}
{"id": "2504.13048", "title": "Design Topological Materials by Reinforcement Fine-Tuned Generative Model", "abstract": "Topological insulators (TIs) and topological crystalline insulators (TCIs) are materials with unconventional electronic properties, making their discovery highly valuable for practical applications. However, such materials, particularly those with a full band gap, remain scarce. Given the limitations of traditional approaches that scan known materials for candidates, we focus on the generation of new topological materials through a generative model. Specifically, we apply reinforcement fine-tuning (ReFT) to a pre-trained generative model, thereby aligning the model's objectives with our material design goals. We demonstrate that ReFT is effective in enhancing the model's ability to generate TIs and TCIs, with minimal compromise on the stability of the generated materials. Using the fine-tuned model, we successfully identify a large number of new topological materials, with Ge$_2$Bi$_2$O$_6$ serving as a representative example--a TI with a full band gap of 0.26 eV, ranking among the largest known in this category.", "date": "2025-04-17", "categories": []}
{"id": "2507.06929", "title": "Machine-Learned Force Fields for Lattice Dynamics at Coupled-Cluster Level Accuracy", "abstract": "We investigate Machine-Learned Force Fields (MLFFs) trained on approximate Density Functional Theory (DFT) and Coupled Cluster (CC) level potential energy surfaces for the carbon diamond and lithium hydride solids. We assess the accuracy and precision of the MLFFs by calculating phonon dispersions and vibrational densities of states (VDOS) that are compared to experiment and reference ab initio results. To overcome limitations from long-range effects and the lack of atomic forces in the CC training data, a delta-learning approach based on the difference between CC and DFT results, as well as a charge aware MLFF approach is explored. Compared to DFT, MLFFs trained on CC theory yield higher vibrational frequencies for optical modes, agreeing better with experiment. Furthermore, the MLFFs are used to estimate anharmonic effects on the VDOS of lithium hydride at the level of CC theory.", "date": "2025-07-09", "categories": []}
{"id": "2605.20242", "title": "LEAP: A closed-loop framework for perovskite precursor additive discovery", "abstract": "Efficient discovery of precursor additives is essential for improving the performance of perovskite solar cells, yet the large chemical space makes conventional trial-and-error screening inefficient. We develop LEAP(LLM-driven Exploration via Active Learning for Perovskites), an expert-in-the-loop closed framework that couples a domain-specialized large language model(LLM) with active learning for iterative additive prioritization. The LLM is trained to extract mechanism-relevant knowledge from the perovskite additive literature and to represent candidate molecules through interpretable descriptors, which are further integrated into a Bayesian optimization workflow for uncertainty-aware prioritization under low-data conditions. Benchmark results on unseen literature show that the domain-specialized model outperforms general-purpose models in mechanism-consistent reasoning. Experimental validation in an expert-in-the-loop proof-of-concept study suggests improved additive prioritization across three screening rounds, leading to average device PCEs of 20.13% and 20.87% for the later-round 6-CDQ- and 2-CNA-treated devices, respectively, compared with 19.25% for the control, with a champion PCE of 21.32%. These results provide preliminary evidence that literature-grounded mechanistic descriptors, when coupled with Bayesian optimization and expert feasibility review, can support mechanism-aware additive prioritization in perovskite photovoltaics.", "date": "2026-05-18", "categories": []}
{"id": "2605.20581", "title": "TriForces: Augmenting Atomistic GNNs for Transferable Representations", "abstract": "Machine learning interatomic potentials (MLIPs) achieve excellent accuracy when trained on large Density Functional Theory (DFT) data. To be useful in practice, they must often be adapted to target chemistries using small and expensive task-specific datasets. However, MLIPs transfer inconsistently across domains, with representations that often loose accessible composition and structure information. To address this, we present TriForces, a model-agnostic three-stream framework that separates composition and structure information, combined with self-supervised learning to preserve transferable representations. TriForces improves performance on MatBench and QM9 over baselines without needing DFT labels and enables efficient similar structure retrieval through its learned latent space. On OMat24, in limited-data training regime, TriForces reduces energy MAE by 57% at 20K samples only and improves force MAE across sample sizes. We release pretrained TriForces variants across multiple MLIP architectures with code at https://github.com/Ramlaoui/triforces.", "date": "2026-05-20", "categories": []}
{"id": "2605.21083", "title": "AIMBio-Mat: An AI-Native FAIR Platform for Closed-Loop Materials Discovery and Biomedical Translation", "abstract": "Materials discovery and biomedical translation increasingly require models that can reason across composition, processing, structure, biological response, manufacturability, safety, and governance constraints. Existing materials and biomedical data ecosystems are powerful but remain poorly coupled for AI-guided discovery. Here we present AIMBio, a conceptual framework for an AI-native, FAIR, and governance-aware decision layer that links materials provenance, biomedical context, knowledge graphs, uncertainty-aware machine learning, and human-in-the-loop active learning. The framework formulates biomedical-materials discovery as constrained multi-objective optimization under uncertainty and introduces practical requirements for metadata, model documentation, risk-tiered governance, evaluation metrics, and phased implementation. To make the roadmap testable, we add a minimum viable prototype specification and a worked pilot for AI-guided nanomaterials for drug delivery. AIMBio is positioned as exploratory and preclinical discovery infrastructure, not as clinical decision-support software; any clinical or regulated-device use would require separate validation, change control, and regulatory review. The central contribution is a publishable platform blueprint for converting fragmented materials and biomedical records into auditable, experimentally actionable, and translationally responsible discovery workflows.", "date": "2026-05-20", "categories": []}
{"id": "2602.01358", "title": "Towards knowledge-based workflows: a semantic approach to atomistic simulations for mechanical and thermodynamic properties", "abstract": "Mechanical and thermodynamic properties, including the influence of crystal defects, are critical for evaluating materials in engineering applications. Molecular dynamics simulations provide valuable insight into these mechanisms at the atomic scale. However, current practice often relies on fragmented scripts with inconsistent metadata and limited provenance, which hinders reproducibility, interoperability, and reuse. FAIR data principles and workflow-based approaches offer a path to address these limitations. We present reusable atomistic workflows that incorporate metadata annotation aligned with application ontologies, enabling automatic provenance capture and FAIR-compliant data outputs. The workflows cover key mechanical and thermodynamic quantities, including equation of state, elastic tensors, mechanical loading, thermal properties, defect formation energies, and nanoindentation. We demonstrate validation of structure-property relations such as the Hall-Petch effect and show that the workflows can be reused across different interatomic potentials and materials within a coherent semantic framework. The approach provides AI-ready simulation data, supports emerging agentic AI workflows, and establishes a generalizable blueprint for knowledge-based mechanical and thermodynamic simulations.", "date": "2026-02-01", "categories": []}
{"id": "2602.02513", "title": "Learning ORDER-Aware Multimodal Representations for Composite Materials Design", "abstract": "Artificial intelligence has shown remarkable success in materials discovery and property prediction, particularly for crystalline and polymer systems where material properties and structures are dominated by discrete graph representations. Such graph-central paradigm breaks down on composite materials, which possess continuous and nonlinear design spaces. General composite descriptors, e.g., fiber volume and misalignment angle, cannot fully capture the fiber distributions that determine microstructural characteristics, necessitating the integration of heterogeneous data sources through multimodal learning. Existing alignment-oriented frameworks have proven effective on abundant crystal or polymer data under discrete, unique graph-property mapping assumptions, but fail to address the highly continuous composite design space under extreme data scarcity. In this work we introduce ORDinal-aware imagE-tabulaR alignment (ORDER), a multimodal pretraining framework that establishes ordinality as a core principle for material representations. ORDER ensures that materials with similar target properties occupy nearby regions in the latent space, which effectively preserves the continuous nature of composite properties and enables meaningful interpolation between sparsely observed designs. We evaluate ORDER on a Nanofiber-reinforced composite dataset and a carbon fiber T700 dataset. ORDER and its variants outperform both alignment-oriented and customized property-aware contrastive baselines across property prediction, cross-modal retrieval, and microstructure generation tasks. We further introduce physics-based ordinal surrogate signals avoiding the need for full property annotation during pretrain. Our work demonstrates learning continuous multimodal features are fundamental for composite materials, and provides a reliable pathway toward data-efficient universal multimodal intelligent systems.", "date": "2026-01-23", "categories": []}
{"id": "2605.18831", "title": "Towards Discovery of Polymers for Insulin Delivery via Physics-Grounded Agentic Workflows", "abstract": "Cold-chain storage limits access to insulin for hundreds of millions of people; a thermally protective patch polymer could help, but the design space is too large for exhaustive experiment. Starting from that problem, we narrow to an agentic workflow: a large language model (LLM) calls physics-based tools through the Model Context Protocol (MCP), searching the discrete PSMILES space under a budget of OpenMM Packmol-matrix evaluations. The LLM acts as an implicit acquisition function conditioned on a persistent \"discovery world\": hypotheses, literature claims, and simulation outcomes updated each iteration. Under matched oracle budgets, the best autonomous campaign reaches an insulin-polymer interaction energy of -2263 kJ/mol, outperforming reinforcement-learning baselines by 68% and Bayesian optimization by 19%. Three independent campaigns converge on one structural motif (dense hydrogen-bond donors and acceptors per repeat unit) while physics checks reject infeasible packings and name-structure mismatches before they steer the next step. The science stage is CPU-bound and runs on commodity hardware. More broadly, the same architecture and workflow designed here applies to other protein-stabilization tasks whenever a tractable screening oracle is available.", "date": "2026-05-12", "categories": []}
{"id": "2605.19124", "title": "Atomistic Modeling of Chemical Disorder in Materials: Bridging Classical Methods and AI-Assisted Approaches", "abstract": "Chemical disorder, originating from the mixed occupation of crystallographic sites by multiple elements, is widespread in alloys, ceramics, and compositionally complex materials, where short- and long-range orderings can strongly influence properties. A central obstacle is the representation gap between experiments and simulations: experiments often report disorder as partial occupancies and ensemble-averaged behaviors, whereas atomistic simulations and AI workflows usually require fully specified configurations. Tackling this gap requires computational methods that convert averaged disorder descriptions into representative configurational ensembles while balancing cost, bias, and fidelity. This challenge has become more urgent in AI-driven computational discovery, where ignoring disorder may cause AI workflows to misrank stability, misjudge novelty, and misdirect experiments with too-idealized representations. This Review highlights how classical and AI-driven methods can bridge this representation gap. We assess the strengths and limitations of approaches spanning mean-field theories, cluster expansion, quasi-random approximations, Monte Carlo, and emerging schemes powered by universal interatomic potentials and generative models. We further highlight how AI can accelerate classical computational schemes by lowering the cost of microstate evaluation, configurational exploration, and atomistic-to-thermodynamic closure. We also emphasize how AI can enable disorder-native capabilities, including workflow triage, ordering-sensitive and alchemical representations, generative models of disordered structures and distributions, and kinetics-aware disorder prediction. Together, this framework outlines a practical roadmap toward disorder-native AI, which can transform chemical disorder from a representational obstacle into a controllable variable for realistic AI-accelerated materials discovery.", "date": "2026-05-18", "categories": []}
{"id": "2505.09203", "title": "InvDesFlow-AL: active learning-based workflow for inverse design of functional materials", "abstract": "Developing inverse design methods for functional materials with specific properties is critical to advancing fields like renewable energy, catalysis, energy storage, and carbon capture. Generative models based on diffusion principles can directly produce new materials that meet performance constraints, thereby significantly accelerating the material design process. However, existing methods for generating and predicting crystal structures often remain limited by low success rates. In this work, we propose a novel inverse material design generative framework called InvDesFlow-AL, which is based on active learning strategies. This framework can iteratively optimize the material generation process to gradually guide it towards desired performance characteristics. In terms of crystal structure prediction, the InvDesFlow-AL model achieves an RMSE of 0.0423 {\\AA}, representing an 32.96% improvement in performance compared to exsisting generative models. Additionally, InvDesFlow-AL has been successfully validated in the design of low-formation-energy and low-Ehull materials. It can systematically generate materials with progressively lower formation energies while continuously expanding the exploration across diverse chemical spaces. These results fully demonstrate the effectiveness of the proposed active learning-driven generative model in accelerating material discovery and inverse design. To further prove the effectiveness of this method, we took the search for BCS superconductors under ambient pressure as an example explored by InvDesFlow-AL. As a result, we successfully identified Li\\(_2\\)AuH\\(_6\\) as a conventional BCS superconductor with an ultra-high transition temperature of 140 K. This discovery provides strong empirical support for the application of inverse design in materials science.", "date": "2025-05-14", "categories": []}
{"id": "2506.03837", "title": "HTSC-2025: A Benchmark Dataset of Ambient-Pressure High-Temperature Superconductors for AI-Driven Critical Temperature Prediction", "abstract": "The discovery of high-temperature superconducting materials holds great significance for human industry and daily life. In recent years, research on predicting superconducting transition temperatures using artificial intelligence~(AI) has gained popularity, with most of these tools claiming to achieve remarkable accuracy. However, the lack of widely accepted benchmark datasets in this field has severely hindered fair comparisons between different AI algorithms and impeded further advancement of these methods. In this work, we present the HTSC-2025, an ambient-pressure high-temperature superconducting benchmark dataset. This comprehensive compilation encompasses theoretically predicted superconducting materials discovered by theoretical physicists from 2023 to 2025 based on BCS superconductivity theory, including the renowned X$_2$YH$_6$ system, perovskite MXH$_3$ system, M$_3$XH$_8$ system, cage-like BCN-doped metal atomic systems derived from LaH$_{10}$ structural evolution, and two-dimensional honeycomb-structured systems evolving from MgB$_2$. The HTSC-2025 benchmark has been open-sourced at https://github.com/xqh19970407/HTSC-2025 and will be continuously updated. This benchmark holds significant importance for accelerating the discovery of superconducting materials using AI-based methods.", "date": "2025-06-04", "categories": []}
{"id": "2512.16085", "title": "Machine Learning Enabled Graph Analysis of Particulate Composites: Application to Solid-state Battery Cathodes", "abstract": "Particulate composites underpin many solid-state chemical and electrochemical systems, where microstructural features such as multiphase boundaries and inter-particle connections strongly influence system performance. Advances in X-ray microscopy enable capturing large-scale, multimodal images of these complex microstructures with an unprecedentedly high throughput. However, harnessing these datasets to discover new physical insights and guide microstructure optimization remains a major challenge. Here, we develop a machine learning (ML) enabled framework that enables automated transformation of experimental multimodal X-ray images of multiphase particulate composites into scalable, topology-aware graphs for extracting physical insights and establishing local microstructure-property relationships at both the particle and network level. Using the multiphase particulate cathode of solid-state lithium batteries as an example, our ML-enabled graph analysis corroborates the critical role of triple phase junctions and concurrent ion/electron conduction channels in realizing desirable local electrochemical activity. Our work establishes graph-based microstructure representation as a powerful paradigm for bridging multimodal experimental imaging and functional understanding, and facilitating microstructure-aware data-driven materials design in a broad range of particulate composites.", "date": "2025-12-18", "categories": []}
{"id": "2605.16612", "title": "PRISMat: Policy-Driven, Permutation-Invariant Autoregressive Material Generation", "abstract": "Rapid identification of candidate materials with target properties has become a key task in materials science. Machine learning has emerged as an alternative to physics-based simulation, offering a faster and cheaper way to filter materials based on their stability and other target properties, reducing the number of candidates that reach the costly synthesis stage. Recently, Large Language Models (LLMs) have been applied to this role, but these models are parameter-heavy and computationally expensive both during training and at inference time, making them unsuitable for high-throughput tasks. This inefficiency stems from both the large over-parameterization of language models and the difficulty of framing material generation as a sequence learning problem. In this paper, we present PRISMat, a cost-effective, permutation-invariant model, which addresses these limitations. We show that PRISMat, despite taking less time for inference, is able to outperform LLMs in generating crystal slabs conditioned on critical materials' surface properties. In targeted material discovery, we achieve mean absolute errors of 0.188 eV/A$^2$ and 2.79 eV for cleavage energy and work function tasks, respectively, reducing the error of the next best model by 4$\\times$.", "date": "2026-05-15", "categories": []}
{"id": "2605.17148", "title": "Evolutionary Extreme Learning Machine of ab-initio Energy Landscapes for Crystal Structure Prediction using Manta Ray Optimization with Levy Flight", "abstract": "The Manta Ray Foraging Optimization algorithm (MRFO) has proven to be a powerful heuristic strategy in the optimal solution of a large number of engineering problems. In this paper, an improvement of MRFO with Levy Flight is suggested for the training of extreme learning machines (ELMs) whose basic model is a Single Layer Feedforward Network (SLFN). The proposed methodology that we called Evolutionary EELM-MRFO-LF for short is implemented to the prediction of unrelaxed and relaxed formation energy compounds relative to ground state crystal structure of pure components in binary systems. EELM-MRFO-LF follows the learning procedure of traditional Evolutionary ELMs in which first MRFO with LF is used to select the input weights and Moore-Penrose (MP) generalized inverse is applied to analytically determine the output weights. Levy Flight trajectory is implemented for increasing the diversity of the population of ELMs against premature convergence and the ability of avoiding getting trapped in a local optima. The performance of the suggested EELM-MRFO-LF is compared with other well-known nature-inspired algorithms under similar conditions.", "date": "2026-05-16", "categories": []}
{"id": "2605.18033", "title": "Real-time Multi-instrument Autonomous Discovery of Novel Phase-change Memory Materials", "abstract": "Autonomous labs enable the integration of automated experiment execution, data analysis and decision making. The main challenge remains the integration of diverse data streams from multiple instruments, where the data is often heterogeneous and unsynchronized. The standard learning process of undetermined synthesis-process-structure-property relationships (SPSPR) usually relies on post-experiment analysis after data is fully collected, not during live experiments, and decision making is carried out independently across characterization equipment. Here, we demonstrate the Multi-instrument Autonomous Discovery (MAD) framework -- combining structural property mapping and functional property optimization simultaneously in a closed-loop manner. As an example, we applied MAD to phase change memory (PCM) materials, and, in particular on the Mn-Sb-Te ternary, a previously unexplored materials system for PCM. A multi-output model is employed to merge data from x-ray diffraction (XRD) and electrical resistance measurements simultaneously through a co-regionalization kernel that models the relationship between them. The output probabilistic posterior and uncertainty quantification facilitate decision making with shared knowledge, while the goals are different across tasks. We aimed to maximize the knowledge of crystal structure distribution using non-negative matrix factorization (NMF), while in parallel, we find the composition with the maximum resistance value, an important figure of merit for PCM. Leveraging MAD, we found promising electrical PCMs and identified the SPSPR within 25 closed-loop iterations, corresponding to a seven-fold speed-up. The framework opens a new path of study in large-scale autonomous facilities, where future experiments can be run in parallel together, not independently.", "date": "2026-05-18", "categories": []}
{"id": "2605.14344", "title": "CrystalReasoner: Reasoning and RL for Property-Conditioned Crystal Structure Generation", "abstract": "Generative modeling has emerged as a promising approach for crystal structure discovery. However, existing LLM-based generative models struggle with low-level atomic precision, while diffusion-based methods fall short in integrating high-level scientific knowledge. As a result, generated structures are often invalid, unstable, or do not possess desirable properties. To address this gap, we propose CrystalReasoner (CrysReas), an end-to-end LLM framework that generates crystal structures from natural language instructions through reasoning and alignment. CrysReas introduces physical priors as thinking tokens, which include crystallographic symmetry, local coordination environments and predicted physical properties before generating atomic coordinates. This bridges the gap between natural language and 3D structures. CrysReas then employs reinforcement learning (RL) with a multi-objective, dense reward function to align generation with physical validity, chemical consistency, and thermodynamic stability. For property-conditioned tasks, we design task-specific reward functions and train specialized models for discrete constraints (e.g., space group) and continuous properties (e.g., elasticity, thermal expansion). Empirical results demonstrate that compared to prior works and baselines without thinking traces or RL, CrysReas obtains better performance on diverse metrics, triples S.U.N. ratio, and achieves better performance for property conditioned generation. CrysReas also exhibits adaptive reasoning, increasing reasoning lengths as the number of atoms increases. Our work demonstrates the potential of leveraging thinking traces and RL for generating valid, stable, and property-conditioned crystal structures.", "date": "2026-05-14", "categories": []}
{"id": "2605.15630", "title": "Reweighting free energy profiles between universal machine learning interatomic potentials for fast consensus building", "abstract": "Free energy profiles serve as a fundamental bridge between microscopic atomic fluctuations and macroscopic thermodynamic observables. Estimating the free energy profile along a reaction coordinate, referred to as the potential of mean force (PMF), with density functional theory (DFT) accuracy is computationally expensive. Universal machine learning interatomic potentials (MLIPs) drastically reduce this cost, but their accuracy is strongly determined by their training data and hence can be uncertain for a given system. In this work, we present a systematic and scalable framework for reweighting PMFs, initially sampled with a single 'source' MLIP, across a representative suite of target MLIPs. Because traditional direct exponential reweighting fails for large system sizes due to low phase-space overlap between potentials, we deploy robust analytical corrections. Applying this to a complex 601-atom system of Li$^+$ transport in a nanoconfined electrolyte, we demonstrate that a mean energy-gap approximation effectively bypasses statistical collapse, producing a highly stable PMF matching the target PMF. Using this approach, we recover high-fidelity target thermodynamics across multiple DFT reference levels (PBE+D3, PBE-sol, r$^2$SCAN,r$^2$SCAN-D4) at a fraction of the computational cost of full simulations. Furthermore, thermodynamic analysis reveals that the studied MLIPs partition into two distinct clusters driven by their training data. Our reweighting framework successfully recovers target thermodynamic properties--specifically, reaction and activation free energies--even when the phase-space overlap between potentials is critically low. Ultimately, this approach establishes a vital diagnostic protocol to achieve affordable cross-model consensus on materials chemistry properties without redundant, resource-intensive simulations.", "date": "2026-05-15", "categories": []}
{"id": "2605.16230", "title": "Universal Magnetic Structure Prediction from Atomic Coordinates with Near-Experimental Accuracy", "abstract": "Magnetic order is a fundamental property of materials, governing collective behavior and enabling a broad range of functionalities. Yet magnetic structure remains difficult to determine: experiments are costly and specialized, while first-principles methods often struggle with the noncollinear and incommensurate orders found in real materials. Here we introduce magnetic structure network (MSN), an E(3) equivariant graph neural network that predicts both collinear and non-collinear magnetic structures directly from atomic crystal structures, trained directly on experimentally determined structures from MAGNDATA. By proposing the primitive modulated structure representation (PMSR), we are able to encode commensurate and incommensurate structures in a unified way without symmetry assumptions. The model achieves strong performance across all modulation components and reconstructs experimental magnetic structures with high fidelity. Our approach provides a scalable framework for rapid magnetic structure prediction and opens a route to data-driven discovery of magnetic materials.", "date": "2026-05-15", "categories": []}
{"id": "2605.14527", "title": "Lang2MLIP: End-to-End Language-to-Machine Learning Interatomic Potential Development with Autonomous Agentic Workflows", "abstract": "Developing machine learning interatomic potentials (MLIPs) for complex materials systems remains challenging because it requires expertise in atomistic simulations, machine learning, and workflow design, as well as iterative active learning procedures. Existing automated pipelines typically assume a fixed sequence of stages or depend on domain experts, which limits their adaptability to heterogeneous materials systems where the optimal curriculum is not known in advance. To lower the barrier to developing MLIPs for non-experts, we propose Lang2MLIP, a multi-agent framework that takes natural-language input and formulates end-to-end MLIP development as a sequential decision-making problem solved by large language models (LLMs). At each step, a decision-making agent observes the current dataset, model, evaluation results, and execution log, and then automatically selects an appropriate action to improve the model. This removes the need for a predefined pipeline and enables the agent to self-correct by revisiting earlier subsystems when new failures arise. We evaluate this approach on a solid electrolyte interphase (SEI) system with multiple components and interfaces. These results suggest that LLM-based multi-agent systems are a promising direction for automating MLIP development and making it more accessible to non-experts.", "date": "2026-05-14", "categories": []}
{"id": "2605.14671", "title": "Agentic Design of Compositional Descriptors via Autoresearch for Materials Science Applications", "abstract": "Autoresearch offers a flexible paradigm for automating scientific tasks, in which an AI agent proposes, implements, evaluates, and refines candidate solutions against a quantitative objective. Here, we use composition-based materials-property prediction to test whether such agents can perform a task beyond model selection and hyperparameter optimization: the design of input descriptors. We introduce Automat, an autoresearch framework where a coding agent based on a large language model generates composition-only descriptors for chemical compounds and evaluates them using a random forest workflow. The agent is restricted to information derivable from chemical formulas and iteratively proposes, implements, and tests chemically motivated descriptor strategies. We apply Automat, with OpenAI Codex using GPT-5.5 as the coding agent, to the prediction of experimental band gaps in inorganic materials and Curie temperatures in ferromagnetic compounds. In both tasks, Automat improves over fractional-composition, Magpie, and combined fractional-composition/Magpie baselines, while producing descriptor families that are chemically interpretable. These results provide a demonstration that autoresearch agents can generate competitive, task-specific materials descriptors without manual feature engineering during the run. They also reveal current limitations, including descriptor redundancy, sensitivity to greedy feature expansion, and the need for explicit complexity control, descriptor pruning, and more sophisticated search strategies.", "date": "2026-05-14", "categories": []}
{"id": "2602.22251", "title": "Zatom-1: Towards a Multimodal Foundation Model for 3D Molecules and Materials", "abstract": "General-purpose 3D modeling in chemistry encompasses molecules and materials, requiring both generative and predictive capabilities. However, most existing AI approaches are optimized for a single domain (molecules or materials) and a single task (generation or prediction), which limits representation sharing and transfer. We introduce Zatom-1, a cross-domain, general-purpose model architecture that unifies generative and predictive learning of 3D molecules and materials. Zatom-1 is a deliberately simplified Transformer trained with a multimodal flow matching objective that jointly models discrete atom types and continuous 3D geometries. This approach supports scalable pretraining with predictable gains as model capacity increases, while enabling fast and stable sampling. We use cross-domain generative pretraining as a universal initialization for downstream multi-task prediction of properties, energies, and forces. Empirically, Zatom-1 outperforms or competes with specialized baselines on both multi-task generative and predictive benchmarks in data-controlled settings, while improving generative inference speed by more than an order of magnitude. Our experiments demonstrate positive predictive transfer between data domains from joint generative pretraining: modeling materials during generative pretraining improves molecular property prediction accuracy. Open-source code and model weights are freely available at https://github.com/Zatom-AI/zatom.", "date": "2026-02-24", "categories": []}
{"id": "2605.12823", "title": "Hessian Matching for Machine-Learned Coarse-Grained Molecular Dynamics", "abstract": "Coarse-grained (CG) molecular dynamics enables simulations of atomic systems such as biomolecules at timescales inaccessible to all-atom (AA) methods, but existing CG neural potentials trained via force matching capture only the gradient of the free-energy surface, leaving its curvature unconstrained. We introduce a framework that augments force matching with stochastic Hessian-vector product (HVP) matching, instilling second-order curvature information into CG potentials without constructing the full Hessian. We derive a decomposition of the target CG Hessian into a model-independent projected AA Hessian, precomputed once before training, and a model-dependent covariance correction computed online at negligible cost. We construct an unbiased stochastic estimator of the Hessian-matching objective by using random probe vectors. We evaluate our method by comparing against force matching on a benchmark of nine fast-folding proteins unseen during training. HVP matching outperforms plain force matching on 8 of 9 proteins on slow-mode metrics, with reductions of up to 85% in the Kullback--Leibler divergence between the CG and reference distributions along the slowest collective mode of the largest protein. Our results demonstrate that higher-order physical supervision is a practical path to more accurate and transferable CG potentials for biomolecular simulation.", "date": "2026-05-12", "categories": []}
{"id": "2605.13297", "title": "PaMM: Periodic Motif Memory for Atomistic Models with an Explicit Local-Structure Interface", "abstract": "Periodic crystals repeatedly instantiate similar local coordination motifs across translated cells and chemically related structures, but current equivariant atomistic models usually encode these patterns only implicitly in dense edge features. We introduce PaMM, a periodic motif memory that augments the UMA eSCN-MD edge encoder with explicit pair and triplet lookup features. Pair motifs are keyed by $(Z_j, Z_i, b_r)$ and triplet motifs by $(Z_j, Z_i, Z_k, b_\\theta)$, hashed into fixed-size tables and fused with the baseline edge representation through lightweight gate-only and affine-equipped variants. We evaluate PaMM in a matched UMA-S OMAT setting and focus on a narrow question: whether explicit motif memory helps at a fixed intermediate training budget. At the 10k-step checkpoint, both PaMM variants improve over the plain baseline; gate-only gives the best energy MAE, while the affine-equipped variant gives the best force MAE. A matched 20k follow-up keeps the same operating-point picture. Aligned controls show that the gain weakens for pair-only, triplet-only, random-bucket, and parameter-matched MLP alternatives, suggesting that the benefit is tied to structured pair/triplet organization rather than generic added capacity. A within-OMAT24 source-family evaluation also shows small but consistent gains across held-out generation families. We therefore make a focused claim: in the studied UMA-S + OMAT regime, explicit pair/ triplet motif memory is a useful inductive bias for periodic atomistic modeling. We do not claim broad cross-dataset transfer, a uniquely preferred fusion variant, or strong scientific interpretability beyond a more inspectable local-structure interface.", "date": "2026-05-13", "categories": []}
{"id": "2605.12194", "title": "Probing Non-Equilibrium Grain Boundary Dynamics with XPCS and Domain-Adaptive Machine Learning", "abstract": "Grain-boundary (GB) dynamics control the stability, mechanical, and functional response of nanocrystalline materials, but direct experimental access to their slow non-equilibrium motion has been limited. Here we establish X-ray photon correlation spectroscopy (XPCS), combined with domain-adaptive machine learning, as a quantitative probe of GB dynamics. Temperature- and grain-size-dependent two-time XPCS measurements in nanocrystalline silicon reveal pronounced departures from time-translation invariance, showing that GB relaxation can remain far from equilibrium over experimental timescales. However, direct extraction of quantitative physical information from these high-dimensional, noisy fluctuation maps faces a significant challenge. To overcome this barrier, we develop a semi-supervised learning framework that transfers physical parameter labels from continuum simulations to unlabeled experimental XPCS maps through domain-adaptive representation alignment. This AI-augmented approach enables the extraction of key kinetic parameters, including bulk diffusivity, GB stiffness, and effective GB concentration, directly from experimental XPCS measurements. Our results show how machine learning can transform indirect fluctuation signals into quantitative materials dynamics, providing a general route to study non-equilibrium defect motion in solids.", "date": "2026-05-12", "categories": []}
{"id": "2512.11077", "title": "A probabilistic framework for crystal structure denoising, phase classification, and order parameters", "abstract": "Atomistic simulations generate large volumes of noisy structural data, yet extracting phase labels and continuous order parameters (OPs) in a robust and general manner remains challenging. Existing tools are often specialized to a limited set of prototypes and split thermal-noise removal, phase classification, and OP construction into separate steps. Here we present a unified probabilistic framework for analyzing noisy atomic configurations with respect to known crystal prototypes. The model predicts per-atom, per-prototype logits and aggregates them into a scalar log-probability (logP) landscape over atomic coordinates. Its gradient defines a conservative denoising field, while the logits provide local phase labels, prototype-resolved OPs, and ambiguity measures through logit margins. We train on AFLOW-mapped crystalline structures from the Materials Project with synthetic positional and elastic perturbations, then test extrapolation to stronger noise, finite-temperature disorder, point defects, water--ice coexistence, binary polymorphs, and shock-compressed Ti. A single differentiable scalar model recovers prototype identity after denoising, tracks smooth transformations such as Bain and Burgers paths, and exposes low-confidence regions near defects and phase boundaries. This provides an integrated and extensible tool for analyzing complex atomistic simulations.", "date": "2025-12-11", "categories": []}
{"id": "2605.08098", "title": "Reinforcement learning for inverse structural design and rapid laser cutting of kirigami prototypes", "abstract": "Kirigami is an increasingly useful fabrication method to produce shape-programmable metamaterial structures. However, inverse design remains difficult because deployment is nonlinear, and feasible cut layouts must satisfy discrete compatibility rules, avoid overlap, and map one target shape to valid designs. We present RL-Kirigami, an inverse design framework that combines optimal-transport conditional flow matching (OT-CFM) with reinforcement learning to generate compatible ratio fields for compact reconfigurable parallelogram quad kirigami. A marching decoder enforces global geometric compatibility, and Group Relative Policy Optimization (GRPO) aligns the generator with nondifferentiable rewards for silhouette matching, feasibility, and ratio-field regularity. Across procedurally generated target shape instances, a single sample from the pretrained OT-CFM prior reached $94.2%$ sIoU and outperformed solver baselines while reducing forward simulator evaluations from hundreds to 1. GRPO improved accuracy to $94.91%$ sIoU and, with regularity included, reduced $\\mathrm{TV}(\\mathbf{x})$ from 0.95 to 0.81 while maintaining $94.83%$ sIoU. Generated layouts were exported to DXF and laser-cut in $50~\\mu\\mathrm{m}$ polymeric sheets to produce deployable prototypes in $8.0 \\pm 1.0$ minutes per part. These results support a manufacturing-aware inverse design workflow for deployable kirigami metamaterials under hard geometric feasibility constraints.", "date": "2026-04-16", "categories": []}
{"id": "2605.08103", "title": "Crystal Fractional Graph Neural Network for Energy Prediction of High-Entropy Alloys", "abstract": "High-entropy alloys (HEAs) have attracted growing attention for their exceptional mechanical and thermal properties arising from complex atomic configurations. In this paper, we propose crystal fractional graph neural network for predicting the energy of high-entropy alloys by explicitly integrating both local atomic environments and global compositional information. The model consists of three components: a crystal graph neural network, which employs graph attention network layers to learn local interactions among 16 on-site atoms within the crystal lattice; fractional neural network, a fully connected network that embeds the global fraction of constituent elements; and feature fusion neural network, which fuses the outputs of the two submodels to predict the total crystal energy. We train the model on a dataset of 1,049 crystal structures and validate it on 198 quaternary structures, optimizing all hyperparameters via Optuna. Our results show that our model achieves an RMSE comparable to first-principles calculations and maintains high accuracy even for low-energy configurations. However, the model exhibits limitations in handling large crystal cells, which we aim to address in future work to extend its applicability to more complex systems.", "date": "2026-04-26", "categories": []}
{"id": "2605.08255", "title": "Can LLMs Predict Polymer Physics Just by Reading Synthesis and Processing Prose?", "abstract": "Can large language models predict physical and mechanical polymer properties simply by reading unstructured scientific prose? Polymer performance is rarely determined by chemical structure alone; identical nominal polymers can exhibit drastically different behaviors depending on their synthesis route, processing history, morphology, and testing conditions. Yet, state-of-the-art polymer property models typically rely on structure-only representations -- such as SMILES or molecular graphs -- which strip away this vital experimental context. In this work, we introduce \\textbf{PolyLM}, a natural-language-only, process- and condition-aware framework that predicts materials performance directly from full-text literature. By circumventing structural inputs entirely, PolyLM preserves the nuanced, unstructured descriptions of synthesis and processing reported by domain scientists. To train this framework, we curated an unprecedented, literature-scale dataset encompassing 185,000 scientific papers and over 276,400 unique polymer samples across 22 physical, mechanical, and thermal properties. We fine-tuned a massive 9-billion-parameter language model (Qwen3.5-9B) using Low-Rank Adaptation (LoRA) and task-level uncertainty weighting. Evaluated on 68,283 held-out observations, the model achieves remarkably high predictive accuracy, establishing new state-of-the-art benchmarks for complex properties. Across the 22 diverse targets, the model achieves a median $R^2$ of 0.74, with predictions for key thermal, mechanical, and physicochemical properties frequently surpassing an $R^2$ of 0.80. These results unequivocally demonstrate that natural language is a powerful, highly scalable interface for realistic materials performance prediction.", "date": "2026-05-07", "categories": []}
{"id": "2605.08960", "title": "CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models", "abstract": "Crystal generative models mainly learn what stable crystals look like, with little explicit supervision for what makes them stable. We reveal a substantial representation gap between state-of-the-art crystal generative models and pretrained universal machine learning interatomic potentials (MLIPs) via energy probing, and show this gap can be closed by a simple training-time alignment. We propose Crystal REPresentation Alignment (CrystalREPA), a plug-and-play framework that aligns the atom-wise hidden states of generative encoders with frozen MLIP representations through an element-aware contrastive objective, transferring stability-aware atomistic priors with marginal training overhead and no additional inference cost. Across three generative frameworks, ten MLIP teachers, and two benchmark datasets, CrystalREPA consistently improves the thermodynamic stability, structural validity, and structural fidelity of generated crystals. Equally important, we find that an MLIP's transfer effectiveness is poorly predicted by its accuracy on standard leaderboards (e.g., Matbench Discovery) but strongly predicted by the distinguishability of its atom-wise representation space, yielding a practical, accuracy-independent criterion for selecting MLIP teachers for generative transfer.", "date": "2026-05-09", "categories": []}
{"id": "2605.10115", "title": "Generating Symmetric Materials using Latent Flow Matching", "abstract": "Tackling the task of materials generation, we aim to enhance the previously proposed All-atom Diffusion Transformer (ADiT) by introducing SymADiT, a symmetry-aware variant. To do so, we use a representation of materials based on Wyckoff positions. We follow ADiT and perform generative modelling in latent space, adapted to our symmetry-aware representation. By forcing the output of the generative model to adhere to the symmetry restrictions imposed by the generated crystal's space group and each atom's Wyckoff-position, the generated materials exhibit more realistic symmetry properties. We benchmark our method against both symmetry-aware and symmetry-agnostic models for materials generation and show competitive performance, generating stable, symmetric materials with a simple Transformer architecture.", "date": "2026-05-11", "categories": []}
{"id": "2605.05866", "title": "XDecomposer: Learning Prior-Free Set Decomposition for Multiphase X-ray Diffraction", "abstract": "Multiphase powder X-ray diffraction (PXRD) analysis remains a fundamental bottleneck in structure identification, as real-world synthesis often produces complex mixtures whose constituent phases (components) cannot be reliably disentangled. While recent advances in representation-based crystal retrieval and generation suggest the possibility of inferring structures directly from PXRD, existing approaches largely assume single-phase inputs and break down in multiphase settings. Here, we present XDecomposer, a prior-free framework for joint decomposition and identification of multiphase XRD patterns without requiring candidate phase lists, structural templates, or prior knowledge of phase number. We formulate multiphase diffraction analysis as a set prediction problem, where the model infers an unordered set of phase-resolved components, their mixture proportions, and corresponding structural representations within a unified architecture. A phase-query-driven decomposition mechanism, together with diffraction-consistent physical reconstruction, enables accurate source separation while preserving crystallographic fidelity. Extensive experiments on both simulated and experimental datasets show that XDecomposer substantially improves reconstruction accuracy and phase identification across diverse chemical systems, while maintaining strong generalization to unseen mixtures. These results provide a practical route toward data-driven, source-resolved multiphase XRD analysis and reduce long-standing dependence on prior-guided iteratively phase matching. The code is openly available at https://github.com/Licht0812/XDecomposer", "date": "2026-05-07", "categories": []}
{"id": "2603.03511", "title": "Orbital Transformers for Predicting Wavefunctions in Time-Dependent Density Functional Theory", "abstract": "We aim to learn wavefunctions simulated by time-dependent density functional theory (TDDFT), which can be efficiently represented as linear combination coefficients of atomic orbitals. In real-time TDDFT, the electronic wavefunctions of a molecule evolve over time in response to an external excitation, enabling first-principles predictions of physical properties such as optical absorption, electron dynamics, and high-order response. However, conventional real-time TDDFT relies on time-consuming propagation of all occupied states with fine time steps. In this work, we propose OrbEvo, which is based on an equivariant graph transformer architecture and learns to evolve the full electronic wavefunction coefficients across time steps. First, to account for external field, we design an equivariant conditioning to encode both strength and direction of external electric field and break the symmetry from SO(3) to SO(2). Furthermore, we design two OrbEvo models, OrbEvo-WF and OrbEvo-DM, using wavefunction pooling and density matrix as interaction method, respectively. Motivated by the central role of the density functional in TDDFT, OrbEvo-DM encodes the density matrix aggregated from all occupied electronic states into feature vectors via tensor contraction, providing a more intuitive approach to learn the time evolution operator. We adopt a training strategy specifically tailored to limit the error accumulation of time-dependent wavefunctions over autoregressive rollout. To evaluate our approach, we generate TDDFT datasets consisting of 5,000 different molecules in the QM9 dataset and 1,500 molecular configurations of the malonaldehyde molecule in the MD17 dataset. Results show that our OrbEvo model accurately captures quantum dynamics of excited states under external field, including time-dependent wavefunctions, time-dependent dipole moment, and optical absorption spectra.", "date": "2026-03-03", "categories": []}
{"id": "2605.05746", "title": "Polarizable atomic multipoles for learning long-range electrostatics", "abstract": "Long-range electrostatics and polarization remain central obstacles to extending machine learning interatomic potentials (MLIPs) to ionic, polar, and interfacial systems. Here, we introduce a semi-local framework for learning electrostatics from energies and forces using polarizable atomic multipoles. Local equivariant descriptors predict environment-dependent latent monopoles, dipoles, and quadrupoles, while residual non-local charge transfer and polarization are captured by non-self-consistent linear response in induced charges and dipoles. Across four diverse benchmarks and four short-range MLIP architectures, the multipole hierarchy and response terms systematically improve potential energy surface accuracy, with the largest gains in systems where long-range effects are essential. More importantly, the learned latent variables recover physically meaningful electrical responses: accurate Born effective charge tensors, emergent polarizabilities, infrared spectra in close agreement with experiments, and semi-quantitative Raman spectra for bulk water and hybrid MAPbI$_3$ perovskite. This systematically improvable, physically transparent framework enables MLIPs trained on standard energy and force labels to predict polarization-sensitive observables.", "date": "2026-05-07", "categories": []}
{"id": "2605.06641", "title": "GlazyBench: A Benchmark for Ceramic Glaze Property Prediction and Image Generation", "abstract": "Developing ceramic glazes is a costly, time-consuming process of trial and error due to complex chemistry, placing a significant burden on independent artists. While recent advances in multimodal AI offer a modern solution, the field lacks the large-scale datasets required to train these models. We propose GlazyBench, the first dataset for AI-assisted glaze design. Comprising 23,148 real glaze formulations, GlazyBench supports two primary tasks: predicting post-firing surface properties, such as color and transparency, from raw materials, and generating accurate visual representations of the glaze based on these properties. We establish comprehensive baselines for property prediction using traditional machine learning and large language models, alongside image generation benchmarks using deep generative and large multimodal models. Our experiments demonstrate promising yet challenging results. GlazyBench pioneers a new research direction in AI-assisted material design, providing a standardized benchmark for systematic evaluation.", "date": "2026-05-07", "categories": []}
{"id": "2605.05104", "title": "Building informative materials datasets beyond targeted objectives", "abstract": "Materials science data collection can be expensive, making the reuse and long-term utility of datasets critical important for future discovery campaigns. In practice, researchers prioritize a subset of properties due to research interests. However, ignoring a subset of outcomes in data collection campaigns potentially generate datasets poorly suited for future learning tasks. Here, we present a framework for dataset construction that maximizes informativeness for target properties of interest while preserving performance on untargeted ones. Our approach uses diversity-aware selection to ensure broad coverage of the materials space. In noisy experimental dataset construction, we find that without our diversity-aware framework, prediction performance on untargeted properties can degrade by up to 40% relative to random sampling, whereas applying our framework yields improvements of up to 10% . For targeted properties, performance can degrade with respect to random sampling by up to 12.5% without diversity, while our framework achieves gains of up to 25%. Incorporating diversity into dataset construction not only preserves informativeness for the targeted properties, but also improves materials coverage for potential future objectives. As a result, the constructed datasets remain broadly informative across considered and unconsidered outcomes, ensuring unbiased quality entries and mitigating cold-start limitations in subsequent modeling and discovery campaigns.", "date": "2026-05-06", "categories": []}
{"id": "2604.23758", "title": "Agentic Fusion of Large Atomic and Language Models to Accelerate Superconductor Discovery", "abstract": "Artificial intelligence has accelerated materials discovery through high-throughput prediction and generation, yet the decision problem remains a formidable bottleneck. While current AI systems readily propose millions of candidates, navigating the decision regarding a viable experimental target requires resolving multi-dimensional judgments across atomic-scale numerical computation and high-level semantic reasoning. Here we present ElementsClaw, an agentic framework for materials discovery that orchestrates a suite of Large Atomic Model (LAM) tools finetuned from our proposed 1-billion-parameter model Elements for numerical computation, while leveraging Large Language Models (LLMs) for semantic reasoning. Applied to superconductors, ElementsClaw rediscovers 66 experimentally verified superconductors that are absent from the standard SuperCon3D database. Scaling to 2.4 million equilibrium crystals, ElementsClaw identifies 68,000 high-confidence candidates in just 28 GPU hours (https://developer.damo-academy.com/material), expanding known superconducting space by orders of magnitude compared to datasets curated over decades. Guided by the agent's reasoning, we experimentally synthesize and verify four novel superconductors: the motif-guided Zr$_3$ScRe$_8$ ($T_c$ = 6.5 K), the de novo generated HfZrRe$_4$ ($T_c$ = 5.9 K), the structurally reinterpreted Zr$_4$VRe$_7$ ($T_c$ = 3.5 K), and the database-latent Hf$_{21}$Re$_{25}$ ($T_c$ = 2.5 K). Together, our results establish a knowledge integrated, autonomously orchestrated, and experimentally grounded paradigm for materials discovery.", "date": "2026-04-26", "categories": []}
{"id": "2605.02871", "title": "Multi-fidelity surrogates for mechanics of composites: from co-kriging to multi-fidelity neural networks", "abstract": "Composite materials exhibit strongly hierarchical and anisotropic properties governed by coupled mechanisms spanning constituents, plies, laminates, structures, and manufacturing history. This intrinsic complexity makes predictive modeling of composites expensive, because repeated experiments and high-fidelity simulations are needed to cover large design spaces of material, structure, and manufacturing. Multi-fidelity surrogate modeling addresses this challenge by combining abundant, less expensive data with limited high-accuracy data to recover reliable high-fidelity predictions. This review presents a structured overview of multi-fidelity modeling for composite mechanics, covering Gaussian-process or Kriging-based methods, including co-Kriging, coregionalization models, autoregressive formulations, nonlinear autoregressive Gaussian processes, multi-fidelity deep Gaussian processes, and multi-fidelity neural networks. Their distinctions are examined in terms of cross-fidelity correlation, discrepancy representation, uncertainty quantification, and scalability. Selected examples of their applications to composites are introduced according to the roles that multi-fidelity surrogates play in engineering problems, including forward prediction for rapid exploration of material design spaces, inverse optimization for composite parameter identification and design search under limited high-fidelity access, and workflow integration, where heterogeneous data sources, constraints, and validation requirements determine model utility. Open question discussions highlight recurring challenges specific to composites, such as regime-dependent fidelity gaps associated with nonlinear damage and manufacturing history, mismatches between simulations and experiments, and uncertainty propagation across multi-fidelity models.", "date": "2026-05-04", "categories": []}
{"id": "2507.09001", "title": "Surprisingly High Redundancy in Electronic Structure Data Across Materials Explained by Low Intrinsic Dimensionality", "abstract": "Machine learning (ML) models for electronic structure typically rely on large datasets generated by computationally expensive Kohn-Sham density functional theory calculations, as it is not known a priori which portions of the data are essential for accurate learning. Here, we reveal significant redundancies in electronic structure datasets across diverse material systems and attribute them to the low intrinsic dimensionality of the underlying data. We show that even random pruning can substantially reduce dataset size with minimal degradation in predictive accuracy. Moreover, a state-of-the-art coverage-based pruning strategy that samples data across all learning difficulties preserves chemical accuracy and model generalizability while using up to two orders of magnitude less data and reducing training time by a factor of three or more. We further demonstrate that the essential electronic structure information lies on a low-dimensional, non-linear manifold, providing a geometric explanation for the observed prunability. These observations are consistent with the predominance of local atomic environments in determining electronic properties, as suggested by nearsightedness arguments, and indicate that large-scale datasets may contain highly overlapping information. Our findings challenge the prevailing assumption that such extensive datasets are necessary for accurate ML-based electronic structure predictions and open a path toward identifying minimal, representative datasets for each material class.", "date": "2025-07-11", "categories": []}
{"id": "2510.18900", "title": "Foundation Models for Discovery and Exploration in Chemical Space", "abstract": "Accurate prediction of atomistic, thermodynamic, and kinetic properties from molecular structures underpins materials innovation. Existing computational and experimental approaches lack the scalability required to navigate chemical space efficiently. Scientific foundation models trained on large unlabelled datasets offer a path towards navigating chemical space across application domains. Here, we develop MIST, a family of molecular foundation models with up to an order of magnitude more parameters and data than prior works. Trained using a novel tokenizer, Smirk, which comprehensively captures nuclear, electronic, and geometric information, MIST learns a diverse range of molecules. MIST models have been fine-tuned to predict more than 400 structure-property relationships and have been shown to match or exceed state-of-the-art performance across diverse benchmarks, from physiology to electrochemistry. We demonstrate the ability of these models to solve real-world problems across chemical space from multiobjective electrolyte solvent screening to stereochemical reasoning for organometallics and mixture property prediction. The clearest demonstration of a foundation model is its ability to solve problems that were neither explicit targets of training nor central to the intentions of its developers. We identify olfactory perception mapping as such a problem, and show that MIST accurately predicted scent profiles and learned a hierarchical representation of olfactory space consistent with hyperbolic geometry. We formulated hyperparameter aware Bayesian neural scaling laws which eliminate the need for hyperparameter sweeps at every scale, making training large compute-optimal models feasible on a limited compute budget. The methods and findings presented here represent a significant step towards accelerating materials discovery, design, and optimization using foundation models.", "date": "2025-10-20", "categories": []}
{"id": "2512.09169", "title": "AI-Driven Expansion and Application of the Alexandria Database", "abstract": "We present a novel multi-stage workflow for computational materials discovery that achieves a 99% success rate in identifying compounds within 100 meV/atom of thermodynamic stability, with a threefold improvement over previous approaches. By combining the Matra-Genoa generative model, Orb-v2 universal machine learning interatomic potential, and ALIGNN graph neural network for energy prediction, we generated 119 million candidate structures and added 1.3 million DFT-validated compounds to the ALEXANDRIA database, including 74 thousand new stable materials. The expanded ALEXANDRIA database now contains 5.8 million structures with 175 thousand compounds on the convex hull. Predicted structural disorder rates (37-43%) match experimental databases, unlike other recent AI-generated datasets. Analysis reveals fundamental patterns in space group distributions, coordination environments, and phase stability networks, including sub-linear scaling of convex hull connectivity. We release the complete dataset, including sAlex25 with 14 million out-of-equilibrium structures containing forces and stresses for training universal force fields. We demonstrate that fine-tuning a GRACE model on this data improves benchmark accuracy. All data, models, and workflows are freely available under Creative Commons licenses.", "date": "2025-12-09", "categories": []}
{"id": "2605.00313", "title": "Beyond Structure: Revolutionising Materials Discovery via AI-Driven Synthesis Protocol-Property Relationships", "abstract": "The current structure-centric paradigm in artificial intelligence (AI)-driven materials discovery, despite delivering thousands of candidate structures, is stalling at a critical barrier: the synthesizability gap. We argue that closing this gap demands a pivot to a synthesis-first paradigm in which executable synthesis protocols, not just atomic configurations, are treated as primary design variables. We outline a roadmap built on three pillars: (i) representing synthesis procedures as machine-readable protocols, (ii) deploying generative and inverse-design models to propose actionable reaction pathways and recipes, and (iii) integrating closed-loop optimisation to refine protocols against experimental realities and sustainability constraints. Framed in terms of the causal backbone P->X->y from protocol P to structure X and properties y, this perspective sets out methodological building blocks, standards needs and self-driving laboratory (SDL) integration strategies to accelerate reproducible, data-first materials discovery.", "date": "2026-05-01", "categories": []}
{"id": "2605.00639", "title": "Born-Qualified: An Autonomous Framework for Deploying Advanced Energy and Electronic Materials", "abstract": "Autonomous science is transforming how we discover materials and chemical systems for advanced energy technologies. However, many initially promising systems never reach deployment. This \"valley of death\" stems from optimization that prioritizes laboratory metrics over industrial viability. We propose a new strategy: \"born-qualified\" autonomous development, which embeds manufacturability, cost, and durability constraints from the outset. This approach is enabled by four pillars, including the development of multi-objective metrics, causal models, a modular infrastructure, and embedding manufacturing in the discovery loop. Realizing this vision will require sustained, community-wide commitment, but the potential return on that investment is commensurate with the scale of the challenge.", "date": "2026-05-01", "categories": []}
{"id": "2508.07798", "title": "Generative Inversion for Property-Targeted Materials Design: Application to Shape Memory Alloys", "abstract": "The design of shape memory alloys (SMAs) with high transformation temperatures and large mechanical work output remains a longstanding challenge in functional materials engineering. Here, we introduce a data-driven framework based on generative adversarial network (GAN) inversion for the inverse design of high-performance SMAs. By coupling a pretrained GAN with a property prediction model, we perform gradient-based latent space optimization to directly generate candidate alloy compositions and processing parameters that satisfy user-defined property targets. The framework is experimentally validated through the synthesis and characterization of five NiTi-based SMAs. Among them, the Ni$_{49.8}$Ti$_{26.4}$Hf$_{18.6}$Zr$_{5.2}$ alloy achieves a high transformation temperature of 404 $^\\circ$C, a large mechanical work output of 9.9 J/cm$^3$, a transformation enthalpy of 43 J/g , and a thermal hysteresis of 29 {\\deg}C, outperforming existing NiTi alloys. The enhanced performance is attributed to a pronounced transformation volume change and a finely dispersed of Ti$_2$Ni-type precipitates, enabled by sluggish Zr and Hf diffusion, and semi-coherent interfaces with localized strain fields. This study demonstrates that GAN inversion offers an efficient and generalizable route for the property-targeted discovery of complex alloys.", "date": "2025-08-11", "categories": []}
{"id": "2603.06526", "title": "Predicting Atomistic Transitions with Transformers", "abstract": "Accurate knowledge of the atomistic transition pathways in materials and material surfaces is crucial for many material science problems. However, conventional simulation techniques used to find these transitions are extremely computationally intensive. Even with large-scale, accelerated material simulations, the computational cost constrains the applicable domain in practice. Machine learning models, with the potential to learn the complex emergent behaviors governing atomistic transitions as a fast surrogate model, have great promise to predict transitions with a vastly reduced computational cost. Here, we demonstrate how transformers can be trained to predict atomistic transitions in nano-clusters. We show how we evaluate physical validity of the predictions and how a multitude of additional, different microstates can be generated by slightly varying the data provided to the model.", "date": "2026-03-05", "categories": []}
{"id": "2604.27300", "title": "METASYMBO: Multi-Agent Language-Guided Metamaterial Discovery via Symbolic Latent Evolution", "abstract": "Metamaterial discovery seeks microstructured materials whose geometry induces targeted mechanical behavior. Existing inverse-design methods can efficiently generate candidates, but they typically require explicit numerical property targets and are less suitable for early-stage exploration, where researchers often begin with incomplete constraints and qualitative intents expressed in natural language. Large language models can interpret such intents, but they lack geometric awareness and physical property validity. To address this gap, we propose MetaSymbO, a multi-agent framework for language-guided Metamaterial discovery via Symbolic-driven latent evOlution. Specifically, MetaSymbO contains three agents: a Designer that interprets free-form design intents and retrieves a semantically consistent scaffold, a Generator that synthesizes candidate microstructures in a disentangled latent space, and a Supervisor that provides fast property-aware feedback for iterative refinement. To move beyond the limitations of reproducing known samples from literature and training data, we further introduce symbolic-driven latent evolution, which applies programmable operators over disentangled latent factors to compose, modify, and refine structures at inference time. Extensive experiments demonstrate that (i) MetaSymbO improves structural validity by up to 34% in symmetry and nearly 98% in periodicity compared to state-of-the-art baselines; (ii) MetaSymbO achieves about 6-7% higher language-guidance scores while maintaining superior structure novelty compared to advanced reasoning LLMs; (iii) qualitative analyses confirm the effectiveness of symbolic logic operators in enabling programmable semantic alignment; and (iv) realworld case studies on auxetic, high-stiffness metamaterial design further validate its practical capability.", "date": "2026-04-30", "categories": []}
{"id": "2604.27613", "title": "AMGenC: Generating Charge Balanced Amorphous Materials", "abstract": "Amorphous (disordered) materials are solids that have shown great potential in various domains, including energy storage, thermal management, and advanced materials. Unlike crystalline materials that can be described by unit cells containing a few to hundreds of atoms, amorphous materials require larger simulation cells with at least hundreds to thousands of atoms. To advance the design of amorphous materials with desired properties and facilitate the exploration of their vast design space, generative inverse design has emerged as a promising approach. It aims to directly output materials with properties closely aligned with the desired ones using probabilistic generative models conditioned on desired properties, which can be more resource efficient than the traditional trial-and-error approach. However, due to the inherent stochasticity of probabilistic generative models, when element assignments are unconstrained, a large portion of generated materials may be charge unbalanced, and no existing methods can effectively mitigate this limitation. In this work, we propose AMGenC, a new generative inverse design method for amorphous materials that can guarantee the generation of charge balanced samples, with minimal additional computational overhead and without sacrificing inverse design accuracy. AMGenC achieves this through an element noise that gives the generation process a starting point centered around charge balance, and the combination of a per-step soft projection and a final discrete projection for steering the elements toward exact charge balance throughout the generation. We perform extensive experiments on two amorphous materials datasets. Experimental results provide evidence that AMGenC achieves its design goal.", "date": "2026-04-30", "categories": []}
{"id": "2604.27685", "title": "VibroML: an automated toolkit for high-throughput vibrational analysis and dynamic instability remediation of crystalline materials using machine-learned potentials", "abstract": "While machine-learned interatomic potentials (MLIPs) accelerate phonon dispersion calculations, merely identifying dynamical instabilities in computationally predicted materials is insufficient; automated pathways to resolve them are required. We introduce VibroML, an open-source Python toolkit driven by foundational MLIPs that shifts the paradigm from stability verification to automated structural remediation. VibroML employs an energy-guided genetic algorithm that vastly outperforms traditional soft-mode following, efficiently navigating the potential energy surface to uncover diverse, dynamically stable polymorphs. As 0 K harmonic stability does not guarantee macroscopic viability, an automated molecular dynamics workflow evaluates finite-temperature structural retention. VibroML also couples with ProtoCSP, our combinatorial structure prediction engine, to stabilize frustrated crystal topologies via targeted alloying, successfully rescuing functional perovskite networks like Cs$_2$KInI$_6$ and KTaSe$_3$. Demonstrating broader applicability, we mined the Alexandria database -- where ~50% of quaternary and 99.5% of quinary elemental combinations lack any structural entries -- to identify thousands of abandoned, high-symmetry stoichiometries. Deploying ProtoCSP's \"cold start\" retrieval and VibroML's evolutionary search on a sample, we successfully identified dynamically stable low-symmetry candidates. Through integrated structural remediation, thermal validation, and systematic compositional exploration, VibroML enables a comprehensive deep-screening approach, yielding physically sound structural propositions that far surpass standard high-throughput workflows.", "date": "2026-04-30", "categories": []}
{"id": "2304.13927", "title": "NIMS-OS: An automation software to implement a closed loop between artificial intelligence and robotic experiments in materials science", "abstract": "NIMS-OS (NIMS Orchestration System) is a Python library created to realize a closed loop of robotic experiments and artificial intelligence (AI) without human intervention for automated materials exploration. It uses various combinations of modules to operate autonomously. Each module acts as an AI for materials exploration or a controller for a robotic experiments. As AI techniques, Bayesian optimization (PHYSBO), boundless objective-free exploration (BLOX), phase diagram construction (PDC), and random exploration (RE) methods can be used. Moreover, a system called NIMS automated robotic electrochemical experiments (NAREE) is available as a set of robotic experimental equipment. Visualization tools for the results are also included, which allows users to check the optimization results in real time. Newly created modules for AI and robotic experiments can be added easily to extend the functionality of the system. In addition, we developed a GUI application to control NIMS-OS.To demonstrate the operation of NIMS-OS, we consider an automated exploration for new electrolytes. NIMS-OS is available at https://github.com/nimsos-dev/nimsos.", "date": "2023-04-27", "categories": []}
{"id": "2502.14912", "title": "Semantic Embeddings of Chemical Elements for Enhanced Materials Inference and Discovery", "abstract": "We present a framework for generating universal semantic embeddings of chemical elements to advance materials inference and discovery. This framework leverages ElementBERT, a domain-specific BERT-based natural language processing model trained on 1.29 million abstracts of alloy-related scientific papers, to capture latent knowledge and contextual relationships specific to alloys. These semantic embeddings serve as robust elemental descriptors, consistently outperforming traditional empirical descriptors with significant improvements across multiple downstream tasks. These include predicting mechanical and transformation properties, classifying phase structures, and optimizing materials properties via Bayesian optimization. Applications to titanium alloys, high-entropy alloys, and shape memory alloys demonstrate up to 23% gains in prediction accuracy. Our results show that ElementBERT surpasses general-purpose BERT variants by encoding specialized alloy knowledge. By bridging contextual insights from scientific literature with quantitative inference, our framework accelerates the discovery and optimization of advanced materials, with potential applications extending beyond alloys to other material classes.", "date": "2025-02-19", "categories": []}
{"id": "2507.22558", "title": "aLLoyM: A large language model for alloy phase diagram prediction", "abstract": "Large Language Models (LLMs) are general-purpose tools with wide-ranging applications, including in materials science. In this work, we introduce aLLoyM, a fine-tuned LLM specifically trained on alloy compositions, temperatures, and their corresponding phase information. To develop aLLoyM, we curated question-and-answer (Q&A) pairs for binary and ternary phase diagrams using the open-source Computational Phase Diagram Database (CPDDB) and assessments based on CALPHAD (CALculation of PHAse Diagrams). We fine-tuned Mistral, an open-source pre-trained LLM, for two distinct Q&A formats: multiple-choice and short-answer. Benchmark evaluations demonstrate that fine-tuning substantially enhances performance on multiple-choice phase diagram questions. Moreover, the short-answer model of aLLoyM exhibits the ability to generate novel phase diagrams from its components alone, underscoring its potential to accelerate the discovery of previously unexplored materials systems. To promote further research and adoption, we have publicly released the short-answer fine-tuned version of aLLoyM, along with the complete benchmarking Q&A dataset, on Hugging Face.", "date": "2025-07-30", "categories": []}
{"id": "2604.26703", "title": "A self-evolving agent for explainable diagnosis of DFT-experiment band-gap mismatch", "abstract": "Standard density functional theory (DFT) routinely misclassifies the electronic ground state of correlated and structurally complex compounds, predicting metallic behaviour for materials that experiments report as semiconductors. Each such mismatch encodes a specific non-ideality -- magnetic ordering, electron correlation, an alternative polymorph, or a defect -- that the calculation excluded, but extracting that signal at scale has remained a manual exercise. Here we introduce XDFT, a closed-loop agent that diagnoses the mismatch automatically: it draws candidate hypotheses from a curated catalogue, executes the corresponding first-principles tests, and updates a global Bayesian posterior over hypothesis usefulness from each verdict. On a verified benchmark of 124 materials, XDFT identifies a resolving mechanism for 70 of 90 mismatch cases (78\\%), an order of magnitude above a uniform-random baseline (19\\%) and a static LLM ordering (20\\%). The internal posterior aligns with empirical performance over the benchmark timeline, and resolved cases collapse into a tri-partite element-class taxonomy that we distil into a four-line static rule. Each diagnosed material is returned with a corrected protocol and a mechanistic attribution; failed cases are flagged as evidence-backed targets for experimental re-examination.", "date": "2026-04-29", "categories": []}
{"id": "2603.21942", "title": "Suiren-1.0 Technical Report: A Family of Molecular Foundation Models", "abstract": "We introduce Suiren-1.0, a family of molecular foundation models for the accurate modeling of diverse organic systems. Suiren-1.0 comprising three specialized variants (Suiren-Base, Suiren-Dimer, and Suiren-ConfAvg) is integrated within an algorithmic framework that bridges the gap between 3D conformational geometry and 2D statistical ensemble spaces. We first pre-train Suiren-Base (1.8B parameters) on a 70M-sample Density Functional Theory dataset using spatial self-supervision and SE(3)-equivariant architectures, achieving robust performance in quantum property prediction. Suiren-Dimer extends this capability through continued pre-training on 13.5M intermolecular interaction samples. To enable efficient downstream application, we propose Conformation Compression Distillation (CCD), a diffusion-based framework that distills complex 3D structural representations into 2D conformation-averaged representations. This yields the lightweight Suiren-ConfAvg, which generates high-fidelity representations from SMILES or molecular graphs. Our extensive evaluations demonstrate that Suiren-1.0 establishes state-of-the-art results across a range of tasks. All models and benchmarks are open-sourced.", "date": "2026-03-23", "categories": []}
{"id": "2604.25568", "title": "Benchmarking bandgap prediction in semiconductors under experimental and realistic evaluation settings", "abstract": "Accurate bandgap prediction is crucial for semiconductor applications, yet machine learning models trained on computational data often struggle to generalize to experimental bandgap measurements. Challenges related to data fidelity, domain generalization, and model interpretability remain insufficiently addressed in existing evaluation frameworks. To bridge this gap, we introduce RealMat-BaG, a benchmark for assessing model reliability under experimentally relevant conditions. We curate an open-access dataset of experimental bandgaps with aligned crystal structures and compare graph neural networks as well as classical machine learning baselines. Our framework evaluates performance across statistical and domain-based splits, examines transfer from DFT-computed to experimental bandgaps, and analyzes interpretability at both elemental-property and structural levels. Our results reveal the fundamental generalization limitations of current bandgap prediction models and establish a benchmark aligned with experimental measurements for developing more reliable learning strategies for materials discovery.", "date": "2026-04-28", "categories": []}
{"id": "2511.21213", "title": "Lattice-to-Total Thermal Conductivity Ratio: A Phonon-Glass Electron-Crystal Descriptor for Data-Driven Thermoelectric Design", "abstract": "Thermoelectrics (TEs) are promising candidates for energy harvesting with performance quantified by figure of merit, $ZT$. To accelerate the discovery of high-$ZT$ materials, efforts have focused on identifying compounds with low thermal conductivity $\\kappa$. Using a curated dataset of 71,913 entries, we show that high-$ZT$ materials reside not only in the low-$\\kappa$ regime but also cluster near a lattice-to-total thermal conductivity ratio ($\\kappa_\\mathrm{L}/\\kappa$) of approximately 0.5. This optimal ratio provides a quantitative descriptor for the well-known phonon-glass electron-crystal (PGEC) design concept. Building on this insight, we construct a framework consisting of two machine learning models for the lattice and electronic components of thermal conductivity that jointly provide both $\\kappa$ and $\\kappa_\\mathrm{L}/\\kappa$ for screening and guiding the optimization of TE materials. By applying this framework to 104,567 inorganic compounds, we identify 2,522 ultralow-$\\kappa$ candidates while simultaneously evaluating their proximity to the optimal PGEC regime. A follow-up case study on chemical doping demonstrates how the framework can qualitatively provide optimization strategies that shift pristine materials toward the ideal $\\kappa_\\mathrm{L}/\\kappa$ $\\approx$ 0.5 target. Ultimately, by integrating rapid screening with PGEC-guided optimization, our data-driven framework takes a critical step towards closing the gap between materials discovery and performance enhancement.", "date": "2025-11-26", "categories": []}
{"id": "2603.12365", "title": "Optimal Experimental Design for Reliable Learning of History-Dependent Constitutive Laws", "abstract": "History-dependent constitutive models serve as macroscopic closures for the aggregated effects of micromechanics. Their parameters are typically learned from experimental data. With a limited experimental budget, eliciting the full range of responses needed to characterize the constitutive relation can be difficult. As a result, the data can be well explained by a range of parameter choices, leading to parameter estimates that are uncertain or unreliable. To address this issue, we propose a Bayesian optimal experimental design framework to quantify, interpret, and maximize the utility of experimental designs for reliable learning of history-dependent constitutive models. In this framework, the design utility is defined as the expected reduction in parametric uncertainty or the expected information gain. This enables in silico design optimization using simulated data and reduces the cost of physical experiments for reliable parameter identification.\n  We introduce two approximations that make this framework practical for advanced material testing with expensive forward models and high-dimensional data: (i) a Gaussian approximation of the expected information gain, and (ii) a surrogate approximation of the Fisher information matrix. The former enables efficient design optimization and interpretation, while the latter extends this approach to batched design optimization by amortizing the cost of repeated utility evaluations. Our numerical studies of uniaxial tests for viscoelastic solids show that optimized specimen geometries and loading paths yield image and force data that significantly improve parameter identifiability relative to random designs, especially for parameters associated with memory effects.", "date": "2026-03-12", "categories": []}
{"id": "2604.23921", "title": "Crystal structure prediction using graph neural combinatorial optimization", "abstract": "Crystalline materials are widely used in technological applications, yet their discovery remains a significant challenge. As their properties are driven by structure, crystal structure prediction (CSP) methods play a central role in computational approaches aiming to accelerate this process. Previously, CSP has been approached from a combinatorial optimization perspective, with the core challenge of allocating atoms on a fine grid of predefined discrete positions within a unit cell while minimizing their interaction energy. Exact mathematical optimization methods provide guaranteed solutions, but they become computationally expensive for large-scale instances, where the atomic configuration space grows rapidly, particularly in the absence of additional symmetry constraints. In this work, we introduce a neural combinatorial optimization approach to the atom allocation challenge and, subsequently, CSP, based on graph neural networks (GNNs), which can effectively sample from the distribution of feasible structures in an unsupervised manner. We leverage expander graphs to construct computational graphs over discrete positions that capture both short- and long-range interactions between atoms, and employ the Gumbel-Sinkhorn approach to enforce the desired stoichiometry of the generated structures. We demonstrate that our method outperforms classical heuristic approaches and is competitive with a commercial optimization solver across a range of chemical compositions. This enables the use of ever-expanding GPU infrastructure to tackle the inherent combinatorial challenges of CSP, paving the way for scaling beyond current capabilities.", "date": "2026-04-27", "categories": []}
{"id": "2410.13713", "title": "CrystalX: High-accuracy Crystal Structure Analysis Using Deep Learning", "abstract": "Atomic structure analysis of crystalline materials is a paramount endeavor in both chemical and material sciences. This sophisticated technique necessitates not only a solid foundation in crystallography but also a profound comprehension of the intricacies of the accompanying software, posing a significant challenge in meeting the rigorous daily demands. For the first time, we confront this challenge head-on by harnessing the power of deep learning for fully automated routine structure analysis at the full-atom level. To validate the performance of the model, named CrystalX, we employed a dataset comprising over 50,000 X-ray diffraction measurements derived from authentic experiments. Under a strict temporal validation scheme that separates training and test data by publication time, CrystalX substantially outperformed the automated baseline and adept at deciphering intricate geometric patterns. Remarkably, CrystalX revealed that even peer-reviewed publications harbor expert interpretation errors that can evade stringent CheckCIF A/B-level alerts, yet CrystalX adeptly rectifies them. It has already been successfully applied in our day-to-day pipeline, enabling fully automated, human-free structure analysis for newly discovered compounds. Overall, CrystalX marks the beginning of a new era in automating routine structural analysis within self-driving laboratories.", "date": "2024-10-17", "categories": []}
{"id": "2503.04492", "title": "Accurate predictive model of band gap with selected important features based on explainable machine learning", "abstract": "In the rapidly advancing field of materials informatics, nonlinear machine learning models have demonstrated exceptional predictive capabilities for material properties. However, their black-box nature limits interpretability, and they may incorporate features that do not contribute to -- or even deteriorate -- model performance. This study employs explainable ML (XML) techniques, including permutation feature importance and the SHapley Additive exPlanation, applied to a pristine support vector regression model designed to predict band gaps at the GW level using 18 input features. Guided by XML-derived individual feature importance, a simple framework is proposed to construct reduced-feature predictive models. Model evaluations indicate that an XML-guided compact model, consisting of the top five features, achieves comparable accuracy to the pristine model on in-domain datasets (0.254 vs. 0.247 eV) while showing improved generalization with lower prediction errors on out-of-domain data (0.348 vs. 0.460 eV). Additionally, the study underscores the necessity for eliminating strongly correlated features (correlation coefficient greater than 0.8) to prevent misinterpretation and overestimation of feature importance before applying XML. This study highlights XML's effectiveness in developing simplified yet highly accurate machine learning models by clarifying feature roles, thereby reducing computational costs for feature acquisition and enhancing model trustworthiness for materials discovery.", "date": "2025-03-06", "categories": []}
{"id": "2604.20899", "title": "Predicting Scale-Up of Metal-Organic Framework Syntheses with Large Language Models", "abstract": "Scalable synthesis remains the gate between MOF discovery and industrial deployment, as scale-up know-how is fragmented across disparate reports. We introduce ESU-MOF, a literature-mined dataset and a positive-unlabeled learning strategy that fine-tunes large language models to predict scalability potential with 91.4% accuracy, enabling rapid data-driven triage for industrial MOF discovery.", "date": "2026-04-21", "categories": []}
{"id": "2604.21068", "title": "Expanding the extreme-k dielectric materials space through physics-validated generative reasoning", "abstract": "The most technologically consequential materials are often the rarest: they occupy narrow regions of chemical space, obey competing physical constraints, and appear only sparsely in existing databases. High-kappa dielectrics, high-Tc superconductors, and ferromagnetic insulators are to name a few. This scarcity fundamentally limits today's data-driven materials discovery, where machine-learning models excel at interpolation but struggle to generate genuinely new candidates. Here, we introduce DielecMIND, an artificial intelligence framework that reframes materials discovery as a reasoning-driven exploration instead of a database-screening problem. Using high-kappa dielectrics as a data-scarce and technologically stringent test case, DielecMIND combines large-language-model hypothesis generation for the first time with physics validated first-principles calculation to navigate chemical space beyond known compounds. Prior to our work, only 14 experimentally or computationally validated materials with kappa > 150 were known. Our framework discovers and validates 5 new such compounds, expanding this rare-materials class by a remarkable = 35% in a single study. Among them, we find that Ba2TiHfO6 exhibits a dielectric constant of 637, minimal loss at low optical frequencies, and stability up to 800 K. Beyond dielectrics, this work demonstrates a new paradigm for artificial-intelligence-guided discovery: one that generates a small number of physically grounded, experimentally plausible candidates yet measurably expands sparsely populated functional materials spaces. Thus, DielecMIND points toward a general strategy for discovering rare, high-impact functional materials where data scarcity has long constrained progress.", "date": "2026-04-22", "categories": []}
{"id": "2604.21073", "title": "Generative Discovery of Magnetic Insulators under Competing Physical Constraints", "abstract": "Discovering materials that must simultaneously satisfy multiple competing constraints remains a central challenge in computational materials design, particularly in data-scarce regimes where conventional data-driven approaches are least effective. Magnetic insulators represent a stringent example: the electronic conditions that favor magnetic order often also promote metallicity, while insulating behavior suppresses the interactions that stabilize magnetism. As a result, experimentally viable magnetic insulators are rare and difficult to identify through conventional screening. Here, we introduce MagMatLLM, a constraint-guided generative discovery framework that integrates language-model-based crystal generation with evolutionary selection, surrogate screening, and first-principles validation to target simultaneous stability, magnetism, and insulating behavior. Unlike stability-first approaches, the framework enforces functional constraints during generation and selection, steering the search toward sparsely populated regions of materials space defined by competing physical requirements. Using this workflow, we identify twelve previously unreported candidate magnetic insulators, including Tm$_4$Co$_2$Cr$_2$O$_{12}$ and Cr$_4$Nb$_2$O$_{12}$. Of these, ten are dynamically stable by phonon analysis and exhibit finite band gaps and nonzero magnetic moments in spin-polarized density functional theory calculations. Beyond the specific compounds identified here, this work establishes a general constraint-guided paradigm for multi-objective materials discovery in sparse chemical spaces and provides a transferable strategy for the design of quantum materials under competing physical constraints.", "date": "2026-04-22", "categories": []}
{"id": "2604.21222", "title": "Neutron and X-ray Diffraction Reveal the Limits of Long-Range Machine Learning Potentials for Medium-Range Order in Silica Glass", "abstract": "Glassy silica is a foundational material in optics and electronics, yet accurately predicting its medium-range order (MRO) remains a major challenge for machine-learning interatomic potentials (MLIPs). While local MLIPs reproduce the short-range SiO4 tetrahedral network well, it remains unclear whether locality alone is sufficient to recover the first sharp diffraction peak (FSDP), the principal experimental signature of MRO. Here, we combine neutron and X-ray diffraction measurements with large-scale molecular dynamics driven by two MACE-based models: a short-range (SR) potential and a long-range (LR) extension incorporating reciprocal-space gated attention. The SR model systematically over-structures the network, producing an overly intense FSDP in both the liquid and glassy states. Incorporating long-range interactions improves agreement with experiment for the liquid structure by reducing this excess ordering, but the LR model still fails to recover the experimental amorphous MRO after quenching. Ring-statistics and bond-angle analyses reveal that SR model exhibits an artificially narrow distribution dominated by six-membered rings, while the LR model produces a broader but still biased ring population. Despite preserving the correct tetrahedral geometry, both models show limited variability in Si-O-Si angles, indicating constrained network flexibility. These structural signatures demonstrate that both models retain excessive memory of the parent liquid network, leading to kinetically trapped and nonphysical medium-range configurations during vitrification. These results show that explicit long-range interactions are necessary but not sufficient for predictive modelling of disordered silica and suggest that accurate MRO further requires training data and sampling strategies that adequately represent the liquid-to-glass transition.", "date": "2026-04-23", "categories": []}
{"id": "2604.20304", "title": "LLM-guided phase diagram construction through high-throughput experimentation", "abstract": "Constructing phase diagrams for multicomponent alloys requires extensive experimental measurements and is a time-consuming task. Here we investigate whether large language models (LLMs) can guide experimental planning for phase diagram construction. In our framework, a general-purpose LLM serves as the experimental planner, suggesting compositions for measurement at each cycle in a closed loop with high-throughput synthesis and X-ray diffraction phase identification. Using this framework, we experimentally constructed the ternary phase diagram of the Co-Al-Ge system at 900 degree C through iterative synthesis and characterization. We compared two strategies that differ in how the initial compositions are selected: one uses predictions from a domain-specific LLM trained on phase diagram data (aLLoyM), while the other relies solely on the general-purpose LLM. The two strategies exhibited complementary strengths. aLLoyM directed the initial measurements toward compositionally complex regions in the interior of the ternary diagram, enabling the earliest discovery of all three novel phases that form only in the ternary system. In contrast, the general-purpose LLM adopted a textbook-like approach which efficiently identified a larger number of phases in fewer cycles. In addition, a simulated benchmark comparing the LLM against conventional machine learning confirmed that the LLM achieves more efficient exploration. The results demonstrate that LLMs have high potential as experimental planners for phase diagram construction.", "date": "2026-04-22", "categories": []}
{"id": "2604.20458", "title": "Surrogate Functionals for Machine-Learned Orbital-Free Density Functional Theory", "abstract": "We introduce surrogate functionals: machine-learned energy functionals for orbital-free density functional theory (OF-DFT) which are defined not by universal fidelity to a physical reference, but merely by the requirement that density optimization with a fixed procedure yields the true ground-state density. Helpfully, training surrogate functionals requires only ground-state densities, no energies or gradients away from the ground state. We here propose a gradient-descent-improvement loss that guarantees exponential convergence of the density to the ground state, and combine it with an adaptive sampling scheme that concentrates learning around the optimization trajectories actually visited during inference. On the QM9 and QMugs benchmarks, surrogate functionals achieve density errors competitive with or improving upon the state of the art for fully supervised machine-learned OF-DFT, while eliminating the need for the $O(N^3)$ orthononormalization step required by prior work, yielding improved runtime scaling for larger systems.", "date": "2026-04-22", "categories": []}
{"id": "2409.06080", "title": "Regression with Large Language Models for Materials and Molecular Property Prediction", "abstract": "We demonstrate the ability of large language models (LLMs) to perform material and molecular property regression tasks, a significant deviation from the conventional LLM use case. We benchmark the Large Language Model Meta AI (LLaMA) 3 on several molecular properties in the QM9 dataset and 28 materials properties. Only composition-based input strings are used as the model input and we fine tune on only the generative loss. We broadly find that LLaMA 3, when fine-tuned using the SMILES representation of molecules, provides useful regression results which can rival standard materials property prediction models like random forest or fully connected neural networks on the QM9 dataset. Not surprisingly, LLaMA 3 errors are 5-10x higher than those of the state-of-the-art models that were trained using far more granular representation of molecules (e.g., atom types and their coordinates) for the same task. Similarly, LLaMA 3 provides comparable, although slightly worse, accuracy relative to random forest and elemental descriptors when using just compound chemical description on our set of 28 materials properties. Interestingly, LLaMA 3 provides improved predictions compared to GPT-3.5 and GPT-4o. This work highlights the versatility of LLMs, suggesting that LLM-like generative models can potentially transcend their traditional applications to tackle complex physical phenomena, thus paving the way for future research and applications in chemistry, materials science and other scientific domains.", "date": "2024-09-09", "categories": []}
{"id": "2604.19383", "title": "Multimodal Transformer for Sample-Aware Prediction of Metal-Organic Framework Properties", "abstract": "Metal-organic frameworks (MOFs) are a major target of machine-learning-based property prediction, yet most models assume that a single framework representation maps to a single property value. This assumption becomes problematic for experimental MOFs, where samples reported as the same framework can exhibit different properties because of differences in crystallinity, phase purity, defects, and other sample-dependent factors. Here we introduce Experimental X-ray Diffraction Integrated Transformer (EXIT), a multimodal transformer for sample-aware prediction of MOF properties that combines MOFid with X-ray diffraction (XRD). In EXIT, MOFid encodes MOF identity, whereas XRD provides complementary information about the experimentally realized sample state. EXIT is pre-trained on one million hypothetical MOFs with simulated XRD to learn transferable representations, leading to improved downstream performance relative to existing approaches. EXIT is fine-tuned on literature-derived experimental datasets for surface area and pore volume prediction. Incorporating experimental XRD improves predictive performance relative to models without experimental XRD, and attention analysis and sample-level case studies further show that EXIT assigns different predictions to samples sharing the same MOF identity when their XRD patterns differ. These results establish a practical step from framework-aware to sample-aware MOF property prediction and highlight the value of incorporating experimental characterization into porous materials informatics.", "date": "2026-04-21", "categories": []}
{"id": "2407.09514", "title": "Machine Learning Based Prediction of Proton Conductivity in Metal-Organic Frameworks", "abstract": "Recently, metal-organic frameworks (MOFs) have demonstrated their potential as solid-state electrolytes in proton exchange membrane fuel cells. However, the number of MOFs reported to exhibit proton conductivity remains limited, and the mechanisms underlying this phenomenon are not fully elucidated, complicating the design of proton-conductive MOFs. In response, we developed a comprehensive database of proton-conductive MOFs and applied machine learning techniques to predict their proton conductivity. Our approach included the construction of both descriptor-based and transformer-based models. Notably, the transformer-based transfer learning (Freeze) model performed the best with a mean absolute error (MAE) of 0.91, suggesting that the proton conductivity of MOFs can be estimated within one order of magnitude using this model. Additionally, we employed feature importance and principal component analysis to explore the factors influencing proton conductivity. The insights gained from our database and machine learning model are expected to facilitate the targeted design of proton-conductive MOFs.", "date": "2024-06-18", "categories": []}
{"id": "2511.03122", "title": "EGMOF: Efficient Generation of Metal-Organic Frameworks Using a Hybrid Diffusion-Transformer Architecture", "abstract": "Designing materials with targeted properties remains challenging due to the vastness of chemical space and the scarcity of property-labeled data. While recent advances in generative models offer a promising way for inverse design, most approaches require large datasets and must be retrained for every new target property. Here, we introduce the EGMOF (Efficient Generation of MOFs), a hybrid diffusion-transformer framework that overcomes these limitations through a modular, descriptor-mediated workflow. EGMOF decomposes inverse design into two steps: (1) a one-dimensional diffusion model (Prop2Desc) that maps desired properties to chemically meaningful descriptors followed by (2) a transformer model (Desc2MOF) that generates structures from these descriptors. This modular hybrid design enables minimal retraining and maintains high accuracy even under small-data conditions. On a hydrogen uptake dataset, EGMOF achieved over 94% validity and 91% hit rate, representing significant improvements of up to 39% in validity and 29% in hit rate compared to existing methods, while remaining effective with only 1,000 training samples. Moreover, our model successfully performed conditional generation across 29 diverse property datasets, including CoREMOF, QMOF, and text-mined experimental datasets, whereas previous models have not. This work presents a data-efficient, generalizable approach to the inverse design of diverse MOFs and highlights the potential of modular inverse design workflows for broader materials discovery.", "date": "2025-11-05", "categories": []}
{"id": "2512.06987", "title": "OXtal: An All-Atom Diffusion Model for Organic Crystal Structure Prediction", "abstract": "Accurately predicting experimentally realizable 3D molecular crystal structures from their 2D chemical graphs is a long-standing open challenge in computational chemistry called crystal structure prediction (CSP). Efficiently solving this problem has implications ranging from pharmaceuticals to organic semiconductors, as crystal packing directly governs the physical and chemical properties of organic solids. In this paper, we introduce OXtal, a large-scale 100M parameter all-atom diffusion model that directly learns the conditional joint distribution over intramolecular conformations and periodic packing. To efficiently scale OXtal, we abandon explicit equivariant architectures imposing inductive bias arising from crystal symmetries in favor of data augmentation strategies. We further propose a novel crystallization-inspired lattice-free training scheme, Stoichiometric Stochastic Shell Sampling ($S^4$), that efficiently captures long-range interactions while sidestepping explicit lattice parametrization -- thus enabling more scalable architectural choices at all-atom resolution. By leveraging a large dataset of 600K experimentally validated crystal structures (including rigid and flexible molecules, co-crystals, and solvates), OXtal achieves orders-of-magnitude improvements over prior ab initio machine learning CSP methods, while remaining orders of magnitude cheaper than traditional quantum-chemical approaches. Specifically, OXtal recovers experimental structures with conformer $\\text{RMSD}_1<0.5$ {\\AA} and attains over 80\\% packing similarity rate, demonstrating its ability to model both thermodynamic and kinetic regularities of molecular crystallization.", "date": "2025-12-07", "categories": []}
{"id": "2603.06082", "title": "Offline Materials Optimization with CliqueFlowmer", "abstract": "Recent advances in deep learning inspired neural network-based approaches to computational materials discovery (CMD). A plethora of problems in this field involve finding materials that optimize a target property. Nevertheless, the increasingly popular generative modeling methods are ineffective at boldly exploring attractive regions of the materials space due to their maximum likelihood training. In this work, we offer an alternative CMD technique based on offline model-based optimization (MBO) that fuses direct optimization of a target material property into generation. To that end, we introduce a domain-specific model, dubbed CliqueFlowmer, that incorporates recent advances of clique-based MBO into transformer and flow generation. We validate this model's optimization abilities and show that materials it produces strongly outperform those from generative baselines. To support specialized materials discovery applications and broader interdisciplinary research, we release our code, model weights, and additional project resources at https://github.com/znowu/CliqueFlowmer, https://colab.research.google.com/drive/1usUg7zezFkcYHlm2MdYwZUNJXf_YkWnY?usp=sharing, and https://x.com/kuba_AI/status/2033382617442345321.", "date": "2026-03-06", "categories": []}
{"id": "2604.16468", "title": "Multi-Label Phase Diagram Prediction in Complex Alloys via Physics-Informed Graph Attention Networks", "abstract": "Accurate phase equilibria are foundational to alloy design because they encode the underlying thermodynamics governing stability, transformations, and processing windows. However, while the CALculation of Phase Diagrams (CALPHAD) provides a rigorous thermodynamic framework, exploring multicomponent composition-temperature space remains computationally expensive and is typically limited to sparse section. To enable rapid phase mapping and alloy screening, we propose a physics-informed graph attention network (GAT) that learns element-aware representations and couples them with thermodynamic constraints for multi-label phase-set prediction in the Ag-Bi-Cu-Sn alloy system. Using about 25,000 equilibrium states generated with pycalphad, each composition-temperature point is represented as a four-node element graph with atomic fractions and elemental descriptors as node features. The model combines graph attention, global pooling, and a multilayer perceptron to predict nine relevant phases. To improve physical consistency, we incorporate thermodynamic constraints, applied as training penalties or as an inference-time projection. Across six binary and three ternary subsystems, the baseline model achieves a macro-F1 score of 0.951 and 93.98% exact-set match, while physics-informed decoding improves robustness and raises exact-set accuracy to about 96% on dense in-domain grids. The surrogate also generalizes to an unseen ternary section with 99.32% exact-set accuracy and to a quaternary section at 700 {\\deg}C with 91.78% accuracy. These results demonstrate that attention-based graph learning coupled with thermodynamic constraint enforcement provides an effective and physically consistent surrogate for high-resolution phase mapping and extrapolative alloy screening.", "date": "2026-04-09", "categories": []}
{"id": "2604.15380", "title": "Exascale Multi-Task Graph Foundation Models for Imbalanced, Multi-Fidelity Atomistic Data", "abstract": "We present an exascale workflow for materials discovery using atomistic graph foundation models built on HydraGNN. We jointly train on 16 open first-principles datasets (544+ million structures covering 85+ elements) using a multi-task architecture with per-dataset heads and a scalable ADIOS2/DDStore data pipeline. On Frontier, we execute six large-scale DeepHyper hyperparameter optimization campaigns in FP64 and promote the top-performing message-passing models to sustained 2,048-node training, yielding a PaiNN-based lead model. The resulting model enables billion-scale screening, evaluating 1.1 billion atomistic structures in 50 seconds, compressing a workload that would require years of first-principles computation, and supports data-scarce fine-tuning across diverse downstream tasks. We quantify precision-performance tradeoffs (BF16/FP32/FP64), demonstrate transfer across twelve chemically diverse downstream tasks, and establish seamless strong- and weak-scaling across Frontier, Aurora, and Perlmutter. This work allows fast and reliable exploration of vast chemical design spaces that are otherwise inaccessible to first-principles methods.", "date": "2026-04-15", "categories": []}
{"id": "2604.15821", "title": "Breaking the Training Barrier of Billion-Parameter Universal Machine Learning Interatomic Potentials", "abstract": "Universal Machine Learning Interatomic Potentials (uMLIPs), pre-trained on massively diverse datasets encompassing inorganic materials and organic molecules across the entire periodic table, serve as foundational models for quantum-accurate physical simulations. However, uMLIP training requires second-order derivatives, which lack corresponding parallel training frameworks; moreover, scaling to the billion-parameter regime causes explosive growth in computation and communication overhead, making its training a tremendous challenge. We introduce MatRIS-MoE, a billion-parameter Mixture-of-Experts model built upon invariant architecture, and {Janus}, a pioneering high-dimensional distributed training framework for uMLIPs with hardware-aware optimizations. Deployed across two Exascale supercomputers, our code attains a peak performance of 1.2/1.0 EFLOPS (24\\%/{35.5\\%} of theoretical peak) in single precision at over 90\\% parallel efficiency, compressing the training of billion-parameter uMLIPs from weeks to hours. This work establishes a new high-water mark for AI-for-Science (AI4S) foundation models at Exascale and provides essential infrastructure for rapid scientific discovery.", "date": "2026-04-17", "categories": []}
{"id": "2604.14562", "title": "Material-Agnostic Zero-Shot Thermal Inference for Metal Additive Manufacturing via a Parametric PINN Framework", "abstract": "Accurate thermal modeling in metal additive manufacturing (AM) is essential for understanding the process-structure-performance relationship. While prior studies have explored generalization across unseen process conditions, they often require extensive datasets, costly retraining, or pre-training. Generalization across different materials also remains relatively unexplored due to the challenges posed by distinct material-dependent thermal behaviors. This paper introduces a parametric physics-informed neural network (PINN) framework for zero-shot generalization across arbitrary materials without labeled data, retraining, or pre-training. The framework adopts a decoupled parametric PINN architecture that separately encodes material properties and spatiotemporal coordinates, fusing them through conditional modulation to better align with the multiplicative role of material parameters in the governing equation and boundary conditions. Physics-guided output scaling derived from Rosenthal's analytical solution and a hybrid optimization strategy are further incorporated to enhance physical consistency, training stability, and convergence. Experiments on bare plate laser powder bed fusion (LPBF) across diverse metal alloys, including both in-distribution and out-of-distribution cases, demonstrate effective zero-shot generalizability along with superior training efficiency. Specifically, the proposed framework achieved up to a 64.2% reduction in relative L2 error compared to the non-parametric baseline while surpassing its performance within only 4.4% of the baseline training epochs. Ablation studies confirm that the proposed framework's components are broadly applicable to other PINN-based approaches. Overall, the proposed framework provides an efficient and scalable material-agnostic solution for zero-shot thermal modeling, contributing to more flexible and practical deployment in metal AM.", "date": "2026-04-16", "categories": []}
{"id": "2507.16005", "title": "Autonomous Multi-objective Alloy Design through Simulation-guided Optimization", "abstract": "Alloy discovery is constrained by vast compositional spaces, competing objectives, and prohibitive experimental costs. Although simulations and machine learning have each accelerated parts of this process, unifying scientific knowledge, scalable search, and experimental confirmation into a data-efficient workflow remains challenging. Here, we present AutoMAT, a hierarchical autonomous framework spanning ideation to experimental validation. Integrating large language models, automated CALPHAD simulations, residual-learning-based correction, and AI-guided optimization, AutoMAT translates design targets into candidate alloys, refines compositions through closed-loop computational search, and validates results experimentally without hand-curated datasets. Targeting lightweight, high-strength alloys, AutoMAT identifies a titanium alloy 8.1% less dense and 13.0% stronger than the aerospace benchmark Ti-185, achieving the highest specific strength among benchmarked systems. In a second case, AutoMAT discovers a high-entropy alloy with 28.2% higher yield strength than the baseline while preserving high ductility. AutoMAT compresses alloy discovery from years to weeks, establishing a generalizable route toward autonomous materials design.", "date": "2025-07-21", "categories": []}
{"id": "2604.13520", "title": "LEGO-MOF: Equivariant Latent Manipulation for Editable, Generative, and Optimizable MOF Design", "abstract": "Metal-organic frameworks (MOFs) are highly promising for carbon capture, yet navigating their vast design space remains challenging. Recent deep generative models enable de novo MOF design but primarily act as feed-forward structure generators. By heavily relying on predefined building block libraries and non-differentiable post-optimization, they fundamentally sever the information flow required for continuous structural editing. Here, we propose a target-driven generative framework focused on continuous structural manipulation. At its core is LinkerVAE, which maps discrete 3D chemical graphs into a continuous, SE(3)-equivariant latent space. This smooth manifold unlocks geometry-aware manipulations, including implicit chemical style transfer and zero-shot isoreticular expansion. Building upon this, we introduce a test-time optimization (TTO) strategy, utilizing an accurate surrogate model to continuously optimize the latent graphs of existing MOFs toward desired properties. This approach systematically enhances carbon capture performance, achieving a striking average relative boost of 147.5% in pure CO2 uptake while strictly preserving structural validity. Integrated with a latent diffusion model and rigid-body assembly for full MOF construction, our framework establishes a scalable, fully differentiable pathway for both the automated discovery, targeted optimization and editing of functional materials.", "date": "2026-04-15", "categories": []}
{"id": "2604.13897", "title": "MolCryst-MLIPs: A Machine-Learned Interatomic Potentials Database for Molecular Crystals", "abstract": "We present an open Molecular Crystal (MC) database of Machine-Learned Interatomic Potentials (MLIP) called MolCryst-MLIPs. The first release comprises fine-tuned MACE models for nine molecular crystal systems -- Benzamide, Benzoic acid, Coumarin, Durene, Isonicotinamide, Niacinamide, Nicotinamide, Pyrazinamide, and Resorcinol -- developed using the Automated Machine Learning Pipeline (AMLP), which streamlines the entire MLIP development workflow, from reference data generation to model training and validation, into a reproducible and user-friendly pipeline. Models are fine-tuned from the MACE-MH-1 foundation model (omol head), yielding a mean energy MAE of 0.141 kJ/mol/atom and a mean force MAE of 0.648 kJ/mol/Angstrom across all systems. Dynamical stability and structural integrity, as assessed through energy conservation, P2 orientational order parameters, and radial distribution functions, are evaluated using molecular dynamics simulations. The released models and datasets constitute a growing open database of validated MLIPs, ready for production MD simulations of molecular crystal polymorphism under different thermodynamic conditions.", "date": "2026-04-15", "categories": []}
{"id": "2503.10471", "title": "Siamese Foundation Models for Crystal Structure Prediction", "abstract": "Predicting crystal structures from chemical compositions is a fundamental challenge in materials discovery, complicated by complex 3D geometries that distinguish it from fields like protein folding. Here, we present Diffusion-based Crystal Omni (DAO), a pretrain-finetune framework for crystal structure prediction integrating two Siamese foundation models: a structure generator and an energy predictor. The generator is pretrained via a two-stage pipeline on a vast dataset of stable and unstable structures, leveraging the predictor to relax unstable configurations and guide the generative sampling. Across two well-known benchmarks, pretraining significantly enhances performance across multiple backbone architectures. Ablation studies confirm that the synergy between the generator and predictor mutually benefits both components. We further validate DAO on three real-world superconductors ($\\text{Cr}_6\\text{Os}_2$, $\\text{Zr}_{16}\\text{Rh}_8\\text{O}_4$, and $\\text{Zr}_{16}\\text{Pd}_8\\text{O}_4$) typically inaccessible to conventional computation. For $\\text{Cr}_6\\text{Os}_2$, DAO achieves a 100\\% match rate with experimental references and an atomic-position error of 0.0012 under 20-shot generation, performing over 2000$\\times$ faster per iteration than DFT-based structure predictors. These compelling results collectively highlight the potential of our approach for advancing materials science research.", "date": "2025-03-13", "categories": []}
{"id": "2604.11827", "title": "Inverse Design of Inorganic Compounds with Generative AI", "abstract": "Machine learning is revolutionizing chemistry. Beyond the value of predictive models accelerating virtual screening, generative AI aims at enabling inverse design, reversing the compound-to-property prediction paradigm into property-to-compound generation. Chemists now have access to a rich AI toolbox for organic chemistry, including drug discovery. However, the application of these methods to inorganic compounds remains limited by the challenges posed by their intrinsic nature. This Review analyzes how these challenges have been addressed, considering widely diverse systems ranging from molecules to crystals, including transition metal complexes and microporous materials. The analysis focuses on how generative AI methods have evolved towards data-representation-model pipelines that address the full complexity of inorganic compounds, including their chemical composition, geometry, symmetry, and electronic structure. Future directions, like benchmark standardization and the development of synthesizability metrics, are also discussed.", "date": "2026-04-11", "categories": []}
{"id": "2604.11957", "title": "Agentic LLM Reasoning in a Self-Driving Laboratory for Air-Sensitive Lithium Halide Spinel Conductors", "abstract": "Self-driving laboratories promise to accelerate materials discovery. Yet current automated solid-state synthesis platforms are limited to ambient conditions, thereby precluding their use for air-sensitive materials. Here, we present A-Lab for Glovebox Powder Solid-state Synthesis (A-Lab GPSS), a robotic platform capable of synthesizing and characterizing air-sensitive inorganic materials under strict air-free conditions. By integrating an agentic AI framework into the A-Lab GPSS platform, we structure autonomous experimental design through abductive and inductive reasoning. We deploy this platform to explore the vast compositional space of lithium halide spinel solid-state ionic conductors. Across a synthesis campaign comprising 352 samples with diverse compositions, the system explores a broad chemical space, experimentally realizing 72% of the 171 possible pairwise combinations among the 19 metals considered in this study. Over the course of the campaign, the fraction of compositions exhibiting both good ionic conductivity (> 0.05 mS/cm) and high halide spinel phase purity increases from 1.33% in the first 75 agent-proposed samples to 5.33% in the final 75. Furthermore, by inspecting the AI's reasoning processes, we reveal distinct yet complementary discovery strategies: abductive reasoning interrogates abnormal observations within already explored regions, whereas inductive reasoning expands the search into broader, previously unvisited chemical space. This work establishes a scalable platform for the autonomous discovery of complex, air-sensitive solid-state materials.", "date": "2026-04-13", "categories": []}
{"id": "2502.02189", "title": "deCIFer: Crystal Structure Prediction from Powder Diffraction Data using Autoregressive Language Models", "abstract": "Novel materials drive advancements in fields ranging from energy storage to electronics, with crystal structure characterization forming a crucial yet challenging step in materials discovery. In this work, we introduce \\emph{deCIFer}, an autoregressive language model designed for powder X-ray diffraction (PXRD)-conditioned crystal structure prediction (PXRD-CSP). Unlike traditional CSP methods that rely primarily on composition or symmetry constraints, deCIFer explicitly incorporates PXRD data, directly generating crystal structures in the widely adopted Crystallographic Information File (CIF) format. The model is trained on nearly 2.3 million crystal structures, with PXRD conditioning augmented by basic forms of synthetic experimental artifacts, specifically Gaussian noise and instrumental peak broadening, to reflect fundamental real-world conditions. Validated across diverse synthetic datasets representative of challenging inorganic materials, deCIFer achieves a 94\\% structural match rate. The evaluation is based on metrics such as the residual weighted profile ($R_{wp}$) and structural match rate (MR), chosen explicitly for their practical relevance in this inherently underdetermined problem. deCIFer establishes a robust baseline for future expansion toward more complex experimental scenarios, bridging the gap between computational predictions and experimental crystal structure determination.", "date": "2025-02-04", "categories": []}
{"id": "2511.08350", "title": "Revealing the Hidden Third Dimension of Point Defects in Two-Dimensional MXenes", "abstract": "Point defects govern many important functional properties of two-dimensional (2D) materials. However, resolving the three-dimensional (3D) arrangement of these defects in multi-layer 2D materials remains a fundamental challenge, hindering rational defect engineering. Here, we overcome this limitation using an artificial intelligence-guided electron microscopy workflow to map the 3D topology and clustering of atomic vacancies in Ti$_3$C$_2$T$_X$ MXene. Our approach reconstructs the 3D coordinates of vacancies across hundreds of thousands of lattice sites, generating robust statistical insight into their distribution that can be correlated with specific synthesis pathways. This large-scale data enables us to classify a hierarchy of defect structures--from isolated vacancies to nanopores--revealing their preferred formation and interaction mechanisms, as corroborated by molecular dynamics simulations. This work provides a generalizable framework for understanding and ultimately controlling point defects across large volumes, paving the way for the rational design of defect-engineered functional 2D materials.", "date": "2025-11-11", "categories": []}
{"id": "2603.27195", "title": "AutoMS: Multi-Agent Evolutionary Search for Cross-Physics Inverse Microstructure Design", "abstract": "Designing microstructures with coupled cross-physics objectives is a fundamental challenge where traditional topology optimization is often computationally prohibitive and deep generative models frequently suffer from physical hallucinations. We introduce AutoMS, a multi-agent neuro-symbolic framework that reformulates inverse design as an LLM-driven evolutionary search. AutoMS leverages LLMs as semantic navigators to decompose complex requirements and coordinate agent workflows, while a novel Simulation-Aware Evolutionary Search (SAES) mechanism handles low-level numerical optimization via local gradient approximation and directed parameter updates. This architecture achieves a state-of-the-art 83.8% success rate on 17 diverse cross-physics tasks, significantly outperforming both traditional evolutionary algorithms and existing agentic baselines. By decoupling open-ended semantic orchestration from simulation-grounded numerical search, AutoMS provides a robust pathway for navigating complex physical landscapes that remain intractable for standard generative or purely linguistic approaches.", "date": "2026-03-28", "categories": []}
{"id": "2604.10157", "title": "Accelerated Dopant Screening in Oxide Semiconductors via Multi-Fidelity Contextual Bandits and a Three-Tier DFT Validation Funnel", "abstract": "Band gap engineering of oxide semiconductors through doping is critical for photocatalysis and optoelectronics, yet the combinatorial space of dopant elements, substitution sites, and co-doping combinations far exceeds typical density functional theory (DFT) budgets. We screen doped candidates across five oxide hosts (ZnO, TiO2, SrTiO3, SnO2, MgO), culminating in a 529-candidate ZnO co-doping campaign, and identify Cu-containing co-doped ZnO systems as consistently achieving visible-light-range band gaps (1.0-1.8 eV), with Y2Cu2 co-doped ZnO as the optimal candidate (1.84 eV). A three-tier validation funnel (PBE, PBE+U, ionic relaxation) reveals that no single level of theory suffices: V-doped ZnO shifts from near-metallic to wide-gap upon Hubbard U correction, while Cu-doped SrTiO3 enters the visible-light window only after correcting for d-electron localization.\n  To make this screening tractable, we introduce a multi-fidelity screening strategy that replaces 81% of DFT evaluations with computationally inexpensive surrogate predictions, reducing a 529-candidate closed-loop Quantum ESPRESSO campaign from an estimated 440 to 62 CPU-hours while finding the global optimum in 100% of 50 independent trials (p = 5.0e-8 versus random screening, Wilcoxon signed-rank). Cross-host analysis of the dopant-host interaction matrix reveals that dopant performance is governed by just two latent chemical dimensions, enabling prediction of rankings in unseen hosts. All 583 DFT calculations, screening code, and stability proofs are released as an open benchmark.", "date": "2026-04-11", "categories": []}
{"id": "2604.10173", "title": "Continuous PT-Symmetry Breaking as a Design Variable for Giant Altermagnetic Spin Splitting", "abstract": "Magnetic point-group analysis classifies altermagnets but returns only a binary symmetry verdict, leaving spin-splitting energy (SSE) inaccessible without spin-polarized density functional theory (DFT). This binary ceiling is not fundamental. Sublattice symmetry breaking is promoted here to a continuous, DFT-free scalar -- the Motif Symmetry-Breaking Index (MSBI) -- that quantifies $\\mathcal{PT}$-symmetry breaking between antiparallel magnetic motifs directly from crystal coordinates. SHAP analysis of an XGBoost surrogate trained on 3,851 DFT-labeled binary structures identifies three dominant descriptors: MSBI (symmetry-breaking axis), motif packing fraction MPF (superexchange axis), and the $p/d$ electron ratio (covalency axis), each mapping onto a directly tunable experimental handle. A controlled VO--CrSb comparison within the same P$6_3$/mmc host lattice demonstrates that composition alone boosts SSE sevenfold. Bayesian optimization over this three-axis space, followed by independent DFT validation, recovers $\\alpha$-NiS (SSE $= 0.823$\\,eV) as cross-validation against an independent symmetry-based prediction and identifies three previously unrecognized high-SSE candidates -- square-planar FeS (1.297\\,eV), octahedral CoS (1.103\\,eV), and FeAs (1.089\\,eV) -- all matching or exceeding CrSb. Square-planar Fe--S is proposed as a transferable coordination motif for giant altermagnetic spin splitting, advancing altermagnet design from symmetry classification to continuous quantitative optimization.", "date": "2026-04-11", "categories": []}
{"id": "2604.10568", "title": "ReadMOF: Structure-Free Semantic Embeddings from Systematic MOF Nomenclature for Machine Learning", "abstract": "Systematic chemical names, such as IUPAC-style nomenclature for metal-organic frameworks (MOFs), contain rich structural and compositional information in a standardized textual format. Here we introduce ReadMOF, which is, to our knowledge, the first nomenclature-free machine learning framework that leverages these names to model structure-property relationships without requiring atomic coordinates or connectivity graphs. By employing pretrained language models, ReadMOF converts systematic MOF names from the Cambridge Structural Database (CSD) into vector embeddings that closely represent traditional structure-based descriptors. These embeddings enable applications in materials informatics, including property prediction, similarity retrieval, and clustering, with performance comparable to geometry-dependent methods. When combined with large language models, ReadMOF also establishes chemically meaningful reasoning ability with textual input only. Our results show that structured chemical language, interpreted through modern natural language processing techniques, can provide a scalable, interpretable, and geometry-independent alternative to conventional molecular representations. This approach opens new opportunities for language-driven discovery in materials science.", "date": "2026-04-12", "categories": []}
{"id": "2604.11540", "title": "A collaborative agent with two lightweight synergistic models for autonomous crystal materials research", "abstract": "Current large language models require hundreds of billions of parameters yet struggle with domain-specific reasoning and tool coordination in materials science. Here, we present MatBrain, a lightweight collaborative agent system with two synergistic models specialization for crystal materials research. MatBrain employs a dual-model architecture: Mat-R1 (30B parameters) as the analytical model providing expert-level domain reasoning, and Mat-T1 (14B parameters) as the executive model orchestrating tool-based actions. Entropy analysis confirms that this architecture resolves the conflict between tool planning and analytical reasoning by decoupling their distinct entropy dynamics. Enabled by this dual-model architecture and structural efficiency, MatBrain significantly outperforms larger general-purpose models while reducing the hardware deployment barrier by over 95%. MatBrain exhibits versatility across structure generation, property prediction, and synthesis planning tasks. Applied to catalyst design, MatBrain generated 30,000 candidate structures and identified 38 promising materials within 48 hours, achieving approximately 100-fold acceleration over traditional approaches. These results demonstrate the potential of lightweight collaborative intelligence for advancing materials research capabilities.", "date": "2026-04-13", "categories": []}
{"id": "2604.08744", "title": "Active Learning for Generalizable Detonation Performance Prediction of Energetic Materials", "abstract": "The discovery of new energetic materials is critical for advancing technologies from defense to private industry. However, experimental approaches remain slow and expensive while computational alternatives require accurate material property inputs that are often costly to obtain, limiting their ability to efficiently predict detonation performance across a vast chemical space. We address this challenge through an active learning strategy that integrates density functional theory calculations, thermochemical modeling, message-passing neural networks, and Bayesian optimization. The resulting high-throughput workflow iteratively expands the training dataset by selecting new molecules in a targeted manner that balances the exploration of broad chemical space with the exploitation of promising high-performing candidates. This approach yields the largest publicly available database of potential CHNO explosives drawn from an initial pool of more than 70 billion candidates and a generalizable surrogate model capable of accurately predicting detonation performance (R$^2$ > 0.98). Feature importance analysis on this largest-to-date dataset reveals that oxygen balance is the dominant driver of detonation performance, complemented by contributions from local electronic structure, density, and the presence of specific functional groups. Cheminformatics analysis highlights how energetic materials with similar performance metrics tend to cluster in distinct chemical spaces offering a clearer direction for future synthesis studies. Together, the surrogate model, database, and resulting chemical insights provide a valuable foundation for high-throughput screening and targeted discovery of new energetic materials spanning diverse and previously unexplored regions of chemical space.", "date": "2026-04-09", "categories": []}
{"id": "2604.07746", "title": "Towards Rapid Constitutive Model Discovery from Multi-Modal Data: Physics Augmented Finite Element Model Updating (paFEMU)", "abstract": "Recent progress in AI-enabled constitutive modeling has concentrated on moving from a purely data-driven paradigm to the enforcement of physical constraints and mechanistic principles, a concept referred to as physics augmentation. Classical phenomenological approaches rely on selecting a pre-defined model and calibrating its parameters, while machine learning methods often focus on discovery of the model itself. Sparse regression approaches lie in between, where large libraries of pre-defined models are probed during calibration. Sparsification in the aforementioned paradigm, but also in the context of neural network architecture, has been shown to enable interpretability, uncertainty quantification, but also heterogeneous software integration due to the low-dimensional nature of the resulting models. Most works in AI-enabled constitutive modeling have also focused on data from a single source, but in reality, materials modeling workflows can contain data from many different sources (multi-modal data), and also from testing other materials within the same materials class (multi-fidelity data). In this work, we introduce physics augmented finite element model updating (paFEMU), as a transfer learning approach that combines AI-enabled constitutive modeling, sparsification for interpretable model discovery, and finite element-based adjoint optimization utilizing multi-modal data. This is achieved by combining simple mechanical testing data, potentially from a distinct material, with digital image correlation-type full-field data acquisition to ultimately enable rapid constitutive modeling discovery. The simplicity of the sparse representation enables easy integration of neural constitutive models in existing finite element workflows, and also enables low-dimensional updating during transfer learning.", "date": "2026-04-09", "categories": []}
{"id": "2604.07276", "title": "Making Room for AI: Multi-GPU Molecular Dynamics with Deep Potentials in GROMACS", "abstract": "GROMACS is a de-facto standard for classical Molecular Dynamics (MD). The rise of AI-driven interatomic potentials that pursue near-quantum accuracy at MD throughput now poses a significant challenge: embedding neural-network inference into multi-GPU simulations retaining high-performance. In this work, we integrate the MLIP framework DeePMD-kit into GROMACS, enabling domain-decomposed, GPU-accelerated inference across multi-node systems. We extend the GROMACS NNPot interface with a DeePMD backend, and we introduce a domain decomposition layer decoupled from the main simulation. The inference is executed concurrently on all processes, with two MPI collectives used each step to broadcast coordinates and to aggregate and redistribute forces. We train an in-house DPA-1 model (1.6 M parameters) on a dataset of solvated protein fragments. We validate the implementation on a small protein system, then we benchmark the GROMACS-DeePMD integration with a 15,668 atom protein on NVIDIA A100 and AMD MI250x GPUs up to 32 devices. Strong-scaling efficiency reaches 66% at 16 devices and 40% at 32; weak-scaling efficiency is 80% to 16 devices and reaches 48% (MI250x) and 40% (A100) at 32 devices. Profiling with the ROCm System profiler shows that >90% of the wall time is spent in DeePMD inference, while MPI collectives contribute <10%, primarily since they act as a global synchronization point. The principal bottlenecks are the irreducible ghost-atom cost set by the cutoff radius, confirmed by a simple throughput model, and load imbalance across ranks. These results demonstrate that production MD with near ab initio fidelity is feasible at scale in GROMACS.", "date": "2026-04-08", "categories": []}
{"id": "2602.00185", "title": "QUASAR: A Universal Autonomous System for Atomistic Simulation and a Benchmark of Its Capabilities", "abstract": "The integration of large language models (LLMs) into materials science offers a transformative opportunity to streamline computational workflows, yet current agentic systems remain constrained by rigid, carefully crafted domain-specific tool-calling paradigms and narrowly scoped agents. In this work, we introduce QUASAR, a universal autonomous system for atomistic simulation designed to facilitate production-grade scientific discovery. QUASAR autonomously orchestrates complex multi-scale workflows across diverse methods, including density functional theory, machine learning potentials, molecular dynamics, and Monte Carlo simulations. The system incorporates robust mechanisms for adaptive planning, context-efficient memory management, and hybrid knowledge retrieval to navigate real-world research scenarios without human intervention. We benchmark QUASAR against a series of three-tiered tasks, progressing from routine tasks to frontier research challenges such as photocatalyst screening and novel material assessment. These results suggest that QUASAR can function as a general atomistic reasoning system rather than a task-specific automation framework. They also provide initial evidence supporting the potential deployment of agentic AI as a component of computational chemistry research workflows, while identifying areas requiring further development.", "date": "2026-01-30", "categories": []}
{"id": "2604.03911", "title": "Align Your Structures: Generating Trajectories with Structure Pretraining for Molecular Dynamics", "abstract": "Generating molecular dynamics (MD) trajectories using deep generative models has attracted increasing attention, yet remains inherently challenging due to the limited availability of MD data and the complexities involved in modeling high-dimensional MD distributions. To overcome these challenges, we propose a novel framework that leverages structure pretraining for MD trajectory generation. Specifically, we first train a diffusion-based structure generation model on a large-scale conformer dataset, on top of which we introduce an interpolator module trained on MD trajectory data, designed to enforce temporal consistency among generated structures. Our approach effectively harnesses abundant structural data to mitigate the scarcity of MD trajectory data and effectively decomposes the intricate MD modeling task into two manageable subproblems: structural generation and temporal alignment. We comprehensively evaluate our method on the QM9 and DRUGS small-molecule datasets across unconditional generation, forward simulation, and interpolation tasks, and further extend our framework and analysis to tetrapeptide and protein monomer systems. Experimental results confirm that our approach excels in generating chemically realistic MD trajectories, as evidenced by remarkable improvements of accuracy in geometric, dynamical, and energetic measurements.", "date": "2026-04-05", "categories": []}
{"id": "2604.04636", "title": "Interpretation of Crystal Energy Landscapes with Kolmogorov-Arnold Networks", "abstract": "Characterizing crystalline energy landscapes is essential to predicting thermodynamic stability, electronic structure, and functional behavior. While machine learning (ML) enables rapid property predictions, the \"black-box\" nature of most models limits their utility for generating new scientific insights. Here, we introduce Kolmogorov-Arnold Networks (KANs) as an interpretable framework to bridge this gap. Unlike conventional neural networks with fixed activation functions, KANs employ learnable functions that reveal underlying physical relationships. We developed the Element-Weighted KAN, a composition-only model that achieves state-of-the-art accuracy in predicting formation energy, band gap, and work function across large-scale datasets. Crucially, without any explicit physical constraints, KANs uncover interpretable chemical trends aligned with the periodic table and quantum mechanical principles through embedding analysis, correlation studies, and principal component analysis. These results demonstrate that KANs provide a powerful framework with high predictive performance and scientific interpretability, establishing a new paradigm for transparent, chemistry-based materials informatics.", "date": "2026-04-06", "categories": []}
{"id": "2604.02524", "title": "AQVolt26: High-Temperature r$^2$SCAN Halide Dataset for Universal ML Potentials and Solid-State Batteries", "abstract": "The demand for safe, high-energy-density batteries has spotlighted halide solid-state electrolytes, which offer the potential for enhanced ionic mobility, electrochemical stability, and interfacial deformability. Accelerating their discovery requires extensive molecular dynamics, which has been increasingly enabled by universal machine learning interatomic potentials trained on foundational datasets. However, the dynamic softness of halides poses a stringent test of whether general-purpose models can reliably replace first-principles calculations under the highly distorted, elevated-temperature regimes necessary to probe ion transport. Here, we present AQVolt26, a dataset of 322,656 r$^2$SCAN single-point calculations for lithium halides, generated via high-temperature configurational sampling across $\\sim$5K structures. We demonstrate that foundational datasets provide a strong baseline for stable halide chemistries and transfer local forces well, however absolute energy predictions degrade in distorted higher-temperature regimes. Co-training with AQVolt26 resolves this blind spot. Furthermore, incorporating Materials Project relaxation data improves near-equilibrium performance but degrades extreme-strain robustness without enhancing high-temperature force accuracy. These results demonstrate that domain-specific configurational sampling is essential for the reliable dynamic screening of halide electrolytes. Furthermore, our findings suggest that while foundational models provide a robust base, they are most effective for dynamically soft solid-state chemistries when augmented with targeted, high-temperature data. Finally, we show that near-equilibrium relaxation data serves as a task-specific complement rather than a universally beneficial addition.", "date": "2026-04-02", "categories": []}
{"id": "2604.02537", "title": "PolyJarvis: LLM Agent for Autonomous Polymer MD Simulations", "abstract": "All-atom molecular dynamics (MD) simulations can predict polymer properties from molecular structure, yet their execution requires specialized expertise in force field selection, system construction, equilibration, and property extraction. We present PolyJarvis, an agent that couples a large language model (LLM) with the RadonPy simulation platform through Model Context Protocol (MCP) servers, enabling end-to-end polymer property prediction from natural language input. Given a polymer name or SMILES string, PolyJarvis autonomously executes monomer construction, charge assignment, polymerization, force field parameterization, GPU-accelerated equilibration, and property calculation. Validation is conducted on polyethylene (PE), atactic polystyrene (aPS), poly(methyl methacrylate) (PMMA), and poly(ethylene glycol) (PEG). Results show density predictions within 0.1--4.8% and bulk moduli within 17--24% of reference values for aPS and PMMA. PMMA glass transition temperature (Tg) (395~K) matches experiment within +10--18~K, while the remaining three polymers overestimate Tg by +38 to +47K (vs upper experimental bounds). Of the 8 property--polymer combinations with directly comparable experimental references, 5 meet strict acceptance criteria. For cases lacking suitable amorphous-phase experimental, agreement with prior MD literature is reported separately. The remaining Tg failures are attributable primarily to the intrinsic MD cooling-rate bias rather than agent error. This work demonstrates that LLM-driven agents can autonomously execute polymer MD workflows producing results consistent with expert-run simulations.", "date": "2026-04-02", "categories": []}
{"id": "2604.01106", "title": "Inverse Design of Optical Multilayer Thin Films using Robust Masked Diffusion Models", "abstract": "Inverse design of optical multilayer stacks seeks to infer layer materials, thicknesses, and ordering from a desired target spectrum. It is a long-standing challenge due to the large design space and non-unique solutions. We introduce \\texttt{OptoLlama}, a masked diffusion language model for inverse thin-film design from optical spectra. Representing multilayer stacks as sequences of material-thickness tokens, \\texttt{OptoLlama} conditions generation on reflectance, absorptance, and transmittance spectra and learns a probabilistic mapping from optical response to structure. Evaluated on a representative test set of 3,000 targets, \\texttt{OptoLlama} reduces the mean absolute spectral error by 2.9-fold relative to a nearest-neighbor template baseline and by 3.45-fold relative to the state-of-the-art data-driven baseline, called \\texttt{OptoGPT}. Case studies on designed and expert-defined targets show that the model reproduces characteristic spectral features and recovers physically meaningful stack motifs, including distributed Bragg reflectors. These results establish diffusion-based sequence modeling as a powerful framework for inverse photonic design.", "date": "2026-04-01", "categories": []}
{"id": "2510.00027", "title": "Learning Inter-Atomic Potentials without Explicit Equivariance", "abstract": "Accurate and scalable machine-learned inter-atomic potentials (MLIPs) are essential for molecular simulations ranging from drug discovery to new material design. Current state-of-the-art models enforce roto-translational symmetries through equivariant neural network architectures, a hard-wired inductive bias that can often lead to reduced flexibility, computational efficiency, and scalability. In this work, we introduce TransIP: Transformer-based Inter-Atomic Potentials, a novel training paradigm for interatomic potentials achieving symmetry compliance without explicit architectural constraints. Our approach guides a generic non-equivariant Transformer-based model to learn SO(3)-equivariance by optimizing its representations in the embedding space. Trained on the recent Open Molecules (OMol25) collection, a large and diverse molecular dataset built specifically for MLIPs and covering different types of molecules (including small organics, biomolecular fragments, and electrolyte-like species), TransIP attains comparable performance in machine-learning force fields versus state-of-the-art equivariant baselines. Further, compared to a data augmentation baseline, TransIP achieves 40% to 60% improvement in performance across varying OMol25 dataset sizes. More broadly, our work shows that learned equivariance can be a powerful and efficient alternative to equivariant or augmentation-based MLIP models. Our code is available at: https://github.com/Ahmed-A-A-Elhag/TransIP.", "date": "2025-09-25", "categories": []}
{"id": "2510.12405", "title": "Continuous SUN (Stable, Unique, and Novel) Metric for Generative Modeling of Inorganic Crystals", "abstract": "To address pressing scientific challenges such as climate change, increasingly sophisticated generative models are being developed to efficiently sample the large chemical space of potential functional materials. The proliferation of these models has necessitated the establishment of rigorous evaluation metrics. While uniqueness (U), novelty (N), and stability (S) of samples serve as standard metrics, their current formulations show several limitations. U and N rely on binary comparisons of crystals, rendering them dependent on heuristic thresholds, incapable of quantifying the degree of similarity, sensitive to atomic coordinate perturbations, and not invariant to sample permutation. Similarly, the binary assessment of S risks a premature exclusion of marginally unstable yet potentially novel candidates. These limitations are addressed by making the aforementioned metrics continuous. Furthermore, we integrate them into a unified metric ``continuous SUN\" (cSUN), which offers a smoother score distribution and greater tunability than the conventional binary SUN metric. Experimental results demonstrate that our continuous metrics provide granular insights into sample distributions and facilitate the identification of the most promising candidates. Finally, the use of cSUN as a reward signal in reinforcement learning is explored, showing that its adjustable weighting scheme effectively mitigates reward hacking and avoids local minima.", "date": "2025-10-14", "categories": []}
{"id": "2603.29152", "title": "SimMOF: AI agent for Automated MOF Simulations", "abstract": "Metal-organic frameworks (MOFs) offer a vast design space, and as such, computational simulations play a critical role in predicting their structural and physicochemical properties. However, MOF simulations remain difficult to access because reliable analysis require expert decisions for workflow construction, parameter selection, tool interoperability, and the preparation of computational ready structures. Here, we introduce SimMOF, a large language model based multi agent framework that automates end-to-end MOF simulation workflows from natural language queries. SimMOF translates user requests into dependency aware plans, generates runnable inputs, orchestrates multiple agents to execute simulations, and summarizes results with analysis aligned to the user query. Through representative case studies, we show that SimMOF enables adaptive and cognitively autonomous workflows that reflect the iterative and decision driven behavior of human researchers and as such provides a scalable foundation for data driven MOF research.", "date": "2026-03-31", "categories": []}
{"id": "2603.29812", "title": "AMShortcut: An Inference- and Training-Efficient Inverse Design Model for Amorphous Materials", "abstract": "Amorphous materials are solids that lack long-range atomic order but possess complex short- and medium-range order. Unlike crystalline materials that can be described by unit cells containing few up to hundreds of atoms, amorphous materials require larger simulation cells with at least hundreds or often thousands of atoms. Inverse design of amorphous materials with probabilistic generative models aims to generate the atomic positions and elements of amorphous materials given a set of desired properties. It has emerged as a promising approach for facilitating the application of amorphous materials in domains such as energy storage and thermal management. In this paper, we introduce AMShortcut, an inference- and training-efficient probabilistic generative model for amorphous materials. AMShortcut enables accurate inference of diverse short- and medium-range structures in amorphous materials with only a few sampling steps, mitigating the need for an excessive number of sampling steps that hinders inference efficiency. AMShortcut can be trained once with all relevant properties and perform inference conditioned on arbitrary combinations of desired properties, mitigating the need for training one model for each combination. Experiments on three amorphous materials datasets with diverse structures and properties demonstrate that AMShortcut achieves its design goals.", "date": "2026-03-31", "categories": []}
{"id": "2409.15675", "title": "The Northeast Materials Database for Magnetic Materials", "abstract": "The discovery of magnetic materials with high operating temperature ranges and optimized performance is essential for advanced applications. Current data-driven approaches are limited by the lack of accurate, comprehensive, and feature-rich databases. This study aims to address this challenge by using Large Language Models (LLMs) to create a comprehensive, experiment-based, magnetic materials database named the Northeast Materials Database (NEMAD), which consists of 67,573 magnetic materials entries(www.nemad.org). The database incorporates chemical composition, magnetic phase transition temperatures, structural details, and magnetic properties. Enabled by NEMAD, we trained machine learning models to classify materials and predict transition temperatures. Our classification model achieved an accuracy of 90% in categorizing materials as ferromagnetic (FM), antiferromagnetic (AFM), and non-magnetic (NM). The regression models predict Curie (N\\'eel) temperature with a coefficient of determination (R2) of 0.87 (0.83) and a mean absolute error (MAE) of 56K (38K). These models identified 25 (13) FM (AFM) candidates with a predicted Curie (N\\'eel) temperature above 500K (100K) from the Materials Project. This work shows the feasibility of combining LLMs for automated data extraction and machine learning models to accelerate the discovery of magnetic materials.", "date": "2024-09-24", "categories": []}
{"id": "2508.13197", "title": "The Rise of Generative AI for Metal-Organic Framework Design and Synthesis", "abstract": "Advances in generative artificial intelligence are transforming how metal-organic frameworks (MOFs) are designed and discovered. This Perspective introduces the shift from laborious enumeration of MOF candidates to generative approaches that can autonomously propose and synthesize in the laboratory new porous reticular structures on demand. We outline the progress of employing deep learning models, such as variational autoencoders, diffusion models, and large language model-based agents, that are fueled by the growing amount of available data from the MOF community and suggest novel crystalline materials designs. These generative tools can be combined with high-throughput computational screening and even automated experiments to form accelerated, closed-loop discovery pipelines. The result is a new paradigm for reticular chemistry in which AI algorithms more efficiently direct the search for high-performance MOF materials for clean air and energy applications. Finally, we highlight remaining challenges such as synthetic feasibility, dataset diversity, and the need for further integration of domain knowledge.", "date": "2025-08-15", "categories": []}
{"id": "2603.27886", "title": "AI-ready design of realistic 2D materials and interfaces with Mat3ra-2D", "abstract": "Artificial intelligence (AI) and machine learning (ML) models in materials science are predominantly trained on ideal bulk crystals, limiting their transferability to real-world applications where surfaces, interfaces, and defects dominate. We present Mat3ra-2D, an open-source framework for the rapid design of realistic two-dimensional materials and related structures, including slabs and heterogeneous interfaces, with support for disorder and defect-driven complexity. The approach combines: (1) well-defined standards for storing and exchanging materials data with a modular implementation of core concepts and (2) transformation workflows expressed as configuration-builder pipelines that preserve provenance and metadata. We implement typical structure generation tasks, such as constructing orientation-specific slabs or strain-matching interfaces, in reusable Jupyter notebooks that serve as both interactive documentation and templates for reproducible runs. To lower the barrier to adoption, we design the examples to run in any web browser and demonstrate how to incorporate these developments into a web application. Mat3ra-2D enables systematic creation and organization of realistic 2D- and interface-aware datasets for AI/ML-ready applications.", "date": "2026-03-29", "categories": []}
{"id": "2603.25894", "title": "Data-Driven Plasticity Modeling via Acoustic Profiling", "abstract": "This paper presents a data-driven framework for modeling plastic deformation in crystalline metals through acoustic emission (AE) analysis. Building on experimental data from compressive loading of nickel micropillars, the study introduces a wavelet-based method using Morlet transforms to detect AE events across distinct frequency bands, enabling identification of both large and previously overlooked small-scale events. The detected events are validated against stress-drop dynamics, demonstrating strong physical consistency and revealing a relationship between AE energy release and strain evolution, including the onset of increased strain rate following major events. Leveraging labeled datasets of events and non-events, the work applies machine learning techniques, showing that engineered time and frequency domain features significantly outperform raw signal classifiers, and identifies key discriminative features such as RMS amplitude, zero crossing rate, and spectral centroid. Finally, clustering analysis uncovers four distinct AE event archetypes corresponding to different deformation mechanisms, highlighting the potential for transitioning from retrospective analysis to predictive modeling of material behavior using acoustic signals.", "date": "2026-03-26", "categories": []}
{"id": "2603.26517", "title": "The internal law of a material can be discovered from its boundary", "abstract": "Since the earliest stages of human civilization, advances in technology have been tightly linked to our ability to understand and predict the mechanical behavior of materials. In recent years, this challenge has increasingly been framed within the broader paradigm of data-driven scientific discovery, where governing laws are inferred directly from observations. However, existing methods require either stress-strain pairs or full-field displacement measurements, which are often inaccessible in practice. We introduce Neural-DFEM, a method that enables unsupervised discovery of hyperelastic material laws even from partial observations, such as boundary-only measurements. The method embeds a differentiable finite element solver within the learning loop, directly linking candidate energy functionals to available measurements. To guarantee thermodynamic consistency and mathematical well-posedness throughout training, the method employs Hyperelastic Neural Networks, a novel structure-preserving neural architecture that enforces frame indifference, material symmetry, polyconvexity, and coercivity by design. The resulting framework enables robust material model discovery in both two- and three-dimensional settings, including scenarios with boundary-only measurements. Neural-DFEM allows for generalization across geometries and loading conditions, and exhibits unprecedented accuracy and strong resilience to measurement noise. Our results demonstrate that reliable identification of material laws is achievable even under partial observability when strong physical inductive biases are embedded in the learning architecture.", "date": "2026-03-27", "categories": []}
{"id": "2603.25373", "title": "Hessian-informed machine learning interatomic potential towards bridging theory and experiments", "abstract": "Local curvature of potential energy surfaces is critical for predicting certain experimental observables of molecules and materials from first principles, yet it remains far beyond reach for complex systems. In this work, we introduce a Hessian-informed Machine Learning Interatomic Potential (Hi-MLIP) that captures such curvature reliably, thereby enabling accurate analysis of associated thermodynamic and kinetic phenomena. To make Hessian supervision practically viable, we develop a highly efficient training protocol, termed Hessian INformed Training (HINT), achieving two to four orders of magnitude reduction for the requirement of expensive Hessian labels. HINT integrates critical techniques, including Hessian pre-training, configuration sampling, curriculum learning and stochastic projection Hessian loss. Enabled by HINT, Hi-MLIP significantly improves transition-state search and brings Gibbs free-energy predictions close to chemical accuracy especially in data-scarce regimes. Our framework also enables accurate treatment of strongly anharmonic hydrides, reproducing phonon renormalization and superconducting critical temperatures in close agreement with experiment while bypassing the computational bottleneck of anharmonic calculations. These results establish a practical route to enhancing curvature awareness of machine learning interatomic potentials, bridging simulation and experimental observables across a wide range of systems.", "date": "2026-03-26", "categories": []}
{"id": "2505.00574", "title": "Transition States Energies from Machine Learning: An Application to Reverse Water-Gas Shift on Single-Atom Alloys", "abstract": "Obtaining accurate transition state (TS) energies is a bottleneck in computational screening of complex materials and reaction networks due to the high cost of TS search methods and first-principles methods such as density functional theory (DFT). Here we propose a machine learning (ML) model for predicting TS energies based on Gaussian process regression with the Wasserstein Weisfeiler-Lehman graph kernel (WWL-GPR). Applying the model to predict adsorption and TS energies for the reverse water-gas shift (RWGS) reaction on single-atom alloy (SAA) catalysts, we show that it can significantly improve the accuracy compared to traditional approaches based on scaling relations or ML models without a graph representation. Further benefitting from the low cost of model training, we train an ensemble of WWL-GPR models to obtain uncertainties through subsampling of the training data and show how these uncertainties propagate to turnover frequency (TOF) predictions through the construction of an ensemble of microkinetic models. Comparing the errors in model-based vs DFT-based TOF predictions, we show that the WWL-GPR model reduces errors by almost an order of magnitude compared to scaling relations. This demonstrates the critical impact of accurate energy predictions on catalytic activity estimation. Finally, we apply our model to screen new materials, identifying promising catalysts for RWGS. This work highlights the power of combining advanced ML techniques with DFT and microkinetic modeling for screening catalysts for complex reactions like RWGS, providing a robust framework for future catalyst design.", "date": "2025-05-01", "categories": []}
{"id": "2603.23943", "title": "ChargeFlow: Flow-Matching Refinement of Charge-Conditioned Electron Densities", "abstract": "Accurate charge densities are central to electronic-structure theory, but computing charge-state-dependent densities with density functional theory remains too expensive for large-scale screening and defect workflows. We present ChargeFlow, a flow-matching refinement model that transforms a charge-conditioned superposition of atomic densities into the corresponding DFT electron density on the native periodic real-space grid using a 3D U-Net velocity field. Trained on 9,502 charged Materials Project-derived calculations and evaluated on an external 1,671-structure benchmark spanning perovskites, charged defects, diamond defects, metal-organic frameworks, and organic crystals, ChargeFlow is not uniformly best on every in-distribution class but is strongest on problems dominated by nonlocal charge redistribution and charge-state extrapolation, improving deformation-density error from 3.62% to 3.21% and charge- response cosine similarity from 0.571 to 0.655 relative to a ResNet baseline. The predicted densities remain chemically useful under downstream analysis, yielding successful Bader partitioning on all 1,671 benchmark structures and high-fidelity electrostatic potentials, which positions flow matching as a practical density-refinement strategy for charged materials.", "date": "2026-03-25", "categories": []}
{"id": "2507.21748", "title": "evoxels: A differentiable physics framework for voxel-based microstructure simulations", "abstract": "Materials science inherently spans disciplines: experimentalists use advanced microscopy to uncover micro- and nanoscale structure, while theorists and computational scientists develop models that link processing, structure, and properties. Bridging these domains is essential for inverse material design where you start from desired performance and work backwards to optimal microstructures and manufacturing routes. Integrating high-resolution imaging with predictive simulations and data-driven optimization accelerates discovery and deepens understanding of process-structure-property relationships. The differentiable physics framework evoxels is based on a fully Pythonic, unified voxel-based approach that integrates segmented 3D microscopy data, physical simulations, inverse modeling, and machine learning.", "date": "2025-07-29", "categories": []}
{"id": "2603.12567", "title": "Foundation-Model Surrogates Enable Data-Efficient Active Learning for Materials Discovery", "abstract": "Active learning (AL) has emerged as a powerful paradigm for accelerating materials discovery by iteratively steering experiments toward promising candidates, reducing the number of costly synthesis-and-characterization cycles needed to identify optimal materials. However, current AL relies predominantly on Gaussian Process (GP) and Random Forest (RF) surrogates, which suffer from complementary limitations: GP underfits complex composition-property landscapes due to rigid kernel assumptions, while RF produces unreliable heuristic uncertainty estimates in small-data regimes. This small-data challenge is pervasive in materials science, making reliable surrogate modeling extremely difficult with models trained from scratch on each new dataset. Here we propose In-Context Active Learning (ICAL), which addresses this bottleneck by replacing conventional surrogates with TabPFN, a transformer-based foundation model (FM) pre-trained on millions of synthetic regression tasks to meta-learn a universal prior over tabular data, upon which TabPFN performs principled Bayesian inference in a single forward pass without dataset-specific retraining, delivering strong small-data regression performance and well-calibrated predictive uncertainty (required for effective AL). We benchmark ICAL against GP and RF across 10 materials datasets and TabPFN wins on 8 out of 10 datasets, achieving a mean saving of 52% in extra evaluations relative to GP and 29.77% relative to RF. Cross-validation analysis confirms that TabPFN's advantage stems from superior uncertainty calibration, achieving the lowest Negative Log-Likelihood and Area Under the Sparsification Error curve among all surrogates. These results demonstrate that pre-trained FMs can serve as effective surrogates for active learning, enabling data-efficient discovery across diverse materials systems and small-data experimental sciences.", "date": "2026-03-13", "categories": []}
{"id": "2603.23210", "title": "Generative Inversion of Spectroscopic Data for Amorphous Structure Elucidation", "abstract": "Determining atomistic structures from characterization data is one of the most common yet intricate problems in materials science. Particularly in amorphous materials, proposing structures that balance realism and agreement with experiments requires expert guidance, good interatomic potentials, or both. Here, we introduce GLASS, a generative framework that inverts multi-modal spectroscopic measurements into realistic atomistic structures without knowledge of the potential energy surface. A score-based model learns a structural prior from low-fidelity data and samples out-of-distribution structures conditioned on differentiable spectral targets. Reconstructions using pair distribution functions (PDFs), X-ray absorption spectroscopy, and diffraction measurements quantify the complementarity between spectral modalities and demonstrate that PDFs is the most informative probe for our framework. We use GLASS to rationalize three contested experimental problems: paracrystallinity in amorphous silicon, a liquid-liquid phase transition in sulfur, and ball-milled amorphous ice. In each case, generated structures reproduce experimental measurements and reveal mechanisms inaccessible to diffraction analysis alone.", "date": "2026-03-24", "categories": []}
{"id": "2312.17251", "title": "MatSegNet: a New Boundary-aware Deep Learning Model for Accurate Carbide Precipitate Analysis in High-Strength Steels", "abstract": "Lower Bainite (LB) and Tempered Martensite (TM) are two common microstructures in modern high-strength steels. LB and TM can render similar mechanical properties for steels, yet LB is often considered superior to TM in resistance to hydrogen embrittlement. Such performance difference has conventionally been attributed to their distinction in certain microstructural features, particularly carbides. The present study developed, MatSegNet, a new contour-aware deep learning (DL) architecture. It is tailored for comprehensive segmentation and quantitative characterization of carbide precipitates with complex contours in high-strength steels, shown to outperform existing state-of-the-art DL architectures. Based on MatSegNet, a high-throughput DL pipeline has been established for precise comparative carbide analysis in LB and TM. The results showed that statistically the two microstructures exhibit similarity in key carbide characteristics with marginal difference, cautioning against the conventional use of carbide orientation as a reliable means to differentiate LB and TM in practice. Through MatSegNet, this work demonstrated the potential of DL to play a critical role in enabling accurate and quantitative microstructure characterization to facilitate development of structure-property relationships for accelerating materials innovation.", "date": "2023-12-02", "categories": []}
{"id": "2509.05909", "title": "Learning Magnetic Order Classification from Large-Scale Materials Databases", "abstract": "The reliable identification of magnetic ground states remains a major challenge in high-throughput materials databases, where density functional theory (DFT) workflows often converge to ferromagnetic (FM) solutions. Here, we partially address this challenge by developing machine learning classifiers trained on experimentally validated MAGNDATA magnetic materials leveraging a limited number of simple compositional, structural, and electronic descriptors sourced from the Materials Project database. Our propagation vector classifiers achieve accuracies above 92%, outperforming recent studies in reliably distinguishing zero from nonzero propagation vector structures, and exposing a systematic ferromagnetic bias inherent to the Materials Project database for more than 7,843 materials. In parallel, LightGBM and XGBoost models trained directly on the Materials Project labels achieve accuracies of 84-86% (with macro F1 average scores of 63-66%), which proves useful for large-scale screening for magnetic classes, if refined by MAGNDATA-trained classifiers. These results underscore the role of machine learning techniques as corrective and exploratory tools, enabling more trustworthy databases and accelerating progress toward the identification of materials with various properties.", "date": "2025-09-07", "categories": []}
{"id": "2510.24523", "title": "Unsupervised Machine-Learning Pipeline for Data-Driven Defect Detection and Characterisation: Application to Displacement Cascades", "abstract": "Neutron irradiation produces, within a few picoseconds, displacement cascades that are sequences of atomic collisions generating point and extended defects which subsequently affects the long-term evolution of materials. The diversity of these defects, characterized morphologically and statistically, defines what is called the \"primary damage\". In this work, we present a fully unsupervised machine learning (ML) workflow that detects and classifies these defects directly from molecular dynamics data. Local environments are encoded by the Smooth Overlap of Atomic Positions (SOAP) vector, anomalous atoms are isolated with autoencoder neural networks (AE), embedded with Uniform Manifold Approximation and Projection (UMAP) and clustered using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). Applied to 80 keV displacement cascades in Ni, Fe$_7$0Ni$_{10}$Cr$_{20}$, and Zr, the AE successfully identify the small fraction of outlier atoms that participate in defect formation. HDBSCAN then partitions the UMAP latent space of AE-flagged SOAP descriptors into well defined groups representing vacancy- and interstitial-dominated regions and, within each, separates small from large aggregates, assigning 99.7 % of outliers to compact physical motifs. A signed cluster-identification score confirms this separation, and cluster size scales with net defect counts (R2 > 0.89). Statistical cross analyses between the ML outlier map and several conventional detectors (centrosymmetry, dislocation extraction, etc.) reveal strong overlap and complementary coverage, all achieved without template or threshold tuning. This ML workflow thus provides an efficient tool for the quantitative mapping of structural anomalies in materials, particularly those arising from irradiation damage in displacement cascades.", "date": "2025-10-28", "categories": []}
{"id": "2511.01946", "title": "COFAP: A Universal Framework for COFs Adsorption Prediction through Designed Multi-Modal Extraction and Cross-Modal Synergy", "abstract": "Covalent organic frameworks (COFs) are promising adsorbents for gas adsorption and separation, while identifying the optimal structures among their vast design space requires efficient high-throughput screening. Conventional machine-learning predictors rely heavily on specific gas-related features. However, these features are time-consuming and limit scalability, leading to inefficiency and labor-intensive processes. Herein, a universal COFs adsorption prediction framework (COFAP) is proposed, which can extract multi-modal structural and chemical features through deep learning, and fuse these complementary features via cross-modal attention mechanism. Without relying on explicit gas-specific thermodynamic descriptors, COFAP achieves state-of-the-art prediction performance on the hypoCOFs dataset under the conditions investigated in this study, outperforming existing approaches. Based on COFAP, we also found that high-performing COFs for gas separation concentrate within a narrow range of pore size and surface area. A weight-adjustable prioritization scheme is also developed to enable flexible, application-specific ranking of candidate COFs for researchers. Superior efficiency and accuracy render COFAP directly deployable in crystalline porous materials.", "date": "2025-11-03", "categories": []}
{"id": "2511.12260", "title": "Reinforcement Learning for Chemical Ordering in Alloy Nanoparticles", "abstract": "We approach the search for optimal element ordering in bimetallic alloy nanoparticles (NPs) as a reinforcement learning (RL) problem and have built an RL agent that learns to perform such global optimization using the geometric graph representation of the NPs. To demonstrate the effectiveness, we train an RL agent to perform composition-conserving atomic swap actions on the icosahedral nanoparticle structure. Trained once on randomized $Ag_{X}Au_{309-X}$ compositions and orderings, the agent discovers previously established ground state structure. We show that this optimization is robust to differently ordered initialisations of the same NP compositions. We also demonstrate that a trained policy can extrapolate effectively to NPs of unseen size. However, the efficacy is limited when multiple alloying elements are involved. Our results demonstrate that RL with pre-trained equivariant graph encodings can navigate combinatorial ordering spaces at the nanoparticle scale, and offer a transferable optimization strategy with the potential to generalize across composition and reduce repeated individual search cost.", "date": "2025-11-15", "categories": []}
{"id": "2511.17848", "title": "Scaling Kinetic Monte-Carlo Simulations of Grain Growth with Combined Convolutional and Graph Neural Networks", "abstract": "Graph neural networks (GNN) have emerged as a promising machine learning method for microstructure simulations such as grain growth. However, accurate modeling of realistic grain boundary networks requires large simulation cells, which GNN has difficulty scaling up to. To alleviate the computational costs and memory footprint of GNN, we propose a hybrid architecture combining a convolutional neural network (CNN) based bijective autoencoder to compress the spatial dimensions, and a GNN that evolves the microstructure in the latent space of reduced spatial sizes. Our results demonstrate that the new design significantly reduces computational costs with using fewer message passing layer (from 12 down to 3) compared with GNN alone. The reduction in computational cost becomes more pronounced as the spatial size increases, indicating strong computational scalability. For the largest mesh evaluated (160^3), our method reduces memory usage and runtime in inference by 117x and 115x, respectively, compared with GNN-only baseline. More importantly, it shows higher accuracy and stronger spatiotemporal capability than the GNN-only baseline, especially in long-term testing. Such combination of scalability and accuracy is essential for simulating realistic material microstructures over extended time scales. The improvements can be attributed to the bijective autoencoder's ability to compress information losslessly from spatial domain into a high dimensional feature space, thereby producing more expressive latent features for the GNN to learn from, while also contributing its own spatiotemporal modeling capability. The training was optimized to learn from the stochastic Potts Monte Carlo method. Our findings provide a highly scalable approach for simulating grain growth.", "date": "2025-11-22", "categories": []}
{"id": "2603.20389", "title": "A chemical language model for reticular materials design", "abstract": "Reticular chemistry has enabled the synthesis of tens of thousands of metal-organic frameworks (MOFs), yet the discovery of new materials still relies largely on intuition-driven linker design and iterative experimentation. As a result, researchers explore only a small fraction of the vast chemical space accessible to reticular materials, limiting the systematic discovery of frameworks with targeted properties. Here, we introduce Nexerra-R1, a building-block chemical language model that enables inverse design in reticular chemistry through the targeted generation of organic linkers. Rather than generating complete frameworks directly, Nexerra-R1 operates at the level of molecular building blocks, preserving the modular logic that underpins reticular synthesis. The model supports both unconstrained generation of low-connectivity linkers and scaffold-constrained design of symmetric multidentate motifs compatible with predefined nodes and topologies. We further combine linker generation with flow-guided distributional targeting to steer the generative process toward application-relevant objectives while maintaining chemical validity and assembly feasibility. The generated linkers are subsequently assembled into three-dimensional frameworks and are structurally optimized to produce candidate materials compatible with experimental synthesis. Using Nexerra-R1, we validate this strategy by rediscovering known MOFs and by proposing the experimental synthesis of a previously unreported framework, CU-525, generated entirely in silico. Together, these results establish a general inverse-design paradigm for reticular materials in which controllable chemical language modelling enables the direct translation from computational design to synthesizable frameworks.", "date": "2026-03-20", "categories": []}
{"id": "2603.20986", "title": "AutoMOOSE: An Agentic AI for Autonomous Phase-Field Simulation", "abstract": "Multiphysics simulation frameworks such as MOOSE provide rigorous engines for phase-field materials modeling, yet adoption is constrained by the expertise required to construct valid input files, coordinate parameter sweeps, diagnose failures, and extract quantitative results. We introduce AutoMOOSE, an open-source agentic framework that orchestrates the full simulation lifecycle from a single natural-language prompt. AutoMOOSE deploys a five-agent pipeline in which the Input Writer coordinates six sub-agents and the Reviewer autonomously corrects runtime failures without user intervention. A modular plugin architecture enables new phase-field formulations without modifying the core framework, and a Model Context Protocol (MCP) server exposes the workflow as ten structured tools for interoperability with any MCP-compatible client. Validated on a four-temperature copper grain growth benchmark, AutoMOOSE generates MOOSE input files with 6 of 12 structural blocks matching a human expert reference exactly and 4 functionally equivalent, executes all runs in parallel with a 1.8x speedup, and performs an end-to-end physical consistency check spanning intent, finite-element execution, and Arrhenius kinetics with no human verification. Grain coarsening kinetics are recovered with R^2 = 0.90-0.95 at T >= 600 K; the recovered activation energy Q_fit = 0.296 eV is consistent with a human-written reference (Q_fit = 0.267 eV) under identical parameters. Three runtime failure classes were diagnosed and resolved autonomously within a single correction cycle, and every run produces a provenance record satisfying FAIR data principles. These results show that the gap between knowing the physics and executing a validated simulation campaign can be bridged by a lightweight multi-agent orchestration layer, providing a pathway toward AI-driven materials discovery and self-driving laboratories.", "date": "2026-03-22", "categories": []}
{"id": "2603.22254", "title": "Characterizing High-Capacity Janus Aminobenzene-Graphene Anode for Sodium-Ion Batteries with Machine Learning", "abstract": "Sodium-ion batteries require anodes that combine high capacity, low operating voltage, fast Na-ion transport, and mechanical stability, which conventional anodes struggle to deliver. Here, we use the SpookyNet machine-learning force field (MLFF) together with all-electron density-functional theory calculations to characterize Na storage in aminobenzene-functionalized Janus graphene (Na$_x$AB) at room-temperature. Simulations across state of charge reveal a three-stage storage mechanism-site-specific adsorption at aminobenzene groups and Na$_n$@AB$_m$ structure formation, followed by interlayer gallery filling-contrasting the multi-stage pore-, graphite-interlayer-, and defect-controlled behavior in hard carbon. This leads to an OCV profile with an extended low-voltage plateau of 0.15 V vs. Na/Na$^{+}$, an estimated gravimetric capacity of $\\sim$400 mAh g$^{-1}$, negligible volume change, and Na diffusivities of $\\sim10^{-6}$ cm$^{2}$ s$^{-1}$, two to three orders of magnitude higher than in hard carbon. Our results establish Janus aminobenzene-graphene as a promising, structurally defined high-capacity Na-ion anode and illustrate the power of MLFF-based simulations for characterizing electrode materials.", "date": "2026-03-23", "categories": []}
{"id": "2512.10989", "title": "Generalization of Long-Range Machine Learning Potentials in Complex Chemical Spaces", "abstract": "The vastness of chemical space makes generalization a central challenge in the development of machine learning interatomic potentials (MLIPs). While MLIPs could enable large-scale atomistic simulations with near-quantum accuracy, their usefulness is often limited by poor transferability to out-of-distribution samples. Here, we systematically evaluate different MLIP architectures with long-range corrections across diverse chemical spaces and show that such schemes are essential, not only for improving in-distribution performance but, more importantly, for enabling significant gains in transferability to unseen regions of chemical space. To enable a more rigorous benchmarking, we introduce biased train-test splitting strategies, which explicitly test the model performance in significantly different regions of chemical space. Together, our findings highlight the importance of long-range modeling for achieving generalizable MLIPs and provide a framework for diagnosing systematic failures across chemical space. Although we demonstrate our methodology on metal-organic frameworks, it is broadly applicable to other materials, offering insights into the design of more robust and transferable MLIPs.", "date": "2025-12-08", "categories": []}
{"id": "2601.18032", "title": "Multimodal Machine Learning for Soft High-k Elastomers under Data Scarcity", "abstract": "Dielectric materials are critical building blocks for modern electronics such as sensors, actuators, and transistors. With rapid advances in soft and stretchable electronics for emerging human- and robot-interfacing applications, there is a growing need for high-performance dielectric elastomers. However, developing soft elastomers that simultaneously exhibit high dielectric constants (k) and low Young's moduli (E) remains a major challenge. Although individual elastomer designs have been reported, structured datasets that systematically integrate molecular sequence, dielectric, and mechanical properties are largely unavailable. To address this gap, we curate a compact, high-quality dataset of acrylate-based dielectric elastomers by aggregating experimental results from the past decade. Building on this dataset, we propose a multimodal learning framework leveraging large-scale pretrained polymer representations. These pretrained embeddings transfer chemical and structural knowledge from vast polymer corpora, enabling accurate few-shot prediction of dielectric and mechanical properties and accelerating data-efficient discovery of soft high-$k$ dielectric elastomers. Our data and implementation are publicly available at: https://github.com/HySonLab/Polymers", "date": "2026-01-25", "categories": []}
{"id": "2603.16925", "title": "Gaussian Process Regression-based Knowledge Distillation Framework for Simultaneous Prediction of Physical and Mechanical Properties of Epoxy Polymers", "abstract": "Epoxy polymers are widely used due to their multifunctional properties, but machine learning (ML) applications remain limited owing to their complex 3D molecular structure, multi-component nature, and lack of curated datasets. Existing ML studies are largely restricted to simulation data, specific properties, or narrow constituent ranges. To address these limitations, we developed an informed Gaussian Process Regression-based Knowledge Distillation (GPR-KD) framework for predicting multiple physical (glass transition temperature, density) and mechanical properties (elastic modulus, tensile strength, compressive strength, flexural strength, fracture energy, adhesive strength) of thermoset epoxy polymers. The model was trained on experimental literature data covering diverse monomer classes (9 resins, 40 hardeners). Individual GPR models serve as teacher models capturing nonlinear feature-property relationships, while a unified neural network student model learns distilled knowledge across all properties simultaneously. By encoding the target property as an input feature, the student model leverages cross-property correlations. Molecular-level descriptors extracted from SMILES representations using RDKit create a physics-informed model. The framework combines GPR interpretability and robustness with deep learning scalability and generalization. Comparative analysis demonstrates superior prediction accuracy over conventional ML models. Simultaneous multi-property prediction further improves accuracy through information sharing across correlated properties. The proposed framework enables accelerated design of novel epoxy polymers with tailored properties.", "date": "2026-03-12", "categories": []}
{"id": "2603.16959", "title": "Machine intelligence supports the full chain of 2D dendrite synthesis", "abstract": "Exemplified by the chemical vapor deposition growth of two-dimensional dendrites, which has potential applications in catalysis and presents a parameter-intensive, data-scarce and reaction process-complex model problem, we devise a machine intelligence-empowered framework for the full chain support of material synthesis, encompassing rapid process optimization, accurate customized synthesis, and comprehensive mechanism deciphering.First, active learning is integrated into the experimental workflow, identifying an optimal recipe for the growth of highly-branched, electrocatalytically-active ReSe2 dendrites through 60 experiments (4 iterations), which account for less than 1.3% of the numerous possible parameter combinations.Then, a prediction accuracy-guided data augmentation strategy is developed combined with a tree-based machine learning (ML) algorithm, unveiling a non-linear correlation between 5 process variables and fractal dimension (DF) of ReSe2 dendrites with only 9 experiment additions, which guides the synthesis of various user-defined DF. Finally, we construct a data-knowledge dual-driven mechanism model by integration of cross-scale characterizations, interpretable ML models, and domain knowledge in thermodynamics and kinetics, unraveling synergistic contributions of multiple process parameters to the product morphology. This work demonstrates the ML potential to transform the research paradigm and is adaptable to broader material synthesis.", "date": "2026-03-17", "categories": []}
{"id": "2603.17196", "title": "Self-Conditioned Denoising for Atomistic Representation Learning", "abstract": "The success of large-scale pretraining in NLP and computer vision has catalyzed growing efforts to develop analogous foundation models for the physical sciences. However, pretraining strategies using atomistic data remain underexplored. To date, large-scale supervised pretraining on DFT force-energy labels has provided the strongest performance gains to downstream property prediction, out-performing existing methods of self-supervised learning (SSL) which remain limited to ground-state geometries, and/or single domains of atomistic data. We address these shortcomings with Self-Conditioned Denoising (SCD), a backbone-agnostic reconstruction objective that utilizes self-embeddings for conditional denoising across any domain of atomistic data, including small molecules, proteins, periodic materials, and 'non-equilibrium' geometries. When controlled for backbone architecture and pretraining dataset, SCD significantly outperforms previous SSL methods on downstream benchmarks and matches or exceeds the performance of supervised force-energy pretraining. We show that a small, fast GNN pretrained by SCD can achieve competitive or superior performance to larger models pretrained on significantly larger labeled or unlabeled datasets, across tasks in multiple domains. Our code is available at: https://github.com/TyJPerez/SelfConditionedDenoisingAtoms", "date": "2026-03-17", "categories": []}
{"id": "2306.12272", "title": "From structure mining to unsupervised exploration of atomic octahedral networks", "abstract": "Understanding the spatial arrangements of atom-centered coordination octahedra is crucial for relating structures to properties for many materials families. Traditional case-by-case inspection becomes a prohibitive task for discovering trends and similarities in large datasets. Here, we operationalize chemical intuition to automate the geometric parsing, quantification, and classification of coordination octahedral networks using unsupervised machine learning. We apply the workflow to analyze two datasets to demonstrate its effectiveness. For computationally generated single oxide perovskite (ABO$_{3}$) polymorphs, we uncover axis-dependent tilting trends which assist in detecting oxidation state changes. For hybrid iodoplumbates (A$_x$Pb$_y$I$_z$) from measured structures, we taxonomize their octahedral networks, revealing a Pauling-like connectivity rule for the coordination environment and the design principles underpinning their structural diversity. Our results offer a glimpse into the vast design space of atomic octahedral networks in materials chemistry and inform high-throughput, targeted screening of specific structure types.", "date": "2023-06-21", "categories": []}
{"id": "2603.15712", "title": "LLM-Driven Discovery of High-Entropy Catalysts via Retrieval-Augmented Generation", "abstract": "CO2 reduction requires efficient catalysts, yet materials discovery remains bottlenecked by 10-20 year development cycles requiring deep domain expertise. This paper demonstrates how large language models can assist the catalyst discovery process by helping researchers explore chemical spaces and interpret results when augmented with retrieval-based grounding. We introduce a retrieval-augmented generation framework that enables GPT-4 to navigate chemical space by accessing a database of 50,000+ known materials, adapting general-purpose language understanding for high-throughput materials design. Our approach generated over 250 catalyst candidates with an 82% thermodynamic stability rate while addressing multi-objective constraints: 68% achieved <$100/kg cost with metallic conductivity (band gap<0.1eV) and mechanical stability (B/G>1.75). The best-performing Fe0.2Co0.2Ni0.2Ir0.1Ru0.3 achieves 0.285V limiting potential (25% improvement over IrO2), while Cr0.2Fe0.2Co0.3Ni0.2Mo0.1 optimally balances performance-cost trade-offs at $18/kg. Volcano plot analysis confirms that 78% of LLM-generated catalysts cluster near the theoretical activity optimum, while our system achieves 200x computational efficiency compared to traditional high-throughput screening. By demonstrating that retrieval-augmented generation can ground AI creativity in physical constraints without sacrificing exploration, this work demonstrates an approach where natural language interfaces can streamline materials discovery workflows, enabling researchers to explore chemical spaces more efficiently while the LLM assists in result interpretation and hypothesis generation.", "date": "2026-03-16", "categories": []}
{"id": "2408.14928", "title": "Targeting the partition function of chemically disordered materials with a generative approach based on inverse variational autoencoders", "abstract": "Computing atomic-scale properties of chemically disordered materials requires an efficient exploration of their vast configuration space. Traditional approaches such as Monte Carlo or Special Quasirandom Structures either entail sampling an excessive amount of configurations or do not ensure that the configuration space has been properly covered. In this work, we propose a novel approach where generative machine learning is used to yield a representative set of configurations for accurate property evaluation and provide accurate estimations of atomic-scale properties with minimal computational cost. Our method employs a specific type of variational autoencoder with inverse roles for the encoder and decoder, enabling the application of an unsupervised active learning scheme that does not require any initial training database. The model iteratively generates configuration batches, whose properties are computed with conventional atomic-scale methods. These results are then fed back into the model to estimate the partition function, repeating the process until convergence. We illustrate our approach by computing point-defect formation energies and concentrations in (U, Pu)O2 mixed-oxide fuels. In addition, the ML model provides valuable insights into the physical factors influencing the target property. Our method is generally applicable to explore other properties, such as atomic-scale diffusion coefficients, in ideally or non-ideally disordered materials like high-entropy alloys.", "date": "2024-08-27", "categories": []}
{"id": "2409.12015", "title": "All-in-one foundational models learning across quantum chemical levels", "abstract": "Machine learning (ML) potentials typically target a single quantum chemical (QC) level while the ML models developed for multi-fidelity learning have not been shown to provide scalable solutions for foundational models. Here we introduce the all-in-one (AIO) ANI model architecture based on multimodal learning which can learn an arbitrary number of QC levels. Our all-in-one learning approach offers a more general and easier-to-use alternative to transfer learning. We use it to train the AIO-ANI-UIP foundational model with the generalization capability comparable to semi-empirical GFN2-xTB and DFT with a double-zeta basis set for organic molecules. We show that the AIO-ANI model can learn across different QC levels ranging from semi-empirical to density functional theory to coupled cluster. We also use AIO models to design the foundational model {\\Delta}-AIO-ANI based on {\\Delta}-learning with increased accuracy and robustness compared to AIO-ANI-UIP. The code and the foundational models are available at https://github.com/dralgroup/aio-ani; they will be integrated into the universal and updatable AI-enhanced QM (UAIQM) library and made available in the MLatom package so that they can be used online at the XACS cloud computing platform (see https://github.com/dralgroup/mlatom for updates).", "date": "2024-09-18", "categories": []}
{"id": "2504.18854", "title": "Predicting Stress in Two-phase Random Materials and Super-Resolution Method for Stress Images by Embedding Physical Information", "abstract": "Stress analysis is an important part of material design. For materials with complex microstructures, such as two-phase random materials (TRMs), material failure is often accompanied by stress concentration. Phase interfaces in two-phase materials are critical for stress concentration. Therefore, the prediction error of stress at phase boundaries is crucial. In practical engineering, the pixels of the obtained material microstructure images are limited, which limits the resolution of stress images generated by deep learning methods, making it difficult to observe stress concentration regions. Existing Image Super-Resolution (ISR) technologies are all based on data-driven supervised learning. However, stress images have natural physical constraints, which provide new ideas for new ISR technologies. In this study, we constructed a stress prediction framework for TRMs. First, the framework uses a proposed Multiple Compositions U-net (MC U-net) to predict stress in low-resolution material microstructures. By considering the phase interface information of the microstructure, the MC U-net effectively reduces the problem of excessive prediction errors at phase boundaries. Secondly, a Mixed Physics-Informed Neural Network (MPINN) based method for stress ISR (SRPINN) was proposed. By introducing the constraints of physical information, the new method does not require paired stress images for training and can increase the resolution of stress images to any multiple. This enables a multiscale analysis of the stress concentration regions at phase boundaries. Finally, we performed stress analysis on TRMs with different phase volume fractions and loading states through transfer learning. The results show the proposed stress prediction framework has satisfactory accuracy and generalization ability.", "date": "2025-04-26", "categories": []}
{"id": "2512.15746", "title": "A Unified Generative-Predictive Framework for Deterministic Inverse Design", "abstract": "Inverse design of heterogeneous material microstructures is a fundamentally ill-posed and famously computationally expensive problem. This is exacerbated by the high-dimensional design spaces associated with finely resolved images, multimodal input property streams, and a highly nonlinear forward physics. Whilst modern generative models excel at accurately modeling such complex forward behavior, most of them are not intrinsically structured to support fast, stable \\emph{deterministic} inversion with a physics-informed bias. This work introduces Janus, a unified generative-predictive framework to address this problem. Janus couples a deep encoder-decoder architecture with a predictive KHRONOS head, a separable neural architecture. Topologically speaking, Janus learns a latent manifold simultaneously isometric for generative inversion and pruned for physical prediction; the joint objective inducing \\emph{disentanglement} of the latent space. Janus is first validated on the MNIST dataset, demonstrating high-fidelity reconstruction, accurate classification and diverse generative inversion of all ten target classes. It is then applied to the inverse design of heterogeneous microstructures labeled with thermal conductivity. It achieves a forward prediction accuracy $R^2=0.98$ (2\\% relative error) and sub-5\\% pixelwise reconstruction error. Inverse solutions satisfy target properties to within $1\\%$ relative error. Inverting a sweep through properties reveal smooth traversal of the latent manifold, and UMAP visualization confirms the emergence of a low-dimensional, disentangled manifold. By unifying prediction and generation within a single latent space, Janus enables real-time, physics-informed inverse microstructure generation at a lower computational cost typically associated with classical optimization-based approaches.", "date": "2025-12-10", "categories": []}
{"id": "2512.17245", "title": "Enhancing Reconstruction Capability of Wavelet Transform Amorphous Radial Distribution Function via Machine Learning Assisted Parameter Tuning", "abstract": "Understanding atomic structures is crucial, yet amorphous materials remain challenging due to their irregular and non-periodic nature. The Wavelet Transform Radial Distribution Function (WT-RDF) offers a physics-based framework for analyzing amorphous structures, reliably reconstructing the first and second Radial Distribution Function (RDF) peaks and overall curve trends in both binary (Ge 0.25 Se 0.75) and ternary Ag x(Ge 0.25 Se 0.75)100-x (x = 5, 10, 15, 20, 25) systems. Despite these strengths, WT-RDF shows limitations in amplitude accuracy, which affects quantitative analyses such as coordination numbers. The shortcoming arises from improper parameter (a, b, Kf, C, and {\\Lambda})) selection, as the parameters intrinsically represent atomic interactions within amorphous materials. This study addresses the issue by optimizing WT-RDF parameters using a machine learning approach via learnable parameter optimization, parameter bounding, and selective loss, producing the enhanced WT-RDF+ framework. WT-RDF+ improves the precision of peak reconstructions and outperforms benchmark Machine Learning (ML) models, including Radial Basis Function (RBF) and Long Short-term Memory (LSTM), when trained on only 25% of the binary dataset. Specifically, the machine learning benchmarks are defined as regressors with radial distance r input and G(r) output taken from Ab Initio Molecular Dynamics (AIMD) RDF simulation, not the reduced structure factor SR(q) to G(r) inversions. These results demonstrate that WT-RDF+ is a robust and reliable model for RDF reconstruction of Ge-Se and Ag-Ge-Se family.", "date": "2025-12-19", "categories": []}
{"id": "2512.23930", "title": "Assessing generative modeling approaches for free energy estimates in condensed matter", "abstract": "The accurate estimation of free energy differences between two states is a long-standing challenge in molecular simulations. Traditional approaches generally rely on sampling multiple intermediate states to ensure sufficient overlap in phase space and are, consequently, computationally expensive. Boltzmann Generators and related generative-model-based methods have recently addressed this challenge by learning a direct probability density transform between two states. However, it remains unclear which approach provides the best trade-off between efficiency, accuracy, and scalability. In this work, we review and benchmark selected generative approaches for condensed-matter systems, including discrete and continuous normalizing flows for targeted free energy perturbation and FEAT (Free Energy Estimators with Adaptive Transport) combined with the escorted Jarzynski equality, using coarse-grained monatomic ice and Lennard-Jones solids as benchmark systems. All models yield highly accurate free energy estimates and, depending on the system, may require fewer energy evaluations than traditional methods. Continuous flows and FEAT are most efficient in energy evaluations, whereas discrete flows have substantially lower inference cost. By releasing all data together with our results, we enable future benchmarking of free energy estimation methods in condensed-phase systems.", "date": "2025-12-30", "categories": []}
{"id": "2601.00581", "title": "AceFF: A State-of-the-Art Machine Learning Potential for Small Molecules", "abstract": "We introduce AceFF, a pre-trained machine learning interatomic potential (MLIP) optimized for small molecule drug discovery. While MLIPs have emerged as efficient alternatives to Density Functional Theory (DFT), generalizability across diverse chemical spaces remains difficult. AceFF addresses this via a refined TensorNet2 architecture trained on a comprehensive dataset of drug-like compounds. This approach yields a force field that balances high-throughput inference speed with DFT-level accuracy. \\mbox{AceFF} fully supports the essential medicinal chemistry elements (H, B, C, N, O, F, Si, P, S, Cl, Br, I) and is explicitly trained to handle charged states. Validation against rigorous benchmarks, including complex torsional energy scans, molecular dynamics trajectories, batched minimizations, and tests of force and energy accuracy, demonstrates that AceFF is state-of-the-art for organic molecules in the accuracy and speed regime important for drug discovery. The AceFF-2 model weights and inference code are available at https://huggingface.co/Acellera/AceFF-2.0.", "date": "2026-01-02", "categories": []}
{"id": "2603.13303", "title": "PolyMon: A Unified Framework for Polymer Property Prediction", "abstract": "Accurate prediction of polymer properties is essential for materials design, but remains challenging due to data scarcity, diverse polymer representations, and the lack of systematic evaluation across modelling choices. Here, we present PolyMon, a unified and accessible framework that integrates multiple polymer representations, machine learning methods, and training strategies within a single, accessible platform. PolyMon supports various descriptors and graph construction strategies for polymer representations, and includes a wide range of models, from tabular models to graph neural networks, along with flexible training strategies including multi-fidelity learning, {\\Delta}-learning, active learning, and ensemble learning. Using five key polymer properties as benchmarks, we perform systematic evaluations to assess how representations and models affect predictive performance. These case studies further illustrate how different training strategies can be applied within a consistent workflow to leverage limited data and incorporate physical model derived information. Overall, PolyMon provides a comprehensive and extensible foundation for benchmarking and advancing machine learning-based polymer property prediction. The code is available at github.com/fate1997/polymon.", "date": "2026-03-02", "categories": []}
{"id": "2603.13445", "title": "Diffusion-based Generative Machine Learning Model for Predicting Crack Propagation in Aluminum Nitride at the Atomic Scale", "abstract": "Predicting atomic-scale crack propagation in aluminum nitride (AlN) is critical for semiconductor reliability but remains prohibitively expensive via molecular dynamics (MD). We develop a diffusion-based generative machine learning model to predict atomic-scale crack propagation in AlN, a critical semiconductor material, by conditioning solely on initial microstructure embeddings. Trained on MD simulations of single-crack systems, the model achieves a significant speedup while accurately forecasting dynamic fracture processes, including stress-driven crack initiation, crack branching, and atomic-scale bridging ligaments. Crucially, it demonstrates inherent physical fidelity by reproducing material-intrinsic mechanisms while disregarding periodic boundary artifacts, and generalizes to unseen multi-crack configurations. Validation against MD ground truth confirms the capability of the model to capture complex fracture physics without auxiliary stress or energy data, enabling rapid exploration of crack-mediated failure for semiconductor reliability optimization.", "date": "2026-03-13", "categories": []}
{"id": "2603.13744", "title": "Research Paradigm of Materials Science Tetrahedra with Artificial Intelligence", "abstract": "The classical material tetrahedron that represents the Structure-Property-Processing-Performance-Characterization relationship is the most important research paradigm in materials science so far. It has served as a protocol to guide experiments, modeling, and theory to uncover hidden relationships between various aspects of a certain material. This substantially facilitates knowledge accumulation and material discovery with desired functionalities to realize versatile applications. In recent years, with the advent of artificial intelligence (AI) techniques, the attention of AI towards scientific research is soaring. The trials of implementing AI in various disciplines are endless, with great potential to revolutionize the research diagram. Despite the success in natural language processing and computer vision, how to effectively integrate AI with natural science is still a grand challenge, bearing in mind their fundamental differences. Inspired by these observations and limitations, we delve into the current research paradigm dictated by the classical material tetrahedron and propose two new paradigms to stimulate data-driven and AI-augmented research. One tetrahedron focuses on AI for materials science by considering the Matter-Data-Model-Potential-Agent diagram. The other demonstrates AI research by discussing Data-Architecture-Encoding-Optimization-Inference relationships. The crucial ingredients of these frameworks and their connections are discussed, which will likely motivate both scientific thinking refinement and technology advancement. Despite the widespread enthusiasm for chasing AI for science, we must analyze issues rationally to come up with well-defined, resolvable scientific problems in order to better master the power of AI.", "date": "2026-03-14", "categories": []}
{"id": "2603.13920", "title": "Generative Inverse Design of Cold Metals for Low-Power Electronics", "abstract": "Cold metals are a class of metals with an intrinsic energy gap located close to the Fermi level, which enables cold-carrier injection for steep-slope transistors and is therefore promising for low-power electronic applications. High-throughput screening has revealed 252 three-dimensional (3D) cold metals in the Materials Project database, but database searches are inherently limited to known compounds. Here we present an inverse-design workflow that generates 3D cold metals using MatterGPT, a conditional autoregressive Transformer trained on SLICES, an invertible and symmetry-invariant crystal string representation. We curate a training set of 26,309 metallic structures labeled with energy above hull and a unified band-edge distance descriptor that merges p-type and n-type cold-metal characteristics to address severe label imbalance. Property-conditioned generation targeting thermodynamic stability and 50-500 meV band-edge distances produces 148,506 unique candidates; 92.1% are successfully reconstructed to 3D structures and down-selected by symmetry, uniqueness and novelty filters, followed by high-throughput DFT validation. We identify 257 cold metals verified as novel with respect to the Materials Project database, with gaps around the Fermi level spanning 50-500 meV. First-principles phonon, electronic-structure, and work-function calculations for representative candidates confirm dynamical stability and contact-relevant work functions. Our results demonstrate that SLICES-enabled generative transformers can expand the chemical space of cold metals beyond high-throughput screening, providing a route to low-power electronic materials discovery.", "date": "2026-03-14", "categories": []}
{"id": "2603.14489", "title": "Predicting Stress-strain Behaviors of Additively Manufactured Materials via Loss-based and Activation-based Physics-informed Machine Learning", "abstract": "Predicting the stress-strain behaviors of additively manufactured materials is crucial for part qualification in additive manufacturing (AM). Conventional physics-based constitutive models often oversimplify material properties, while data-driven machine learning (ML) models often lack physical consistency and interpretability. To address these issues, we propose a physics-informed machine learning (PIML) framework to improve the predictive performance and physical consistency for predicting the stress-strain curves of additively manufactured polymers and metals. A polynomial regression model is used to predict the yield point from AM process parameters, then stress-strain curves are segmented into elastic and plastic regions. Two long short-term memory (LSTM) models are trained to predict two regions separately. For the elastic region, Hooke's law is embedded into the LSTM model for both polymer and metal. For the plastic region, Voce hardening law and Hollomon's law are embedded into the LSTM model for polymer and metal, respectively. The loss-based and activation-based PIML architectures are developed by embedding the physical laws into the loss and activation functions, respectively. The performance of the two PIML architectures are compared with two LSTM-based ML models, three additional ML models, and a physics-based constitutive model. These models are built on experimental data collected from two additively manufactured polymers (i.e., Nylon and carbon fiber-acrylonitrile butadiene styrene) and two additively manufactured metals (i.e., AlSi10Mg and Ti6Al4V). Experimental results demonstrate that two PIML architectures consistently outperform the other models. The segmental predictive model with activation-based PIML architecture achieves the lowest MAPE of 10.46+/-0.81% and the highest R^2 of 0.82+/-0.05 arocss four datasets.", "date": "2026-03-15", "categories": []}
{"id": "2603.14695", "title": "Scaling Autoregressive Models for Lattice Thermodynamics", "abstract": "Predicting how materials behave under realistic conditions requires understanding the statistical distribution of atomic configurations on crystal lattices, a problem central to alloy design, catalysis, and the study of phase transitions. Traditional Markov-chain Monte Carlo sampling suffers from slow convergence and critical slowing down near phase transitions, motivating the use of generative models that directly learn the thermodynamic distribution. Existing autoregressive models (ARMs), however, generate configurations in a fixed sequential order and incur high memory and training costs, limiting their applicability to realistic systems. Here, we develop a framework combining any-order ARMs, which generate configurations flexibly by conditioning on any known subset of lattice sites, with marginalization models (MAMs), which approximate the probability of any partial configuration in a single forward pass and substantially reduce memory requirements. This combination enables models trained on smaller lattices to be reused for sampling larger systems, while supporting expressive Transformer architectures with lattice-aware positional encodings at manageable computational cost. We demonstrate that Transformer-based any-order MAMs achieve more accurate free energies than multilayer perceptron-based ARMs on both the two-dimensional Ising model and CuAu alloys, faithfully capturing phase transitions and critical behavior. Overall, our framework scales from $10 \\times 10$ to $20 \\times 20$ Ising systems and from $2 \\times 2 \\times 4$ to $4 \\times 4 \\times 8$ CuAu supercells at reduced computational cost compared to conventional sampling methods.", "date": "2026-03-16", "categories": []}
{"id": "2603.15582", "title": "Benchmarking Machine Learning Approaches for Polarization Mapping in Ferroelectrics Using 4D-STEM", "abstract": "Four-dimensional scanning transmission electron microscopy (4D-STEM) provides rich, atomic-scale insights into materials structures. However, extracting specific physical properties - such as polarization directions essential for understanding functional properties of ferroelectrics - remains a significant challenge. In this study, we systematically benchmark multiple machine learning models, namely ResNet, VGG, a custom convolutional neural network, and PCA-informed k-Nearest Neighbors, to automate the detection of polarization directions from 4D-STEM diffraction patterns in ferroelectric potassium sodium niobate. While models trained on synthetic data achieve high accuracy on idealized synthetic diffraction patterns of equivalent thickness, the domain gap between simulation and experiment remains a critical barrier to real-world deployment. In this context, a custom made prototype representation training regime and PCA-based methods, combined with data augmentation and filtering, can better bridge this gap. Error analysis reveals periodic missclassification patterns, indicating that not all diffraction patterns carry enough information for a successful classification. Additionally, our qualitative analysis demonstrates that irregularities in the model's prediction patterns correlate with defects in the crystal structure, suggesting that supervised models could be used for detecting structural defects. These findings guide the development of robust, transferable machine learning tools for electron microscopy analysis.", "date": "2026-03-16", "categories": []}
{"id": "2409.01523", "title": "Machine learning approach for vibronically renormalized electronic band structures", "abstract": "We present a machine learning (ML) method for efficient computation of vibrational thermal expectation values of physical properties from first principles. Our approach is based on the non-perturbative frozen phonon formulation in which stochastic Monte Carlo algorithm is employed to sample configurations of nuclei in a supercell at finite temperatures based on a first-principles phonon model. A deep-learning neural network is trained to accurately predict physical properties associated with sampled phonon configurations, thus bypassing the time-consuming {\\em ab initio} calculations. To incorporate the point-group symmetry of the electronic system into the ML model, group-theoretical methods are used to develop a symmetry-invariant descriptor for phonon configurations in the supercell. We apply our ML approach to compute the temperature dependent electronic energy gap of silicon based on density functional theory (DFT). We show that, with less than a hundred DFT calculations for training the neural network model, an order of magnitude larger number of sampling can be achieved for the computation of the vibrational thermal expectation values. Our work highlights the promising potential of ML techniques for finite temperature first-principles electronic structure methods.", "date": "2024-09-03", "categories": []}
{"id": "2509.20276", "title": "Extended Low-Rank Approximation Accelerates Learning of Elastic Response in Heterogeneous Materials", "abstract": "Predicting how the microstructure governs the mechanical response of heterogeneous materials is essential for optimizing design and performance. Yet this task remains difficult due to the complex, high dimensional nature of microstructural features. Relying on physics based simulations to probe the microstructural space is computationally prohibitive. This motivates the development of computational tools to efficiently learn structure property linkages governing mechanical behavior. While contemporary data driven approaches offer new possibilities, they often require large datasets. To address this challenge, this work presents the Extended Low Rank Approximation (xLRA), a framework that employs canonical polyadic tensor decomposition. It efficiently maps high dimensional microstructural information to the local elastic response by adaptively incorporating higher rank terms. xLRA accurately predicts the local elastic strain fields in porous microstructures, requiring a maximum rank of only 4. The compact formulation of xLRA achieves accurate predictions when trained on just 5% of the dataset, demonstrating significant data efficiency. Moreover, xLRA proves transferability by delivering results across representative material systems, including two phase composites and single and dual phase polycrystals. Despite being compact, xLRA retains essential microstructural details, enabling accurate predictions on unseen microstructures. Benchmarking shows that xLRA outperforms contemporary methods in predictive accuracy, generalizability, and computational efficiency, while requiring 6 orders of magnitude fewer floating point operations. In summary, xLRA provides an efficient framework for predicting the elastic response from microstructures, enabling scalable mapping of structure property linkages.", "date": "2025-09-24", "categories": []}
{"id": "2603.12183", "title": "Proof-Carrying Materials: Falsifiable Safety Certificates for Machine-Learned Interatomic Potentials", "abstract": "Machine-learned interatomic potentials (MLIPs) are deployed for high-throughput materials screening without formal reliability guarantees. We show that a single MLIP used as a stability filter misses 93% of density functional theory (DFT)-stable materials (recall 0.07) on a 25,000-material benchmark. Proof-Carrying Materials (PCM) closes this gap through three stages: adversarial falsification across compositional space, bootstrap envelope refinement with 95% confidence intervals, and Lean 4 formal certification. Auditing CHGNet, TensorNet and MACE reveals architecture-specific blind spots with near-zero pairwise error correlations (r <= 0.13; n = 5,000), confirmed by independent Quantum ESPRESSO validation (20/20 converged; median DFT/CHGNet force ratio 12x). A risk model trained on PCM-discovered features predicts failures on unseen materials (AUC-ROC = 0.938 +/- 0.004) and transfers across architectures (cross-MLIP AUC-ROC ~ 0.70; feature importance r = 0.877). In a thermoelectric screening case study, PCM-audited protocols discover 62 additional stable materials missed by single-MLIP screening - a 25% improvement in discovery yield.", "date": "2026-03-12", "categories": []}
{"id": "2603.07977", "title": "Scaling Machine Learning Interatomic Potentials with Mixtures of Experts", "abstract": "Machine Learning Interatomic Potentials (MLIPs) enable accurate large-scale atomistic simulations, yet improving their expressive capacity efficiently remains challenging. Here we systematically develop Mixture-of-Experts (MoE) and Mixture-of-Linear-Experts (MoLE) architectures for MLIPs and analyze the effects of routing strategies and expert designs. We show that sparse activation combined with shared experts yields substantial performance gains, and that nonlinear MoE formulations outperform MoLE when shared experts are present, underscoring the importance of nonlinear expert specialization. Furthermore, element-wise routing consistently surpasses configuration-level routing, while global MoE routing often leads to numerical instability. The resulting element-wise MoE model achieves state-of-the-art accuracy across the OMol25, OMat24, and OC20M benchmarks. Analysis of routing patterns reveals chemically interpretable expert specialization aligned with periodic-table trends, indicating that the model effectively captures element-specific chemical characteristics for precise interatomic modeling.", "date": "2026-03-09", "categories": []}
{"id": "2603.11068", "title": "From Phase Prediction to Phase Design: A ReAct Agent Framework for High-Entropy Alloy Discovery", "abstract": "Discovering high-entropy alloy (HEA) compositions that reliably form a target crystal phase is a high-dimensional inverse design problem that conventional trial-and-error experimentation and forward-only machine learning models cannot efficiently solve. Here we present a ReAct (Reasoning + Acting) LLM agent that autonomously proposes, validates, and iteratively refines HEA compositions by querying a calibrated XGBoost surrogate trained on 4,753 experimental records across four phases (FCC, BCC, BCC+FCC, BCC+IM), achieving 94.66\\% accuracy (F1 macro = 0.896). Against Bayesian optimisation (BO) and random search baselines, the full-prompt agent achieves descriptor-space rediscovery rates of 38\\%, 18\\%, and 38\\% for FCC, BCC, and BCC+FCC (Mann--Whitney $p \\leq 0.039$), with proposals lying 2.4--22.8$\\times$ closer to the experimental phase manifold than random search. An ablation reveals that domain priors shift the agent from landmark-alloy recall toward compositionally diverse exploration -- an uninformed agent scores higher rediscovery by concentrating on literature-dense families, while the full-prompt agent explores underrepresented space (unique ratio 1.0 vs.\\ 0.39 for BCC+FCC). These regimes represent distinct criteria: proximity to known literature versus genuine discovery. Spearman analysis confirms agent reasoning is statistically aligned with empirical phase distributions ($\\rho = 0.736$, $p = 0.004$ for BCC). This work establishes LLM-guided agentic reasoning as a principled, transparent, and manifold-aware complement to gradient-free optimisation for inverse alloy design.", "date": "2026-03-10", "categories": []}
{"id": "2603.11695", "title": "PolyCrysDiff: Controllable Generation of Three-Dimensional Computable Polycrystalline Material Structures", "abstract": "The three-dimensional (3D) microstructures of polycrystalline materials exert a critical influence on their mechanical and physical properties. Realistic, controllable construction of these microstructures is a key step toward elucidating structure-property relationships, yet remains a formidable challenge. Herein, we propose PolyCrysDiff, a framework based on conditional latent diffusion that enables the end-to-end generation of computable 3D polycrystalline microstructures. Comprehensive qualitative and quantitative evaluations demonstrate that PolyCrysDiff faithfully reproduces target grain morphologies, orientation distributions, and 3D spatial correlations, while achieving an $R^2$ over 0.972 on grain attributes (e.g., size and sphericity) control, thereby outperforming mainstream approaches such as Markov random field (MRF)- and convolutional neural network (CNN)-based methods. The computability and physical validity of the generated microstructures are verified through a series of crystal plasticity finite element method (CPFEM) simulations. Leveraging PolyCrysDiff's controllable generative capability, we systematically elucidate how grain-level microstructural characteristics affect the mechanical properties of polycrystalline materials. This development is expected to pave a key step toward accelerated, data-driven optimization and design of polycrystalline materials.", "date": "2026-03-12", "categories": []}
{"id": "2603.10205", "title": "Flexible Cutoff Learning: Optimizing Machine Learning Potentials After Training", "abstract": "We introduce Flexible Cutoff Learning (FCL), a method for training machine learning interatomic potentials (MLIPs) whose cutoff radii can be adjusted after training. Unlike conventional MLIPs that fix the cutoff radius during training, FCL models are trained by randomly sampling cutoff radii independently for each atom. The resulting model can then be deployed with different per-atom cutoff radii depending on the application, enabling application-specific optimization of the accuracy-cost tradeoff. Using a differentiable cost model, these per-atom cutoffs can be optimized for specific target systems after training. We demonstrate FCL with a modified MACE architecture trained on the MAD dataset. For a subset featuring molecular crystals, optimized per-atom cutoffs reduce computational cost by more than 60% while increasing force errors by less than 1%. These results show that FCL enables training of a single general-purpose MLIP that can be adapted to diverse applications through post-training cutoff optimization, eliminating the need for retraining.", "date": "2026-03-10", "categories": []}
{"id": "2503.08104", "title": "Functional Unit: A New Perspective on Materials Science Research Paradigms", "abstract": "New materials have long marked the civilization level, serving as an impetus for technological progress and societal transformation. The classic structure-property correlations were key of materials science and engineering. However, the knowledge of materials faces significant challenges in adapting to exclusively data-driven approaches for new material discovery. This perspective introduces the concepts of functional units (FUs) to fill the gap in understanding of material structure-property correlations and knowledge inheritance as the \"composition-microstructure\" paradigm transitions to a data-driven AI paradigm transitions. Firstly, we provide a bird's-eye view of the research paradigm evolution from early \"process-structure-properties-performance\" to contemporary data-driven AI new trend. Next, we highlight recent advancements in the characterization of functional units across diverse material systems, emphasizing their critical role in multiscale material design. Finally, we discuss the integration of functional units into the new AI-driven paradigm of materials science, addressing both opportunities and challenges in computational materials innovation.", "date": "2025-03-11", "categories": []}
{"id": "2602.16020", "title": "MolCrystalFlow: Molecular Crystal Structure Prediction via Flow Matching", "abstract": "Molecular crystal structure prediction represents a grand challenge in computational chemistry due to large sizes of constituent molecules and complex intra- and intermolecular interactions. While generative modeling has revolutionized structure discovery for molecules, inorganic solids, and metal-organic frameworks, extending such approaches to fully periodic molecular crystals is still elusive. Here, we present MolCrystalFlow, a flow-based generative model for molecular crystal structure prediction. The framework disentangles intramolecular complexity from intermolecular packing by embedding molecules as rigid bodies and jointly learning the lattice matrix, molecular orientations, and centroid positions. Centroids and orientations are represented on their native Riemannian manifolds, allowing geodesic flow construction and graph neural network operations that respects geometric symmetries. We benchmark our model against a state-of-the-art generative model (MOFFlow) for large-size periodic crystals and a rule-based structure generation method (Genarris) on two open-source molecular crystal datasets. MolCrystalFlow outperforms MOFFlow while achieving competitive performance against Genarris. We also demonstrate an integration of MolCrystalFlow model with universal machine learning potential to accelerate molecular crystal structure prediction, paving the way for data-driven generative discovery of molecular crystals.", "date": "2026-02-17", "categories": []}
{"id": "2603.08881", "title": "From Word2Vec to Transformers: Text-Derived Composition Embeddings for Filtering Combinatorial Electrocatalysts", "abstract": "Compositionally complex solid solution electrocatalysts span vast composition spaces, and even one materials system can contain more candidate compositions than can be measured exhaustively. Here we evaluate a label-free screening strategy that represents each composition using embeddings derived from scientific texts and prioritizes candidates based on similarity to two property concepts. We compare a corpus-trained Word2Vec baseline with transformer-based embeddings, where compositions are encoded either by linear element-wise mixing or by short composition prompts. Similarities to `concept directions', the terms conductivity and dielectric, define a 2-dimensional descriptor space, and a symmetric Pareto-front selection is used to filter candidate subsets without using electrochemical labels. Performance is assessed on 15 materials libraries including noble metal alloys and multicomponent oxides. In this setting, the lightweight Word2Vec baseline, which uses a simple linear combination of element embeddings, often achieves the highest number of reductions of possible candidate compositions while staying close to the best measured performance.", "date": "2026-03-09", "categories": []}
{"id": "2603.09693", "title": "Physics-informed neural operator for predictive parametric phase-field modelling", "abstract": "Predicting the microstructural and morphological evolution of materials through phase-field modelling is computationally intensive, particularly for high-throughput parametric studies. While neural operators such as the Fourier neural operator (FNO) show promise in accelerating the solution of parametric partial differential equations (PDEs), the lack of explicit physical constraints, may limit generalisation and long-term accuracy for complex phase-field dynamics. Here, we develop a physics-informed neural operator framework to learn parametric phase-field PDEs, namely PF-PINO. By embedding the residuals of phase-field governing equations into the data-fidelity loss function, our framework effectively enforces physical constraints during training. We validate PF-PINO against benchmark phase-field problems, including electrochemical corrosion, dendritic crystal solidification, and spinodal decomposition. Our results demonstrate that PF-PINO significantly outperforms conventional FNO in accuracy, generalisation capability, and long-term stability. This work provides a robust and efficient computational tool for phase-field modelling and highlights the potential of physics-informed neural operators to advance scientific machine learning for complex interfacial evolution problems.", "date": "2026-03-10", "categories": []}
{"id": "2509.01370", "title": "CbLDM: A Diffusion Model for recovering nanostructure from atomic pair distribution function", "abstract": "The nanostructure inverse problem is an attractive problem that helps researchers to understand the relationship between the properties and the structure of nanomaterials. This study focuses on the problem of recovering the model system of monometallic nanoparticles (MMNPs) from their pair distribution function (PDF) and regards it as a highly ill-posed conditional generation task. This study proposes a Condition-based Latent Diffusion Model (CbLDM) as a feasible solution to this problem. This model demonstrates an acceleration approach within the framework of a latent diffusion model by using conditional priors to estimate the conditional posterior distribution, which is an approximate distribution of p(z|x). In addition, this study uses Laplacian matrix instead of distance matrix to recover the nanostructure, which helps to improve stability. Our study demonstrates that a latent diffusion model with a conditional prior can generate nanostructures that are consistent with PDF observations and physically meaningful, thereby laying the groundwork for subsequent more complex inverse problems.", "date": "2025-09-01", "categories": []}
{"id": "2603.06939", "title": "Physics-Consistent Neural Networks for Learning Deformation and Director Fields in Microstructured Media with Loss-Based Validation Criteria", "abstract": "In this work, we study the mechanical behavior of solids with microstructure using the framework of Cosserat elasticity with a single unit director. This formulation captures the coupling between deformation and orientational fields that arises in many structured materials. To compute equilibrium configurations of such media, we develop two complementary computational approaches: a finite element formulation based on variational principles and a neural network-based solver that directly minimizes the total potential energy. The neural architecture is constructed to respect the fundamental kinematic structure of the theory. In particular, it enforces frame invariance of the energy, satisfies the unit-length constraint on the director field, and represents deformation and director fields through separate networks to preserve their kinematic independence in the variational setting. Beyond satisfying balance laws, however, physically admissible solutions must also correspond to stable energy minimizers. To assess this requirement, we derive the quasiconvexity condition, rank-one convexity condition, and the Legendre-Hadamard inequalities for the Cosserat model and formulate them in a manner suitable for evaluating neural network predictions. These necessary stability conditions provide a physics-based validation framework: network outputs that violate these necessary conditions cannot correspond to stable energy minimizers and can therefore be rejected. In this way, we integrate classical variational stability theory with modern machine-learning solvers, establishing a computational workflow in which equilibrium solutions are not only learned but also assessed for energetic consistency.", "date": "2026-03-06", "categories": []}
{"id": "2509.14961", "title": "Spectral/Spatial Tensor Atomic Cluster Expansion with Universal Embeddings in Cartesian Space", "abstract": "Equivariant atomistic machine learning models have largely been built on spherical-tensor representations, where explicit angular-momentum coupling introduces substantial complexity and systematic extensions beyond energies and forces remain challenging, often requires problem-specific architectural choices. Here we introduce the Tensor Atomic Cluster Expansion (TACE), which unifies scalar and tensorial modeling in Cartesian and space by decomposing local environments into irreducible Cartesian tensors (ICT) constructing a controlled many-body hierarchy with atomic cluster expansion (ACE). In addition to performing ACE in the frequency domain, we propose an efficient Clebsch-Gordan-free alternative in the spatial domain. TACE provides universal invariant (e.g., fidelity tags and charges) and equivariant (e.g., external electric fields and non-collinear magnetic moments) embeddings and predicted tensorial observables are handled on equal footing and enabling explicit control at inference. We demonstrate the accuracy, stability, and efficiency across finite molecules and extended materials, including in-domain and out-of-domain benchmarks, spectra, Hessian, external-field responses, charged systems, and multi-fidelity/head training. We further show its robustness on nonequilibrium/reactive datasets and controlled scaling when extending to large foundation model datasets.", "date": "2025-09-18", "categories": []}
{"id": "2603.02002", "title": "MatRIS: Toward Reliable and Efficient Pretrained Machine Learning Interatomic Potentials", "abstract": "Foundation MLIPs demonstrate broad applicability across diverse material systems and have emerged as a powerful and transformative paradigm in chemical and computational materials science. Equivariant MLIPs achieve state-of-the-art accuracy in a wide range of benchmarks by incorporating equivariant inductive bias. However, the reliance on tensor products and high-degree representations makes them computationally costly. This raises a fundamental question: as quantum mechanical-based datasets continue to expand, can we develop a more compact model to thoroughly exploit high-dimensional atomic interactions? In this work, we present MatRIS (\\textbf{Mat}erials \\textbf{R}epresentation and \\textbf{I}nteraction \\textbf{S}imulation), an invariant MLIP that introduces attention-based modeling of three-body interactions. MatRIS leverages a novel separable attention mechanism with linear complexity $O(N)$, enabling both scalability and expressiveness. MatRIS delivers accuracy comparable to that of leading equivariant models on a wide range of popular benchmarks (Matbench-Discovery, MatPES, MDR phonon, Molecular dataset, etc). Taking Matbench-Discovery as an example, MatRIS achieves an F1 score of up to 0.847 and attains comparable accuracy at a lower training cost. The work indicates that our carefully designed invariant models can match or exceed the accuracy of equivariant models at a fraction of the cost, shedding light on the development of accurate and efficient MLIPs.", "date": "2026-03-02", "categories": []}
{"id": "2510.22503", "title": "LLEMA: Evolutionary Search with LLMs for Multi-Objective Materials Discovery", "abstract": "Materials discovery requires navigating vast chemical and structural spaces while satisfying multiple, often conflicting, objectives. We present LLM-guided Evolution for MAterials discovery (LLEMA), a unified framework that couples the scientific knowledge embedded in large language models with chemistry-informed evolutionary rules and memory-based refinement. At each iteration, an LLM proposes crystallographically specified candidates under explicit property constraints; a surrogate-augmented oracle estimates physicochemical properties; and a multi-objective scorer updates success/failure memories to guide subsequent generations. Evaluated on 14 realistic tasks that span electronics, energy, coatings, optics, and aerospace, LLEMA discovers candidates that are chemically plausible, thermodynamically stable, and property-aligned, achieving higher hit rates and improved Pareto front quality relative to generative and LLM-only baselines. Ablation studies confirm the importance of rule-guided generation, memory-based refinement, and surrogate prediction. By enforcing synthesizability and multi-objective trade-offs, LLEMA provides a principled approach to accelerating practical materials discovery. Project website: https://scientific-discovery.github.io/llema-project/", "date": "2025-10-26", "categories": []}
{"id": "2511.19500", "title": "CycleChemist: A Dual-Pronged Machine Learning Framework for Organic Photovoltaic Discovery", "abstract": "Organic photovoltaic (OPV) materials offer a promising path toward sustainable energy generation, but their development is limited by the difficulty of identifying high performance donor and acceptor pairs with strong power conversion efficiencies (PCEs). Existing design strategies typically focus on either the donor or the acceptor alone, rather than using a unified approach capable of modeling both components. In this work, we introduce a dual machine learning framework for OPV discovery that combines predictive modeling with generative molecular design. We present the Organic Photovoltaic Donor Acceptor Dataset (OPV2D), the largest curated dataset of its kind, containing 2000 experimentally characterized donor acceptor pairs. Using this dataset, we develop the Organic Photovoltaic Classifier (OPVC) to predict whether a material exhibits OPV behavior, and a hierarchical graph neural network that incorporates multi task learning and donor acceptor interaction modeling. This framework includes the Molecular Orbital Energy Estimator (MOE2) for predicting HOMO and LUMO energy levels, and the Photovoltaic Performance Predictor (P3) for estimating PCE. In addition, we introduce the Material Generative Pretrained Transformer (MatGPT) to produce synthetically accessible organic semiconductors, guided by a reinforcement learning strategy with three objective policy optimization. By linking molecular representation learning with performance prediction, our framework advances data driven discovery of high performance OPV materials.", "date": "2025-11-23", "categories": []}
{"id": "2603.05188", "title": "Escaping the Hydrolysis Trap: An Agentic Workflow for Inverse Design of Durable Photocatalytic Covalent Organic Frameworks", "abstract": "Covalent organic frameworks (COFs) are promising photocatalysts for solar hydrogen production, yet the most electronically favorable linkages, imines, hydrolyze rapidly in water, creating a stability--activity trade-off that limits practical deployment. Navigating the combinatorial design space of nodes, linkers, linkages, and functional groups to identify candidates that are simultaneously active and durable remains a formidable challenge. Here we introduce Ara, a large-language-model (LLM) agent that leverages pretrained chemical knowledge, donor--acceptor theory, conjugation effects, and linkage stability hierarchies, to guide the search for photocatalytic COFs satisfying joint band-gap, band-edge, and hydrolytic-stability criteria. Evaluated against random search and Bayesian optimization (BO) over a space consisting of candidates with various nodes, linkers, linkages, and r-groups, screened with a GFN1-xTB fragment pipeline, Ara achieves a 52.7\\% hit rate (11.5$\\times$ random, p = 0.006), finds its first hit at iteration 12 versus 25 for random search, and significantly outperforms BO (p = 0.006). Inspection of the agent's reasoning traces reveals interpretable chemical logic: early convergence on vinylene and beta-ketoenamine linkages for stability, node selection informed by electron-withdrawing character, and systematic R-group optimization to center the band gap at 2.0 eV. Exhaustive evaluation of the full search space uncovers a complementary exploitation--exploration trade-off between the agent and BO, suggesting that hybrid strategies may combine the strengths of both approaches. These results demonstrate that LLM chemical priors can substantially accelerate multi-criteria materials discovery.", "date": "2026-03-05", "categories": []}
{"id": "2310.04925", "title": "Crystal-GFN: sampling crystals with desirable properties and constraints", "abstract": "The discovery of novel solid-state materials, such as electrocatalysts, super-ionic conductors, or photovoltaic materials, plays a critical role in addressing various global challenges. It has, for instance, the potential to significantly improve the efficiency of renewable energy production and storage, thereby making substantial contributions to climate crisis mitigation strategies. In this paper, we introduce Crystal-GFN, a generative model of crystal structures possessing desirable properties and constraints. Operating as a multi-environment, continuous-discrete GFlowNet, it sequentially samples structural attributes of crystalline materials, namely space group, composition and lattice parameters. This domain-inspired approach enables the flexible incorporation of physicochemical and geometric hard constraints. We demonstrate the capabilities of Crystal-GFN to efficiently discover diverse and valid crystals with various properties: low predicted formation energy (median -3.2 eV/atom), band gap close to a target value and high density. Overall, Crystal-GFN is a crystal generation method that addresses several existing challenges in the literature and opens promising paths for accelerating materials discovery with machine learning.", "date": "2023-10-07", "categories": []}
{"id": "2603.03946", "title": "Lang2Str: Two-Stage Crystal Structure Generation with LLMs and Continuous Flow Models", "abstract": "Generative models hold great promise for accelerating material discovery but are often limited by their inflexible single-stage generative process in designing valid and diverse materials. To address this, we propose a two-stage generative framework, Lang2Str, that combines the strengths of large language models (LLMs) and flow-based models for flexible and precise material generation. Our method frames the generative process as a conditional generative task, where an LLM provides high-level conditions by generating descriptions of material unit cells' geometric layouts and properties. These descriptions, informed by the LLM's extensive background knowledge, ensure reasonable structure designs. A conditioned flow model then decodes these textual conditions into precise continuous coordinates and unit cell parameters. This staged approach combines the structured reasoning of LLMs and the distribution modeling capabilities of flow models. Experimental results show that our method achieves competitive performance on \\textit{ab initio} material generation and crystal structure prediction tasks, with generated structures exhibiting closer alignment to ground truth in both geometry and energy levels, surpassing state-of-the-art models. The flexibility and modularity of our framework further enable fine-grained control over the generation process, potentially leading to more efficient and customizable material design.", "date": "2026-03-04", "categories": []}
{"id": "2502.15483", "title": "MoMa: A Modular Deep Learning Framework for Material Property Prediction", "abstract": "Deep learning methods for material property prediction have been widely explored to advance materials discovery. However, the prevailing pre-train then fine-tune paradigm often fails to address the inherent diversity and disparity of material tasks. To overcome these challenges, we introduce MoMa, a Modular framework for Materials that first trains specialized modules across a wide range of tasks and then adaptively composes synergistic modules tailored to each downstream scenario. Evaluation across 17 datasets demonstrates the superiority of MoMa, with a substantial 14% average improvement over the strongest baseline. Few-shot and continual learning experiments further highlight MoMa's potential for real-world applications. Pioneering a new paradigm of modular material learning, MoMa will be open-sourced to foster broader community collaboration.", "date": "2025-02-21", "categories": []}
{"id": "2509.19877", "title": "Advancing Universal Deep Learning for Electronic-Structure Hamiltonian Prediction of Materials", "abstract": "Deep learning methods for electronic-structure Hamiltonian prediction has offered significant computational efficiency advantages over traditional DFT methods, yet the diversity of atomic types, structural patterns, and the high-dimensional complexity of Hamiltonians pose substantial challenges to the generalization performance. In this work, we contribute on both the methodology and dataset sides to advance universal deep learning paradigm for Hamiltonian prediction. On the method side, we propose NextHAM, a neural E(3)-symmetry and expressive correction method for efficient and generalizable materials electronic-structure Hamiltonian prediction. First, we introduce the zeroth-step Hamiltonians, which can be efficiently constructed by the initial charge density of DFT, as informative descriptors of neural regression model in the input level and initial estimates of the target Hamiltonian in the output level, so that the regression model directly predicts the correction terms to the target ground truths, thereby significantly simplifying the input-output mapping for learning. Second, we present a neural Transformer architecture with strict E(3)-Symmetry and high non-linear expressiveness for Hamiltonian prediction. Third, we propose a novel training objective to ensure the accuracy performance of Hamiltonians in both real space and reciprocal space, preventing error amplification and the occurrence of \"ghost states\" caused by the large condition number of the overlap matrix. On the dataset side, we curate a high-quality broad-coverage large benchmark, namely Materials-HAM-SOC, comprising 17,000 material structures spanning 68 elements from six rows of the periodic table and explicitly incorporating SOC effects. Experimental results on Materials-HAM-SOC demonstrate that NextHAM achieves excellent accuracy and efficiency in predicting Hamiltonians and band structures.", "date": "2025-09-24", "categories": []}
{"id": "2511.22652", "title": "Generative Models for Crystalline Materials", "abstract": "Understanding structure-property relationships in materials is fundamental in condensed matter physics and materials science. Over the past few years, machine learning (ML) has emerged as a powerful tool for advancing this understanding and accelerating materials discovery. Early ML approaches primarily focused on constructing and screening large material spaces to identify promising candidates for various applications. More recently, research efforts have increasingly shifted toward generating crystal structures using end-to-end generative models. This review analyzes the current state of generative modeling for crystal structure prediction and de novo generation. It examines crystal representations, outlines the generative models used to design crystal structures, and evaluates their respective strengths and limitations. Furthermore, the review highlights experimental considerations for evaluating generated structures and provides recommendations for suitable existing software tools. Emerging topics, such as modeling disorder and defects, integration in advanced characterization, incorporating synthetic feasibility constraints, and model explainability are explored. Ultimately, this work aims to inform both experimental scientists looking to adapt suitable ML models to their specific circumstances and ML specialists seeking to understand the unique challenges related to inverse materials design and discovery.", "date": "2025-11-27", "categories": []}
{"id": "2512.13989", "title": "A Single Architecture for Representing Invariance Under Any Space Group", "abstract": "Incorporating known symmetries in data into machine learning models has consistently improved predictive accuracy, robustness, and generalization. However, achieving exact invariance to specific symmetries typically requires designing bespoke architectures for each group, limiting scalability and preventing knowledge transfer across related symmetries. In the case of the space groups, symmetries critical to modeling crystalline solids in materials science and condensed matter physics, this challenge is particularly salient as there are 230 such groups in three dimensions. In this work we present a new approach to such crystallographic symmetries by developing a single machine learning architecture that is capable of adapting its weights automatically to enforce invariance to any input space group. Our approach is based on constructing symmetry-adapted Fourier bases through an explicit characterization of constraints that group operations impose on Fourier coefficients. Encoding these constraints into a neural network layer enables weight sharing across different space groups, allowing the model to leverage structural similarities between groups and overcome data sparsity when limited measurements are available for specific groups. We demonstrate the effectiveness of this approach in achieving competitive performance on material property prediction tasks and performing zero-shot learning to generalize to unseen groups.", "date": "2025-12-16", "categories": []}
{"id": "2602.07543", "title": "MSP-LLM: A Unified Large Language Model Framework for Complete Material Synthesis Planning", "abstract": "Material synthesis planning (MSP) remains a fundamental and underexplored bottleneck in AI-driven materials discovery, as it requires not only identifying suitable precursor materials but also designing coherent sequences of synthesis operations to realize a target material. Although several AI-based approaches have been proposed to address isolated subtasks of MSP, a unified methodology for solving the entire MSP task has yet to be established. We propose MSP-LLM, a unified LLM-based framework that formulates MSP as a structured process composed of two constituent subproblems: precursor prediction (PP) and synthesis operation prediction (SOP). Our approach introduces a discrete material class as an intermediate decision variable that organizes both tasks into a chemically consistent decision chain. For SOP, we further incorporate hierarchical precursor types as synthesis-relevant inductive biases and employ an explicit conditioning strategy that preserves precursor-related information in the autoregressive decoding state. Extensive experiments show that MSP-LLM consistently outperforms existing methods on both PP and SOP, as well as on the complete MSP task, demonstrating an effective and scalable framework for MSP that can accelerate real-world materials discovery.", "date": "2026-02-07", "categories": []}
{"id": "2603.00092", "title": "A comparative study of transformer models and recurrent neural networks for path-dependent composite materials", "abstract": "Accurate modeling of Short Fiber Reinforced Composites (SFRCs) remains computationally expensive for full-field simulations. Data-driven surrogate models using Artificial Neural Networks (ANNs) have been proposed as an efficient alternative to numerical modeling, where Recurrent Neural Networks (RNNs) are increasingly being used for path-dependent multiscale modeling by predicting the homogenized response of a Representative Volume Element (RVE). However, recently, transformer models have been developed and they offer scalability and efficient parallelization, yet have not been systematically compared with RNNs in this field. In this study, we perform a systematic comparison between RNNs and transformer models trained on sequences of homogenized response of SFRC RVEs. We study the effect on two types of hyperparameters, namely architectural hyperparameters (such as the number of GRU layers, hidden size, number of attention heads, and encoder blocks) and training hyperparameters (such as learning rate and batch size). Both sets of hyperparameters are tuned using Bayesian optimization. We then analyze scaling laws with respect to dataset size and inference accuracy in interpolation and extrapolation regimes. The results show that while transformer models remain competitive in terms of accuracy on large datasets, the RNNs demonstrate better accuracy on small datasets and show better extrapolation performance. Furthermore, under extrapolation, there is a clear difference, where the RNN remains accurate, while the transformer model performs poorly. On the other hand, the transformer model is 7 times faster at inference, requiring 0.5 ms per prediction compared to the 3.5 ms per prediction for the RNN model.", "date": "2026-02-17", "categories": []}
{"id": "2603.00093", "title": "Adaptive Uncertainty-Guided Surrogates for Efficient phase field Modeling of Dendritic Solidification", "abstract": "The high computational cost of phase field simulations remains a major limitation for predicting dendritic solidification in metals, particularly in additive manufacturing, where microstructural control is critical. This work presents a surrogate model for dendritic solidification that employs uncertainty-driven adaptive sampling with XGBoost and CNNs, including a self-supervised strategy, to efficiently approximate the spatio-temporal evolution while reducing costly phase field simulations. The proposed adaptive strategy leverages model uncertainty, approximated via Monte Carlo dropout for CNNs and bagging for XGBoost, to identify high-uncertainty regions where new samples are generated locally within hyperspheres, progressively refining the spatio-temporal design space and achieving accurate predictions with significantly fewer phase field simulations than an Optimal Latin Hypercube Sampling optimized via discrete Particle Swarm Optimization (OLHS-PSO). The framework systematically investigates how temporal instance selection, adaptive sampling, and the choice between domain-informed and data-driven surrogates affect spatio-temporal model performance. Evaluation considers not only computational cost but also the number of expensive phase field simulations, surrogate accuracy, and associated $CO_2$ emissions, providing a comprehensive assessment of model performance as well as their related environmental impact.", "date": "2026-02-17", "categories": []}
{"id": "2504.19372", "title": "Composable and adaptive design of machine learning interatomic potentials guided by Fisher-information analysis", "abstract": "An adaptive physics-inspired model design strategy for machine-learning interatomic potentials (MLIPs) is proposed. This strategy relies on iterative reconfigurations of composite models from single-term models, followed by a unified training procedure. A model evaluation method based on the Fisher information matrix (FIM) and multiple-property error metrics is also proposed to guide the model reconfiguration and hyperparameter optimization. By combining the reconfiguration and the evaluation subroutines, we provide an adaptive MLIP design strategy that balances flexibility and extensibility. In a case study of designing models against a structurally diverse niobium dataset, we managed to obtain an optimal model configuration with 75 parameters generated by our framework that achieved a force RMSE of 0.172 eV/{\\AA} and an energy RMSE of 0.013 eV/atom.", "date": "2025-04-27", "categories": []}
{"id": "2602.22086", "title": "MBD-ML: Many-body dispersion from machine learning for molecules and materials", "abstract": "Van der Waals (vdW) interactions are essential for describing molecules and materials, from drug design and catalysis to battery applications. These omnipresent interactions must also be accurately included in machine-learned force fields. The many-body dispersion (MBD) method stands out as one of the most accurate and transferable approaches to capture vdW interactions, requiring only atomic $C_6$ coefficients and polarizabilities as input. We present MBD-ML, a pretrained message passing neural network that predicts these atomic properties directly from atomic structures. Through seamless integration with libMBD, our method enables the immediate calculation of MBD-inclusive total energies, forces, and stress tensors. By eliminating the need for intermediate electronic structure calculations, MBD-ML offers a practical and streamlined tool that simplifies the incorporation of state-of-the-art vdW interactions into any electronic structure code, as well as empirical and machine-learned force fields.", "date": "2026-02-25", "categories": []}
{"id": "2602.19153", "title": "Constrained Diffusion for Accelerated Structure Relaxation of Inorganic Solids with Point Defects", "abstract": "Point defects affect material properties by altering electronic states and modifying local bonding environments. However, high-throughput first-principles simulations of point defects are costly due to large simulation cells and complex energy landscapes. To this end, we propose a generative framework for simulating point defects, overcoming the limits of costly first-principles simulators. By leveraging a primal-dual algorithm, we introduce a constraint-aware diffusion model which outperforms existing constrained diffusion approaches in this domain. Across six defect configuration settings for Bi2Te3, the proposed approach provides state-of-the-art performance generating physically grounded structures.", "date": "2026-02-22", "categories": []}
{"id": "2602.19915", "title": "Fully Convolutional Spatiotemporal Learning for Microstructure Evolution Prediction", "abstract": "Understanding and predicting microstructure evolution is fundamental to materials science, as it governs the resulting properties and performance of materials. Traditional simulation methods, such as phase-field models, offer high-fidelity results but are computationally expensive due to the need to solve complex partial differential equations at fine spatiotemporal resolutions. To address this challenge, we propose a deep learning-based framework that accelerates microstructure evolution predictions while maintaining high accuracy. Our approach utilizes a fully convolutional spatiotemporal model trained in a self-supervised manner using sequential images generated from simulations of microstructural processes, including grain growth and spinodal decomposition. The trained neural network effectively learns the underlying physical dynamics and can accurately capture both short-term local behaviors and long-term statistical properties of evolving microstructures, while also demonstrating generalization to unseen spatiotemporal domains and variations in configuration and material parameters. Compared to recurrent neural architectures, our model achieves state-of-the-art predictive performance with significantly reduced computational cost in both training and inference. This work establishes a robust baseline for spatiotemporal learning in materials science and offers a scalable, data-driven alternative for fast and reliable microstructure simulations.", "date": "2026-02-23", "categories": []}
{"id": "2602.17730", "title": "Clever Materials: When Models Identify Good Materials for the Wrong Reasons", "abstract": "Machine learning can accelerate materials discovery. Models perform impressively on many benchmarks. However, strong benchmark performance does not imply that a model learned chemistry. I test a concrete alternative hypothesis: that property prediction can be driven by bibliographic confounding. Across five tasks spanning MOFs (thermal and solvent stability), perovskite solar cells (efficiency), batteries (capacity), and TADF emitters (emission wavelength), models trained on standard chemical descriptors predict author, journal, and publication year well above chance. When these predicted metadata (\"bibliographic fingerprints\") are used as the sole input to a second model, performance is sometimes competitive with conventional descriptor-based predictors. These results show that many datasets do not rule out non-chemical explanations of success. Progress requires routine falsification tests (e.g., group/time splits and metadata ablations), datasets designed to resist spurious correlations, and explicit separation of two goals: predictive utility versus evidence of chemical understanding.", "date": "2026-02-18", "categories": []}
{"id": "2602.17750", "title": "Inelastic Constitutive Kolmogorov-Arnold Networks: A generalized framework for automated discovery of interpretable inelastic material models", "abstract": "A key problem of solid mechanics is the identification of the constitutive law of a material, that is, the relation between strain and stress. Machine learning has lead to considerable advances in this field lately. Here we introduce inelastic Constitutive Kolmogorov-Arnold Networks (iCKANs). This novel artificial neural network architecture can discover in an automated manner symbolic constitutive laws describing both the elastic and inelastic behavior of materials. That is, it can translate data from material testing into corresponding elastic and inelastic potential functions in closed mathematical form. We demonstrate the advantages of iCKANs using both synthetic data and experimental data of the viscoelastic polymer materials VHB 4910 and VHB 4905. The results demonstrate that iCKANs accurately capture complex viscoelastic behavior while preserving physical interpretability. It is a particular strength of iCKANs that they can process not only mechanical data but also arbitrary additional information available about a material (e.g., about temperature-dependent behavior). This makes iCKANs a powerful tool to discover in the future also how specific processing or service conditions affect the properties of materials.", "date": "2026-02-19", "categories": []}
{"id": "2602.18213", "title": "Machine-learning force-field models for dynamical simulations of metallic magnets", "abstract": "We review recent advances in machine learning (ML) force-field methods for Landau-Lifshitz-Gilbert (LLG) simulations of itinerant electron magnets, focusing on scalability and transferability. Built on the principle of locality, a deep neural network model is developed to efficiently and accurately predict the electron-mediated forces governing spin dynamics. Symmetry-aware descriptors constructed through a group-theoretical approach ensure rigorous incorporation of both lattice and spin-rotation symmetries. The framework is demonstrated using the prototypical s-d exchange model widely employed in spintronics. ML-enabled large-scale simulations reveal novel nonequilibrium phenomena, including anomalous coarsening of tetrahedral spin order on the triangular lattice and the freezing of phase separation dynamics in lightly hole-doped, strong-coupling square-lattice systems. These results establish ML force-field frameworks as scalable, accurate, and versatile tools for modeling nonequilibrium spin dynamics in itinerant magnets.", "date": "2026-02-20", "categories": []}
{"id": "2602.15923", "title": "A fully differentiable framework for training proxy Exchange Correlation Functionals for periodic systems", "abstract": "Density Functional Theory (DFT) is widely used for first-principles simulations in chemistry and materials science, but its computational cost remains a key limitation for large systems. Motivated by recent advances in ML-based exchange-correlation (XC) functionals, this paper introduces a differentiable framework that integrates machine learning models into density functional theory (DFT) for solids and other periodic systems. The framework defines a clean API for neural network models that can act as drop in replacements for conventional exchange-correlation (XC) functionals and enables gradients to flow through the full self-consistent DFT workflow. The framework is implemented in Python using a PyTorch backend, making it fully differentiable and easy to use with standard deep learning tools. We integrate the implementation with the DeepChem library to promote the reuse of established models and to lower the barrier for experimentation. In initial benchmarks against established electronic structure packages (GPAW and PySCF), our models achieve relative errors on the order of 5-10%.", "date": "2026-02-17", "categories": []}
{"id": "2602.13312", "title": "PeroMAS: A Multi-agent System of Perovskite Material Discovery", "abstract": "As a pioneer of the third-generation photovoltaic revolution, Perovskite Solar Cells (PSCs) are renowned for their superior optoelectronic performance and cost potential. The development process of PSCs is precise and complex, involving a series of closed-loop workflows such as literature retrieval, data integration, experimental design, and synthesis. However, existing AI perovskite approaches focus predominantly on discrete models, including material design, process optimization,and property prediction. These models fail to propagate physical constraints across the workflow, hindering end-to-end optimization. In this paper, we propose a multi-agent system for perovskite material discovery, named PeroMAS. We first encapsulated a series of perovskite-specific tools into Model Context Protocols (MCPs). By planning and invoking these tools, PeroMAS can design perovskite materials under multi-objective constraints, covering the entire process from literature retrieval and data extraction to property prediction and mechanism analysis. Furthermore, we construct an evaluation benchmark by perovskite human experts to assess this multi-agent system. Results demonstrate that, compared to single Large Language Model (LLM) or traditional search strategies, our system significantly enhances discovery efficiency. It successfully identified candidate materials satisfying multi-objective constraints. Notably, we verify PeroMAS's effectiveness in the physical world through real synthesis experiments.", "date": "2026-02-10", "categories": []}
{"id": "2602.14867", "title": "Fast and accurate quasi-atom method for simultaneous atomistic and continuum simulation of solids", "abstract": "We report a novel hybrid method of simultaneous atomistic simulation of solids in critical regions (contacts surfaces, cracks areas, etc.), along with continuum modeling of other parts. The continuum is treated in terms of quasi-atoms of different size, comprising composite medium. The parameters of interaction potential between the quasi-atoms are optimized to match elastic properties of the composite medium to those of the atomic one. The optimization method coincides conceptually with the online Machine Learning (ML) methods, making it computationally very efficient. Such an approach allows a straightforward application of standard software packages for molecular dynamics (MD), supplemented by the ML-based optimizer. The new method is applied to model systems with a simple, pairwise Lennard-Jones potential, as well with multi-body Tersoff potential, describing covalent bonds. Using LAMMPS software we simulate collision of particles of different size. Comparing simulation results, obtained by the novel method, with full-atomic simulations, we demonstrate its accuracy, validity and overwhelming superiority in computational speed. Furthermore, we compare our method with other hybrid methods, specifically, with the closest one -- AtC (Atomic to Continuum) method. We demonstrate a significant superiority of our approach in computational speed and implementation convenience. Finally, we discuss a possible extension of the method for modeling other phenomena.", "date": "2026-02-16", "categories": []}
{"id": "2602.13140", "title": "FlashSchNet: Fast and Accurate Coarse-Grained Neural Network Molecular Dynamics", "abstract": "Graph neural network (GNN) potentials such as SchNet improve the accuracy and transferability of molecular dynamics (MD) simulation by learning many-body interactions, but remain slower than classical force fields due to fragmented kernels and memory-bound pipelines that underutilize GPUs. We show that a missing principle is making GNN-MD IO-aware, carefully accounting for reads and writes between GPU high-bandwidth memory (HBM) and on-chip SRAM. We present FlashSchNet, an efficient and accurate IO-aware SchNet-style GNN-MD framework built on four techniques: (1) flash radial basis, which fuses pairwise distance computation, Gaussian basis expansion, and cosine envelope into a single tiled pass, computing each distance once and reusing it across all basis functions; (2) flash message passing, which fuses cutoff, neighbor gather, filter multiplication, and reduction to avoid materializing edge tensors in HBM; (3) flash aggregation, which reformulates scatter-add via CSR segment reduce, reducing atomic writes by a factor of feature dimension and enabling contention-free accumulation in both forward and backward passes; (4) channel-wise 16-bit quantization that exploits the low per-channel dynamic range in SchNet MLP weights to further improve throughput with negligible accuracy loss. On a single NVIDIA RTX PRO 6000, FlashSchNet achieves 1000 ns/day aggregate simulation throughput over 64 parallel replicas on coarse-grained (CG) protein containing 269 beads (6.5x faster than CGSchNet baseline with 80% reduction of peak memory), surpassing classical force fields (e.g. MARTINI) while retaining SchNet-level accuracy and transferability.", "date": "2026-02-13", "categories": []}
{"id": "2602.12045", "title": "Fourier Transformers for Latent Crystallographic Diffusion and Generative Modeling", "abstract": "The discovery of new crystalline materials calls for generative models that handle periodic boundary conditions, crystallographic symmetries, and physical constraints, while scaling to large and structurally diverse unit cells. We propose a reciprocal-space generative pipeline that represents crystals through a truncated Fourier transform of the species-resolved unit-cell density, rather than modeling atomic coordinates directly. This representation is periodicity-native, admits simple algebraic actions of space-group symmetries, and naturally supports variable atomic multiplicities during generation, addressing a common limitation of particle-based approaches. Using only nine Fourier basis functions per spatial dimension, our approach reconstructs unit cells containing up to 108 atoms per chemical species. We instantiate this pipeline with a transformer variational autoencoder over complex-valued Fourier coefficients, and a latent diffusion model that generates in the compressed latent space. We evaluate reconstruction and latent diffusion on the LeMaterial benchmark and compare unconditional generation against coordinate-based baselines in the small-cell regime ($\\leq 16$ atoms per unit cell).", "date": "2026-02-12", "categories": []}
{"id": "2601.16376", "title": "Polymer-Agent: Large Language Model Agent for Polymer Design", "abstract": "On-demand Polymer discovery is essential for various industries, ranging from biomedical to reinforcement materials. Experiments with polymers have a long trial-and-error process, leading to use of extensive resources. For these processes, machine learning has accelerated scientific discovery at the property prediction and latent space search fronts. However, laboratory researchers cannot readily access codes and these models to extract individual structures and properties due to infrastructure limitations. We present a closed-loop polymer structure-property predictor integrated in a terminal for early-stage polymer discovery. The framework is powered by LLM reasoning to provide users with property prediction, property-guided polymer structure generation, and structure modification capabilities. The SMILES sequences are guided by the synthetic accessibility score and the synthetic complexity score (SC Score) to ensure that polymer generation is as close as possible to synthetically accessible monomer-level structures. This framework addresses the challenge of generating novel polymer structures for laboratory researchers, thereby providing computational insights into polymer research.", "date": "2026-01-23", "categories": []}
{"id": "2602.11123", "title": "From Natural Language to Materials Discovery:The Materials Knowledge Navigation Agent", "abstract": "Accelerating the discovery of high-performance materials remains a central challenge across energy, electronics, and aerospace technologies, where traditional workflows depend heavily on expert intuition and computationally expensive simulations. Here we introduce the Materials Knowledge Navigation Agent (MKNA), a language-driven system that translates natural-language scientific intent into executable actions for database retrieval, property prediction, structure generation, and stability evaluation. Beyond automating tool invocation, MKNA autonomously extracts quantitative thresholds and chemically meaningful design motifs from literature and database evidence, enabling data-grounded hypothesis formation. Applied to the search for high-Debye-temperature ceramics, the agent identifies a literature-supported screening criterion (Theta_D > 800 K), rediscovers canonical ultra-stiff materials such as diamond, SiC, SiN, and BeO, and proposes thermodynamically stable, previously unreported Be-C-rich compounds that populate the sparsely explored 1500-1700 K regime. These results demonstrate that MKNA not only finds stable candidates but also reconstructs interpretable design heuristics, establishing a generalizable platform for autonomous, language-guided materials exploration.", "date": "2026-02-11", "categories": []}
{"id": "2602.09093", "title": "Predicting magnetism with first-principles AI", "abstract": "Computational discovery of magnetic materials remains challenging because magnetism arises from the competition between kinetic energy and Coulomb interaction that is often beyond the reach of standard electronic-structure methods. Here we tackle this challenge by directly solving the many-electron Schr\\\"odinger equation with neural-network variational Monte Carlo, which provides a highly expressive variational wavefunction for strongly correlated systems. Applying this technique to transition metal dichalcogenide moir\\'e semicondutors, we predict itinerant ferromagnetism in WSe$_2$/WS$_2$ and an antiferromagnetic insulator in twisted $\\Gamma$-valley homobilayer, using the same neural network without any physics input beyond the microscopic Hamiltonian. Crucially, both types of magnetic states are obtained from a single calculation within the $S_z=0$ sector, removing the need to compute and compare multiple $S_z$ sectors. This significantly reduces computational cost and paves the way for faster and more reliable magnetic material design.", "date": "2026-02-09", "categories": []}
{"id": "2308.07897", "title": "Probabilistic Phase Labeling and Lattice Refinement for Autonomous Material Research", "abstract": "X-ray diffraction (XRD) is an essential technique to determine a material's crystal structure in high-throughput experimentation, and has recently been incorporated in artificially intelligent agents in autonomous scientific discovery processes. However, rapid, automated and reliable analysis method of XRD data matching the incoming data rate remains a major challenge. To address these issues, we present CrystalShift, an efficient algorithm for probabilistic XRD phase labeling that employs symmetry-constrained pseudo-refinement optimization, best-first tree search, and Bayesian model comparison to estimate probabilities for phase combinations without requiring phase space information or training. We demonstrate that CrystalShift provides robust probability estimates, outperforming existing methods on synthetic and experimental datasets, and can be readily integrated into high-throughput experimental workflows. In addition to efficient phase-mapping, CrystalShift offers quantitative insights into materials' structural parameters, which facilitate both expert evaluation and AI-based modeling of the phase space, ultimately accelerating materials identification and discovery.", "date": "2023-08-15", "categories": []}
{"id": "2512.16545", "title": "Predictive Inorganic Synthesis based on Machine Learning using Small Data sets: a case study of size-controlled Cu Nanoparticles", "abstract": "Copper nanoparticles (Cu NPs) have a broad applicability, yet their synthesis is sensitive to subtle changes in reaction parameters. This sensitivity, combined with the time- and resource-intensive nature of experimental optimization, poses a major challenge in achieving reproducible and size-controlled synthesis. While Machine Learning (ML) shows promise in materials research, its application is often limited by scarcity of large high-quality experimental data sets. This study explores ML to predict the size of Cu NPs from microwave-assisted polyol synthesis using a small data set of 25 in-house performed syntheses. Latin Hypercube Sampling is used to efficiently cover the parameter space while creating the experimental data set. Ensemble regression models successfully predict particle sizes with high accuracy ($R^2 = 0.74$), outperforming classical statistical approaches ($R^2 = 0.60$). Additionally, classification models using both random forests and Large Language Models (LLMs) are evaluated to distinguish between large and small particles. While random forests show moderate performance, LLMs offer no significant advantages under data-scarce conditions. Overall, this study demonstrates that carefully curated small data sets, paired with robust classical ML, can effectively predict the synthesis of Cu NPs and highlights that for lab-scale studies, complex models like LLMs may offer limited benefit over simpler techniques.", "date": "2025-12-18", "categories": []}
{"id": "2601.04510", "title": "Towards Spatio-Temporal Extrapolation of Phase-Field Simulations with Convolution-Only Neural Networks", "abstract": "Phase-field simulations of liquid metal dealloying (LMD) can capture complex microstructural evolutions but can be prohibitively expensive for large domains and long time horizons. In this paper, we introduce a fully convolutional, conditionally parameterized U-Net surrogate designed to extrapolate far beyond its training data in both space and time. The architecture integrates convolutional self-attention, physically informed padding, and a flood-fill corrector method to maintain accuracy under extreme extrapolation, while conditioning on simulation parameters allows for flexible time-step skipping and adaptation to varying alloy compositions. To remove the need for costly solver-based initialization, we couple the surrogate with a conditional diffusion model that generates synthetic, physically consistent initial conditions. We train our surrogate on simulations generated over small domain sizes and short time spans, but, by taking advantage of the convolutional nature of U-Nets, we are able to run and extrapolate surrogate simulations for longer time horizons than what would be achievable with classic numerical solvers. Across multiple alloy compositions, the framework is able to reproduce the LMD physics accurately. It predicts key quantities of interest and spatial statistics with relative errors typically below 5% in the training regime and under 15% during large-scale, long time-horizon extrapolations. Our framework can also deliver speed-ups of up to 36,000 times, bringing the time to run weeks-long simulations down to a few seconds. This work is a first stepping stone towards high-fidelity extrapolation in both space and time of phase-field simulation for LMD.", "date": "2026-01-08", "categories": []}
{"id": "2602.07192", "title": "Systematic Performance Assessment of Deep Material Networks for Multiscale Material Modeling", "abstract": "Deep Material Networks (DMNs) are structure-preserving, mechanistic machine learning models that embed micromechanical principles into their architectures, enabling strong extrapolation capabilities and significant potential to accelerate multiscale modeling of complex microstructures. A key advantage of these models is that they can be trained exclusively on linear elastic data and then generalized to nonlinear inelastic regimes during online prediction. Despite their growing adoption, systematic evaluations of their performance across the full offline-online pipeline remain limited. This work presents a comprehensive comparative assessment of DMNs with respect to prediction accuracy, computational efficiency, and training robustness. We investigate the effects of offline training choices, including initialization, batch size, training data size, and activation regularization on online generalization performance and uncertainty. The results demonstrate that both prediction error and variance decrease with increasing training data size, while initialization and batch size can significantly influence model performance. Moreover, activation regularization is shown to play a critical role in controlling network complexity and therefore generalization performance. Compared with the original DMN, the rotation-free Interaction-based Material Network (IMN) formulation achieves a 3.4x - 4.7x speed-up in offline training, while maintaining comparable online prediction accuracy and computational efficiency. These findings clarify key trade-offs between model expressivity and efficiency in structure-preserving material networks and provide practical guidance for their deployment in multiscale material modeling.", "date": "2026-02-06", "categories": []}
{"id": "2602.07310", "title": "Optimization of Precipitate Segmentation Through Linear Genetic Programming of Image Processing", "abstract": "Current analysis of additive manufactured niobium-based copper alloys relies on hand annotation due to varying contrast, noise, and image artifacts present in micrographs, slowing iteration speed in alloy development. We present a filtering and segmentation algorithm for detecting precipitates in FIB cross-section micrographs, optimized using linear genetic programming (LGP), which accounts for the various artifacts. To this end, the optimization environment uses a domain-specific language for image processing to iterate on solutions. Programs in this language are a list of image-filtering blocks with tunable parameters that sequentially process an input image, allowing for reliable generation and mutation by a genetic algorithm. Our environment produces optimized human-interpretable MATLAB code representing an image filtering pipeline. Under ideal conditions--a population size of 60 and a maximum program length of 5 blocks--our system was able to find a near-human accuracy solution with an average evaluation error of 1.8% when comparing segmentations pixel-by-pixel to a human baseline using an XOR error evaluation. Our automation work enabled faster iteration cycles and furthered exploration of the material composition and processing space: our optimized pipeline algorithm processes a 3.6 megapixel image in about 2 seconds on average. This ultimately enables convergence on strong, low-activation, precipitation hardened copper alloys for additive manufactured fusion reactor parts.", "date": "2026-02-07", "categories": []}
{"id": "2501.18064", "title": "Learning Metal Microstructural Heterogeneity through Spatial Mapping of Diffraction Latent Space Features", "abstract": "To leverage advancements in machine learning for metallic materials design and property prediction, it is crucial to develop a data-reduced representation of metal microstructures that surpasses the limitations of current physics-based discrete microstructure descriptors. This need is particularly relevant for metallic materials processed through additive manufacturing, which exhibit complex hierarchical microstructures that cannot be adequately described using the conventional metrics typically applied to wrought materials. Furthermore, capturing the spatial heterogeneity of microstructures at the different scales is necessary within such framework to accurately predict their properties. To address these challenges, we propose the physical spatial mapping of metal diffraction latent space features. This approach integrates (i) point diffraction data encoding via variational autoencoders or contrastive learning and (ii) the physical mapping of the encoded values. Together these steps offer a method offers a novel means to comprehensively describe metal microstructures. We demonstrate this approach on a wrought and additively manufactured alloy, showing that it effectively encodes microstructural information and enables direct identification of microstructural heterogeneity not directly possible by physics-based models. This data-reduced microstructure representation opens the application of machine learning models in accelerating metallic material design and accurately predicting their properties.", "date": "2025-01-30", "categories": []}
{"id": "2510.12091", "title": "ToPolyAgent: AI Agents for Coarse-Grained Topological Polymer Simulations", "abstract": "We introduce ToPolyAgent, a multi-agent AI framework for performing coarse-grained molecular dynamics (MD) simulations of topological polymers through natural language instructions. By integrating large language models (LLMs) with domain-specific computational tools, ToPolyAgent supports both interactive and autonomous simulation workflows across diverse polymer architectures, including linear, ring, brush, and star polymers, as well as dendrimers. The system consists of four LLM-powered agents: a Config Agent for generating initial polymer-solvent configurations, a Simulation Agent for executing LAMMPS-based MD simulations and conformational analyses, a Report Agent for compiling markdown reports, and a Workflow Agent for streamlined autonomous operations. Interactive mode incorporates user feedback loops for iterative refinements, while autonomous mode enables end-to-end task execution from detailed prompts. We demonstrate ToPolyAgent's versatility through case studies involving diverse polymer architectures under varying solvent condition, thermostats, and simulation lengths. Furthermore, we highlight its potential as a research assistant by directing it to investigate the effect of interaction parameters on the linear polymer conformation, and the influence of grafting density on the persistence length of the brush polymer. By coupling natural language interfaces with rigorous simulation tools, ToPolyAgent lowers barriers to complex computational workflows and advances AI-driven materials discovery in polymer science. It lays the foundation for autonomous and extensible multi-agent scientific research ecosystems.", "date": "2025-10-14", "categories": []}
{"id": "2602.00169", "title": "Towards Agentic Intelligence for Materials Science", "abstract": "The convergence of artificial intelligence and materials science presents a transformative opportunity, but achieving true acceleration in discovery requires moving beyond task-isolated, fine-tuned models toward agentic systems that plan, act, and learn across the full discovery loop. This survey advances a unique pipeline-centric view that spans from corpus curation and pretraining, through domain adaptation and instruction tuning, to goal-conditioned agents interfacing with simulation and experimental platforms. Unlike prior reviews, we treat the entire process as an end-to-end system to be optimized for tangible discovery outcomes rather than proxy benchmarks. This perspective allows us to trace how upstream design choices-such as data curation and training objectives-can be aligned with downstream experimental success through effective credit assignment.\n  To bridge communities and establish a shared frame of reference, we first present an integrated lens that aligns terminology, evaluation, and workflow stages across AI and materials science. We then analyze the field through two focused lenses: From the AI perspective, the survey details LLM strengths in pattern recognition, predictive analytics, and natural language processing for literature mining, materials characterization, and property prediction; from the materials science perspective, it highlights applications in materials design, process optimization, and the acceleration of computational workflows via integration with external tools (e.g., DFT, robotic labs). Finally, we contrast passive, reactive approaches with agentic design, cataloging current contributions while motivating systems that pursue long-horizon goals with autonomy, memory, and tool use. This survey charts a practical roadmap towards autonomous, safety-aware LLM agents aimed at discovering novel and useful materials.", "date": "2026-01-29", "categories": []}
{"id": "2602.05702", "title": "Broken neural scaling laws in materials science", "abstract": "In materials science, data are scarce and expensive to generate, whether computationally or experimentally. Therefore, it is crucial to identify how model performance scales with dataset size and model capacity to distinguish between data- and model-limited regimes. Neural scaling laws provide a framework for quantifying this behavior and guide the design of materials datasets and machine learning architectures. Here, we investigate neural scaling laws for a paradigmatic materials science task: predicting the dielectric function of metals, a high-dimensional response that governs how solids interact with light. Using over 200,000 dielectric functions from high-throughput ab initio calculations, we study two multi-objective graph neural networks trained to predict the frequency-dependent complex interband dielectric function and the Drude frequency. We observe broken neural scaling laws with respect to dataset size, whereas scaling with the number of model parameters follows a simple power law that rapidly saturates.", "date": "2026-02-05", "categories": []}
{"id": "2511.11388", "title": "Robust inverse material design with physical guarantees using the Voigt-Reuss Net", "abstract": "We propose a spectrally normalized surrogate for forward and inverse mechanical homogenization with hard physical guarantees. Leveraging the Voigt-Reuss bounds, we factor their difference via a Cholesky-like operator and learn a dimensionless, symmetric positive semi-definite representation with eigenvalues in $[0,1]$; the inverse map returns symmetric positive-definite predictions that lie between the bounds in the L\\\"owner sense. In 3D linear elasticity on an open dataset of stochastic biphasic microstructures, a fully connected Voigt-Reuss net trained on $>\\!7.5\\times 10^{5}$ FFT-based labels with 236 isotropy-invariant descriptors and three contrast parameters recovers the isotropic projection with near-perfect fidelity (isotropy-related entries: $R^2 \\ge 0.998$), while anisotropy-revealing couplings are unidentifiable from $SO(3)$-invariant inputs. Tensor-level relative Frobenius errors have median $\\approx 1.7\\%$ and mean $\\approx 3.4\\%$ across splits. For 2D plane strain on thresholded trigonometric microstructures, coupling spectral normalization with a differentiable renderer and a CNN yields $R^2>0.99$ on all components, subpercent normalized losses, accurate tracking of percolation-induced eigenvalue jumps, and robust generalization to out-of-distribution images. Treating the parametric microstructure as design variables, batched first-order optimization with a single surrogate matches target tensors within a few percent and returns diverse near-optimal designs. Overall, the Voigt-Reuss net unifies accurate, physically admissible forward prediction with large-batch, constraint-consistent inverse design, and is generic to elliptic operators and coupled-physics settings.", "date": "2025-11-14", "categories": []}
{"id": "2602.04323", "title": "Efficient Equivariant High-Order Crystal Tensor Prediction via Cartesian Local-Environment Many-Body Coupling", "abstract": "End-to-end prediction of high-order crystal tensor properties from atomic structures remains challenging: while spherical-harmonic equivariant models are expressive, their Clebsch-Gordan tensor products incur substantial compute and memory costs for higher-order targets. We propose the Cartesian Environment Interaction Tensor Network (CEITNet), an approach that constructs a multi-channel Cartesian local environment tensor for each atom and performs flexible many-body mixing via a learnable channel-space interaction. By performing learning in channel space and using Cartesian tensor bases to assemble equivariant outputs, CEITNet enables efficient construction of high-order tensor. Across benchmark datasets for order-2 dielectric, order-3 piezoelectric, and order-4 elastic tensor prediction, CEITNet surpasses prior high-order prediction methods on key accuracy criteria while offering high computational efficiency.", "date": "2026-02-04", "categories": []}
{"id": "2602.04435", "title": "Machine Learning-Driven Crystal System Prediction for Perovskites Using Augmented X-ray Diffraction Data", "abstract": "Prediction of crystal system from X-ray diffraction (XRD) spectra is a critical task in materials science, particularly for perovskite materials which are known for their diverse applications in photovoltaics, optoelectronics, and catalysis. In this study, we present a machine learning (ML)-driven framework that leverages advanced models, including Time Series Forest (TSF), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and a simple feedforward neural network (NN), to classify crystal systems, point groups, and space groups from XRD data of perovskite materials. To address class imbalance and enhance model robustness, we integrated feature augmentation strategies such as Synthetic Minority Over-sampling Technique (SMOTE), class weighting, jittering, and spectrum shifting, along with efficient data preprocessing pipelines. The TSF model with SMOTE augmentation achieved strong performance for crystal system prediction, with a Matthews correlation coefficient (MCC) of 0.9, an F1 score of 0.92, and an accuracy of 97.76%. For point and space group prediction, balanced accuracies above 95% were obtained. The model demonstrated high performance for symmetry-distinct classes, including cubic crystal systems, point groups 3m and m-3m, and space groups Pnma and Pnnn. This work highlights the potential of ML for XRD-based structural characterization and accelerated discovery of perovskite materials", "date": "2026-02-04", "categories": []}
{"id": "2602.04734", "title": "DMFlow: Disordered Materials Generation by Flow Matching", "abstract": "The design of materials with tailored properties is crucial for technological progress. However, most deep generative models focus exclusively on perfectly ordered crystals, neglecting the important class of disordered materials. To address this gap, we introduce DMFlow, a generative framework specifically designed for disordered crystals. Our approach introduces a unified representation for ordered, Substitutionally Disordered (SD), and Positionally Disordered (PD) crystals, and employs a flow matching model to jointly generate all structural components. A key innovation is a Riemannian flow matching framework with spherical reparameterization, which ensures physically valid disorder weights on the probability simplex. The vector field is learned by a novel Graph Neural Network (GNN) that incorporates physical symmetries and a specialized message-passing scheme. Finally, a two-stage discretization procedure converts the continuous weights into multi-hot atomic assignments. To support research in this area, we release a benchmark containing SD, PD, and mixed structures curated from the Crystallography Open Database. Experiments on Crystal Structure Prediction (CSP) and De Novo Generation (DNG) tasks demonstrate that DMFlow significantly outperforms state-of-the-art baselines adapted from ordered crystal generation. We hope our work provides a foundation for the AI-driven discovery of disordered materials.", "date": "2026-02-04", "categories": []}
{"id": "2602.02671", "title": "MARA: Continuous SE(3)-Equivariant Attention for Molecular Force Fields", "abstract": "Machine learning force fields (MLFFs) have become essential for accurate and efficient atomistic modeling. Despite their high accuracy, most existing approaches rely on fixed angular expansions, limiting flexibility in weighting local geometric interactions. We introduce Modular Angular-Radial Attention (MARA), a module that extends spherical attention -- originally developed for SO(3) tasks -- to the molecular domain and SE(3), providing an efficient approximation of equivariant interactions. MARA operates directly on the angular and radial coordinates of neighboring atoms, enabling flexible, geometrically informed, and modular weighting of local environments. Unlike existing attention mechanisms in SE(3)-equivariant architectures, MARA can be integrated in a plug-and-play manner into models such as MACE without architectural modifications. Across molecular benchmarks, MARA improves energy and force predictions, reduces high-error events, and enhances robustness. These results demonstrate that continuous spherical attention is an effective and generalizable geometric operator that increases the expressiveness, stability, and reliability of atomistic models.", "date": "2026-02-02", "categories": []}
{"id": "2602.03438", "title": "Acceleration of Atomistic NEGF: Algorithms, Parallelization, and Machine Learning", "abstract": "The Non-equilibrium Green's function (NEGF) formalism is a particularly powerful method to simulate the quantum transport properties of nanoscale devices such as transistors, photo-diodes, or memory cells, in the ballistic limit of transport or in the presence of various scattering sources such as electronphonon, electron-photon, or even electron-electron interactions. The inclusion of all these mechanisms has been first demonstrated in small systems, composed of a few atoms, before being scaled up to larger structures made of thousands of atoms. Also, the accuracy of the models has kept improving, from empirical to fully ab-initio ones, e.g., density functional theory (DFT). This paper summarizes key (algorithmic) achievements that have allowed us to bring DFT+NEGF simulations closer to the dimensions and functionality of realistic systems. The possibility of leveraging graph neural networks and machine learning to speed up ab-initio device simulations is discussed as well.", "date": "2026-02-03", "categories": []}
{"id": "2502.11085", "title": "On the Importance of Pretraining Data Alignment for Atomic Property Prediction", "abstract": "This paper challenges the recent paradigm in atomic property prediction that links progress to growing dataset sizes and computational resources. We show that pretraining on a carefully selected task-aligned dataset can match or even surpass large-scale joint pretraining while using only 1/24th of the pretraining budget. We introduce the Chemical Similarity Index (CSI), a simple metric for molecular graphs inspired by the Fr\\'echet Inception Distance in computer vision, which quantifies the alignment between upstream pretraining datasets and downstream tasks. By selecting the most aligned dataset with minimal CSI distance, we show that models pretrained on a smaller, focused dataset consistently achieve better performance on downstream tasks than those pretrained on massive, mixed datasets such as JMP. This holds even when the mixed dataset includes the upstream dataset most aligned with the downstream task. Counterintuitively, we also find that indiscriminately adding more data can degrade model performance when the additional data is poorly aligned with the target task. Our findings highlight that quality often outperforms quantity in pretraining for atomic property prediction.", "date": "2025-02-16", "categories": []}
{"id": "2512.01067", "title": "On The Finetuning of MLIPs Through the Lens of Iterated Maps With BPTT", "abstract": "Accurate structural relaxation is critical for advanced materials design. Traditional approaches built on physics-derived first-principles calculations are computationally expensive, motivating the creation of machine-learning interatomic potentials (MLIPs), which strive to faithfully reproduce first-principles computed forces. We propose a fine-tuning method to be used on a pretrained MLIP in which we create a fully-differentiable end-to-end simulation loop that optimizes the predicted final structures directly. Trajectories are unrolled and gradients are tracked through the entire relaxation. We show that this method consistently improves performance across all evaluated pretrained models; resulting in an average of roughly 32% reduction in prediction error. Interestingly, we show the process is robust to substantial variation in the relaxation setup, achieving negligibly different results across varied hyperparameter and procedural modifications.", "date": "2025-11-30", "categories": []}
{"id": "2601.11126", "title": "Shape-morphing programming of soft materials on complex geometries via neural operator", "abstract": "Shape-morphing soft materials can enable diverse target morphologies through voxel-level material distribution design, offering significant potential for various applications. Despite progress in basic shape-morphing design with simple geometries, achieving advanced applications such as conformal implant deployment or aerodynamic morphing requires accurate and diverse morphing designs on complex geometries, which remains challenging. Here, we present a Spectral and Spatial Neural Operator (S2NO), which enables high-fidelity morphing prediction on complex geometries. S2NO effectively captures global and local morphing behaviours on irregular computational domains by integrating Laplacian eigenfunction encoding and spatial convolutions. Combining S2NO with evolutionary algorithms enables voxel-level optimisation of material distributions for shape morphing programming on various complex geometries, including irregular-boundary shapes, porous structures, and thin-walled structures. Furthermore, the neural operator's discretisation-invariant property enables super-resolution material distribution design, further expanding the diversity and complexity of morphing design. These advancements significantly improve the efficiency and capability of programming complex shape morphing.", "date": "2026-01-16", "categories": []}
{"id": "2602.00103", "title": "Autonomous Multi-Agent AI for High-Throughput Polymer Informatics: From Property Prediction to Generative Design Across Synthetic and Bio-Polymers", "abstract": "We present an integrated multiagent AI ecosystem for polymer discovery that unifies high-throughput materials workflows, artificial intelligence, and computational modeling within a single Polymer Research Lifecycle (PRL) pipeline. The system orchestrates specialized agents powered by state-of-the-art large language models (DeepSeek-V2 and DeepSeek-Coder) to retrieve and reason over scientific resources, invoke external tools, execute domain-specific code, and perform metacognitive self-assessment for robust end-to-end task execution. We demonstrate three practical capabilities: a high-fidelity polymer property prediction and generative design pipeline, a fully automated multimodal workflow for biopolymer structure characterization, and a metacognitive agent framework that can monitor performance and improve execution strategies over time. On a held-out test set of 1,251 polymers, our PolyGNN agent achieves strong predictive accuracy, reaching R2 = 0.89 for glass-transition temperature (Tg ), R2 = 0.82 for tensile strength, R2 = 0.75 for elongation, and R2 = 0.91 for density. The framework also provides uncertainty estimates via multiagent consensus and scales with linear complexity to at least 10,000 polymers, enabling high-throughput screening at low computational cost. For a representative workload, the system completes inference in 16.3 s using about 2 GB of memory and 0.1 GPU hours, at an estimated cost of about $0.08. On a dedicated Tg benchmark, our approach attains R2 = 0.78, outperforming strong baselines including single-LLM prediction (R2 = 0.67), group-contribution methods (R2 = 0.71), and ChemCrow (R2 = 0.66). We further demonstrate metacognitive control in a polystyrene case study, where the system not only produces domain-level scientific outputs but continually monitors and optimizes its own behavior through tactical, strategic, and meta-strategic self-assessment.", "date": "2026-01-25", "categories": []}
{"id": "2602.00756", "title": "A New Workflow for Materials Discovery Bridging the Gap Between Experimental Databases and Graph Neural Networks", "abstract": "Incorporating Machine Learning (ML) into material property prediction has become a crucial step in accelerating materials discovery. A key challenge is the severe lack of training data, as many properties are too complicated to calculate with high-throughput first principles techniques. To address this, recent research has created experimental databases from information extracted from scientific literature. However, most existing experimental databases do not provide full atomic coordinate information, which prevents them from supporting advanced ML architectures such as Graph Neural Networks (GNNs). In this work, we propose to bridge this gap through an alignment process between experimental databases and Crystallographic Information Files (CIF) from the Inorganic Crystal Structure Database (ICSD). Our approach enables the creation of a database that can fully leverage state-of-the-art model architectures for material property prediction. It also opens the door to utilizing transfer learning to improve prediction accuracy. To validate our approach, we align NEMAD with the ICSD and compare models trained on the resulting database to those trained on NEMAD originally. We demonstrate significant improvements in both Mean Absolute Error (MAE) and Correct Classification Rate (CCR) in predicting the ordering temperatures and magnetic ground states of magnetic materials, respectively.", "date": "2026-01-31", "categories": []}
{"id": "2602.01149", "title": "Robust Machine Learning Framework for Reliable Discovery of High-Performance Half-Heusler Thermoelectrics", "abstract": "Machine learning (ML) can facilitate efficient thermoelectric (TE) material discovery essential to address the environmental crisis. However, ML models often suffer from poor experimental generalizability despite high metrics. This study presents a robust workflow, applied to the half-Heusler (hH) structural prototype, for figure of merit (zT) prediction, to improve the generalizability of ML models. To resolve challenges in dataset handling and feature filtering, we first introduce a rigorous PCA-based splitting method that ensures training and test sets are unbiased and representative of the full chemical space. We then integrate Bayesian hyperparameter optimization with k-best feature filtering across three architectures-Random Forest, XGBoost, and Neural Networks - while employing SISSO symbolic regression for physical insight and comparison. Using SHAP and SISSO analysis, we identify A-site dopant concentration (xA'), and A-site Heat of Vaporization (HVA) as the primary drivers of zT besides Temperature (T). Finally, a high-throughput screening of approximately 6.6x10^8 potential compositions, filtered by stability constraints, yielded several novel high-zT candidates. Breaking from the traditional focus of improving test RMSE/R^2 values of the models, this work shifts the attention on establishing the test set a true proxy for model generalizability and strengthening the often neglected modules of the existing ML workflows for the data-driven design of next-generation thermoelectric materials.", "date": "2026-02-01", "categories": []}
{"id": "2602.01215", "title": "AI Meets Plasticity: A Comprehensive Survey", "abstract": "Artificial intelligence (AI) is rapidly emerging as a new paradigm of scientific discovery, namely data-driven science, across nearly all scientific disciplines. In materials science and engineering, AI has already begun to exert a transformative influence, making it both timely and necessary to examine its interaction with materials plasticity. In this study, we present a holistic survey of the convergence between AI and plasticity, highlighting state-of-the-art AI methodologies employed to discover, construct surrogate models for, and emulate the plastic behavior of materials. From a materials science perspective, we examine cause-and-effect relationships governing plastic deformation, including microstructural characterization and macroscopic responses described through plasticity constitutive models. From the perspective of AI methodology, we review a broad spectrum of applied approaches, ranging from frequentist techniques such as classical machine learning (ML), deep learning (DL), and physics-informed models to probabilistic frameworks that incorporate uncertainty quantification and generative AI methods. These data-driven approaches are discussed in the context of materials characterization and plasticity-related applications. The primary objective of this survey is to develop a comprehensive and well-organized taxonomy grounded in AI methodologies, with particular emphasis on distinguishing critical aspects of these techniques, including model architectures, data requirements, and predictive performance within the specific domain of materials plasticity. By doing so, this work aims to provide a clear road map for researchers and practitioners in the materials community, while offering deeper physical insight and intuition into the role of AI in advancing materials plasticity and characterization, an area of growing importance in the emerging AI-driven era.", "date": "2026-02-01", "categories": []}
{"id": "2503.01431", "title": "How simple can you go? An off-the-shelf transformer approach to molecular dynamics", "abstract": "Most current neural networks for molecular dynamics (MD) include physical inductive biases, resulting in specialized and complex architectures. This is in contrast to most other machine learning domains, where specialist approaches are increasingly replaced by general-purpose architectures trained on vast datasets. In line with this trend, several recent studies have questioned the necessity of architectural features commonly found in MD models, such as built-in rotational equivariance or energy conservation. In this work, we contribute to the ongoing discussion by evaluating the performance of an MD model with as few specialized architectural features as possible. We present a recipe for MD using an Edge Transformer, an ``off-the-shelf'' transformer architecture that has been minimally modified for the MD domain, termed MD-ET. Our model implements neither built-in equivariance nor energy conservation. We use a simple supervised pre-training scheme on $\\sim$30 million molecular structures from the QCML database. Using this ``off-the-shelf'' approach, we show state-of-the-art results on several benchmarks after fine-tuning for a small number of steps. Additionally, we examine the effects of being only approximately equivariant and energy conserving for MD simulations, proposing a novel method for distinguishing the errors resulting from non-equivariance from other sources of inaccuracies like numerical rounding errors. While our model exhibits runaway energy increases on larger structures, we show approximately energy-conserving NVE simulations for a range of small structures.", "date": "2025-03-03", "categories": []}
{"id": "2601.20996", "title": "MADE: Benchmark Environments for Closed-Loop Materials Discovery", "abstract": "Existing benchmarks for computational materials discovery primarily evaluate static predictive tasks or isolated computational sub-tasks. While valuable, these evaluations neglect the inherently iterative and adaptive nature of scientific discovery. We introduce MAterials Discovery Environments (MADE), a novel framework for benchmarking end-to-end autonomous materials discovery pipelines. MADE simulates closed-loop discovery campaigns in which an agent or algorithm proposes, evaluates, and refines candidate materials under a constrained oracle budget, capturing the sequential and resource-limited nature of real discovery workflows. We formalize discovery as a search for thermodynamically stable compounds relative to a given convex hull, and evaluate efficacy and efficiency via comparison to baseline algorithms. The framework is flexible; users can compose discovery agents from interchangeable components such as generative models, filters, and planners, enabling the study of arbitrary workflows ranging from fixed pipelines to fully agentic systems with tool use and adaptive decision making. We demonstrate this by conducting systematic experiments across a family of systems, enabling ablation of components in discovery pipelines, and comparison of how methods scale with system complexity.", "date": "2026-01-28", "categories": []}
{"id": "2601.21056", "title": "Better without U: Impact of Selective Hubbard U Correction on Foundational MLIPs", "abstract": "The training of foundational machine learning interatomic potentials (fMLIPs) relies on diverse databases with energies and forces calculated using ab initio methods. We show that fMLIPs trained on large datasets such as MPtrj, Alexandria, and OMat24 encode inconsistencies from the Materials Project's selective use of the Hubbard U correction, which is applied to certain transition metals only if O or F atoms are present in the simulation cell. This inconsistent use of +U creates two incompatible potential-energy surfaces (PES): a lower-energy GGA surface and a higher-energy GGA+U one. When trained on both, MLIPs interpolate between them, leading to systematic underbinding, or even spurious repulsion, between U-corrected metals and oxygen- or fluorine-containing species. Models such as MACE-OMAT and -MPA exhibit repulsion between U-corrected metals and their oxides, limiting their value for studying catalysis and oxidation. We link the severity of this pathology to the oxygen number density in U-corrected training configurations. This explains why OMAT-trained models are most affected and suggests the issue might worsen as expanding future datasets increasingly include configurations with low oxygen content, such as those generated through combinatorial exploration of multi-element or defect-containing systems.\n  Our simple per-U-corrected-atom shift aligns PBE+U and PBE energies for identical structures, yielding a smoother PES compared to existing correction schemes, which target phase diagram accuracy. As a result, models trained on datasets with our shift applied exhibit smaller mean absolute errors for the adsorption energies of oxygen on U-corrected elemental slabs. Since datasets omitting +U entirely (e.g. MatPES, MP-ALOE) avoid these pathologies, we recommend excluding +U in future fMLIP datasets. For existing datasets, our post-hoc correction provides a low-cost improvement.", "date": "2026-01-28", "categories": []}
{"id": "2601.21147", "title": "Smooth Dynamic Cutoffs for Machine Learning Interatomic Potentials", "abstract": "Machine learning interatomic potentials (MLIPs) have proven to be wildly useful for molecular dynamics simulations, powering countless drug and materials discovery applications. However, MLIPs face two primary bottlenecks preventing them from reaching realistic simulation scales: inference time and memory consumption. In this work, we address both issues by challenging the long-held belief that the cutoff radius for the MLIP must be held to a fixed, constant value. For the first time, we introduce a dynamic cutoff formulation that still leads to stable, long timescale molecular dynamics simulation. In introducing the dynamic cutoff, we are able to induce sparsity onto the underlying atom graph by targeting a specific number of neighbors per atom, significantly reducing both memory consumption and inference time. We show the effectiveness of a dynamic cutoff by implementing it onto 4 state of the art MLIPs: MACE, Nequip, Orbv3, and TensorNet, leading to 2.26x less memory consumption and 2.04x faster inference time, depending on the model and atomic system. We also perform an extensive error analysis and find that the dynamic cutoff models exhibit minimal accuracy dropoff compared to their fixed cutoff counterparts on both materials and molecular datasets. All model implementations and training code will be fully open sourced.", "date": "2026-01-29", "categories": []}
{"id": "2601.22009", "title": "MEIDNet: Multimodal generative AI framework for inverse materials design", "abstract": "In this work, we present Multimodal Equivariant Inverse Design Network (MEIDNet), a framework that jointly learns structural information and materials properties through contrastive learning, while encoding structures via an equivariant graph neural network (EGNN). By combining generative inverse design with multimodal learning, our approach accelerates the exploration of chemical-structural space and facilitates the discovery of materials that satisfy predefined property targets. MEIDNet exhibits strong latent-space alignment with cosine similarity 0.96 by fusion of three modalities through cross-modal learning. Through implementation of curriculum learning strategies, MEIDNet achieves ~60 times higher learning efficiency than conventional training techniques. The potential of our multimodal approach is demonstrated by generating low-bandgap perovskite structures at a stable, unique, and novel (SUN) rate of 13.6 %, which are further validated by ab initio methods. Our inverse design framework demonstrates both scalability and adaptability, paving the way for the universal learning of chemical space across diverse modalities.", "date": "2026-01-29", "categories": []}
{"id": "2508.13251", "title": "\"DIVE\" into Hydrogen Storage Materials Discovery with AI Agents", "abstract": "Data-driven artificial intelligence (AI) approaches are fundamentally transforming the discovery of new materials. Despite the unprecedented availability of materials data in the scientific literature, much of this information remains trapped in unstructured figures and tables, hindering the construction of large language model (LLM)-based AI agent for automated materials design. Here, we present the Descriptive Interpretation of Visual Expression (DIVE) multi-agent workflow, which systematically reads and organizes experimental data from graphical elements in scientific literatures. We focus on solid-state hydrogen storage materials-a class of materials central to future clean-energy technologies and demonstrate that DIVE markedly improves the accuracy and coverage of data extraction compared to the direct extraction by multimodal models, with gains of 10-15% over commercial models and over 30% relative to open-source models. Building on a curated database of over 30,000 entries from 4,000 publications, we establish a rapid inverse design workflow capable of identifying previously unreported hydrogen storage compositions in two minutes. The proposed AI workflow and agent design are broadly transferable across diverse materials, providing a paradigm for AI-driven materials discovery.", "date": "2025-08-18", "categories": []}
{"id": "2509.21778", "title": "Beyond Structure: Invariant Crystal Property Prediction with Pseudo-Particle Ray Diffraction", "abstract": "Crystal property prediction, governed by quantum mechanical principles, is computationally prohibitive to solve exactly for large many-body systems using traditional density functional theory. While machine learning models have emerged as efficient approximations for large-scale applications, their performance is strongly influenced by the choice of atomic representation. Although modern graph-based approaches have progressively incorporated more structural information, they often fail to capture long-range atomic interactions due to finite receptive fields and local encoding schemes. This limitation leads to distinct crystals being mapped to identical representations, hindering accurate property prediction. To address this, we introduce PRDNet that leverages unique reciprocal-space diffraction besides graph representations. To enhance sensitivity to elemental and environmental variations, we employ a data-driven pseudo-particle to generate a synthetic diffraction pattern. PRDNet ensures full invariance to crystallographic symmetries. Extensive experiments are conducted on Materials Project, JARVIS-DFT, and MatBench, demonstrating that the proposed model achieves state-of-the-art performance. The code is openly available at https://github.com/Bin-Cao/PRDNet.", "date": "2025-09-26", "categories": []}
{"id": "2506.08224", "title": "AI-Assisted Rapid Crystal Structure Generation Towards a Target Local Environment", "abstract": "In the field of material design, traditional crystal structure prediction approaches require extensive structural sampling through computationally expensive energy minimization methods using either force fields or quantum mechanical simulations. While emerging artificial intelligence (AI) generative models have shown great promise in generating realistic crystal structures more rapidly, most existing models fail to account for the unique symmetries and periodicity of crystalline materials, and they are limited to handling structures with only a few tens of atoms per unit cell. Here, we present a symmetry-informed AI generative approach called Local Environment Geometry-Oriented Crystal Generator (LEGO-xtal) that overcomes these limitations. Our method generates initial structures using AI models trained on an augmented small dataset, and then optimizes them using machine learning structure descriptors rather than traditional energy-based optimization. We demonstrate the effectiveness of LEGO-xtal by expanding from 25 known low-energy sp2 carbon allotropes to over 1,700, all within 0.5 eV/atom of the ground-state energy of graphite. This framework offers a generalizable strategy for the targeted design of materials with modular building blocks, such as metal-organic frameworks and next-generation battery materials.", "date": "2025-06-09", "categories": []}
{"id": "2509.21330", "title": "InSpecLearn4SDL: Interpretable Spectral Features Predict Conductivity in Self-Driving Doped Conjugated Polymer Labs", "abstract": "To accelerate materials discovery using self-driving labs (SDLs), we present a machine learning pipeline that predicts the electrical conductivity of doped conjugated polymers using rapid, non-destructive optical spectroscopy. Our approach automates spectral featurization by combining a genetic algorithm with adaptive area-under-the-curve (AUC) computations, creating a quantitative structure-property relationship (QSPR) that links optical response and processing parameters to conductivity. By incorporating SHAP-guided selection and domain-knowledge-based feature expansion, the model matches expert-curated performance while theoretically reducing experimental effort by $\\sim 33\\%$ by minimizing the need for costly direct conductivity measurements. Notably, the model recovers known physical descriptors in pBTTT and identifies informative tail-state regions correlated with polymer bleaching upon successful doping. This generic, interpretable, small-data-friendly methodology can be extended to other spectroscopic modalities, such as Raman or FTIR, providing a framework for autonomous decision-making in SDLs.", "date": "2025-09-06", "categories": []}
{"id": "2510.22345", "title": "Uncertainty quantification in model discovery by distilling interpretable material constitutive models from Gaussian process posteriors", "abstract": "Constitutive model discovery refers to the task of identifying an appropriate model structure, usually from a predefined model library, while simultaneously inferring its material parameters. The data used for model discovery are measured in mechanical tests and are thus inevitably affected by noise which, in turn, induces uncertainties. Previously proposed methods for uncertainty quantification in model discovery either require the selection of a prior for the material parameters, are restricted to linear coefficients of the model library or are limited in the flexibility of the inferred parameter probability distribution. We therefore propose a partially Bayesian framework for uncertainty quantification in model discovery that does not require prior selection for the material parameters and also allows for the discovery of constitutive models with inner-non-linear parameters: First, we augment the available stress-deformation data with a Gaussian process. Second, we approximate the parameter distribution by a normalizing flow, which allows for modeling complex joint distributions. Third, we distill the parameter distribution by matching the distribution of stress-deformation functions induced by the parameters with the Gaussian process posterior. Fourth, we perform a Sobol' sensitivity analysis to obtain a sparse and interpretable model. We demonstrate the capability of our framework for both isotropic and experimental anisotropic data.", "date": "2025-10-25", "categories": []}
{"id": "2512.19494", "title": "Kolmogorov-Arnold graph neural networks for chemically informed prediction tasks on inorganic nanomaterials", "abstract": "The recent development of Kolmogorov-Arnold Networks (KANs) has found its application in the field of Graph Neural Networks (GNNs) particularly in molecular data modeling and potential drug discovery. Kolmogorov-Arnold Graph Neural Networks (KAGNNs) expand on the existing set of GNN models with KAN-based counterparts. KAGNNs have been demonstrably successful in surpassing the accuracy of MultiLayer Perceptron (MLP)-based GNNs in the task of molecular property prediction. These models were widely tested on the graph datasets consisting of organic molecules. In this study, we explore the application of KAGNNs towards inorganic nanomaterials. In 2024, a large scale inorganic nanomaterials dataset was published under the title CHILI (Chemically-Informed Large-scale Inorganic Nanomaterials Dataset), and various MLP-based GNNs have been tested on this dataset. We adapt and test our own KAGNNs appropriate for eight defined tasks. Our experiments reveal that, KAGNNs frequently surpass the performance of their counterpart GNNs. Most notably, on crystal system and space group classification tasks in CHILI-3K, KAGNNs achieve the new state-of-the-art results of 99.5 percent and 96.6 percent accuracy, respectively, compared to the previous 65.7 and 73.3 percent each.", "date": "2025-12-22", "categories": []}
{"id": "2505.19350", "title": "FlashMD: long-stride, universal prediction of molecular dynamics", "abstract": "Molecular dynamics (MD) provides insights into atomic-scale processes by integrating over time the equations that describe the motion of atoms under the action of interatomic forces. Machine learning models have substantially accelerated MD by providing inexpensive predictions of the forces, but they remain constrained to minuscule time integration steps, which are required by the fast time scale of atomic motion. In this work, we propose FlashMD, a method to predict the evolution of positions and momenta over strides that are between one and two orders of magnitude longer than typical MD time steps. We incorporate considerations on the mathematical and physical properties of Hamiltonian dynamics in the architecture, generalize the approach to allow the simulation of any thermodynamic ensemble, and carefully assess the possible failure modes of such a long-stride MD approach. We validate FlashMD's accuracy in reproducing equilibrium and time-dependent properties, using both system-specific and general-purpose models, extending the ability of MD simulation to reach the long time scales needed to model microscopic processes of high scientific and technological relevance.", "date": "2025-05-25", "categories": []}
{"id": "2507.05024", "title": "A Generative Diffusion Model for Amorphous Materials", "abstract": "Generative models show great promise for the inverse design of molecules and inorganic crystals, but remain largely ineffective within more complex structures such as amorphous materials. Here, we present a diffusion model that reliably generates amorphous structures up to 1000 times faster than conventional simulations across processing conditions, compositions, and data sources. Generated structures recovered the short- and medium-range order, sampling diversity, and macroscopic properties of silica glass, as validated by simulations and an information-theoretical strategy. Conditional generation allowed sampling large structures at low cooling rates of 10$^{-2}$ K/ps to uncover a ductile-to-brittle transition and mesoporous silica structures. Extension to metallic glassy systems accurately reproduced local structures and properties from both computational and experimental datasets, demonstrating how synthetic data can be generated from characterization results. Our methods provide a roadmap for the design and simulation of amorphous materials previously inaccessible to computational methods.", "date": "2025-07-07", "categories": []}
{"id": "2601.12554", "title": "Artificial Intelligence in Materials Science and Engineering: Current Landscape, Key Challenges, and Future Trajectorie", "abstract": "Artificial Intelligence is rapidly transforming materials science and engineering, offering powerful tools to navigate complexity, accelerate discovery, and optimize material design in ways previously unattainable. Driven by the accelerating pace of algorithmic advancements and increasing data availability, AI is becoming an essential competency for materials researchers. This review provides a comprehensive and structured overview of the current landscape, synthesizing recent advancements and methodologies for materials scientists seeking to effectively leverage these data-driven techniques. We survey the spectrum of machine learning approaches, from traditional algorithms to advanced deep learning architectures, including CNNs, GNNs, and Transformers, alongside emerging generative AI and probabilistic models such as Gaussian Processes for uncertainty quantification. The review also examines the pivotal role of data in this field, emphasizing how effective representation and featurization strategies, spanning compositional, structural, image-based, and language-inspired approaches, combined with appropriate preprocessing, fundamentally underpin the performance of machine learning models in materials research. Persistent challenges related to data quality, quantity, and standardization, which critically impact model development and application in materials science and engineering, are also addressed.", "date": "2026-01-18", "categories": []}
{"id": "2501.05903", "title": "Discovery of sustainable energy materials via the machine-learned material space", "abstract": "Does a machine learning model actually gain an understanding of the material space? We answer this question in the affirmative on the example of the OptiMate model, a graph attention network trained to predict the optical properties of semiconductors and insulators. By applying the UMAP dimensionality reduction technique to its latent embeddings, we demonstrate that the model captures a nuanced and interpretable representation of the materials space, reflecting chemical and physical principles, without any user-induced bias. This enables clustering of almost 10,000 materials based on optical properties and chemical similarities. Beyond this understanding, we demonstrate how the learned material space can be used to identify more sustainable alternatives to critical materials in energy-related technologies, such as photovoltaics. These findings demonstrate the dual utility of machine learning models in materials science: Accurately predicting material properties while providing insights into the underlying materials space. The approach demonstrates the broader potential of leveraging learned materials spaces for the discovery and design of materials for diverse applications, and is easily applicable to any state-of-the-art machine learning model.", "date": "2025-01-10", "categories": []}
{"id": "2504.02367", "title": "Reinforcement Fine-Tuning for Materials Design", "abstract": "Reinforcement fine-tuning played an instrumental role in enhancing the instruction-following and reasoning abilities of large language models. In this work, we employ reinforcement fine-tuning for materials design, in which discriminative machine learning models are used to provide rewards to the autoregressive transformer-based materials generative model CrystalFormer. By optimizing the reward signals-such as energy above the convex hull and material properties figures of merit-reinforcement fine-tuning infuses knowledge from discriminative models into generative models. The resulting model, CrystalFormer-RL, shows enhanced stability in generated crystals and successfully discovers crystals with desirable yet conflicting material properties, such as substantial dielectric constant and band gap simultaneously. Notably, we observe that reinforcement fine-tuning not only enables the property-guided material design but also unlocks property-based material retrieval behavior of pretrained generative model. The present framework opens an exciting gateway to the synergies of the machine learning ecosystem for materials design.", "date": "2025-04-03", "categories": []}
{"id": "2508.14022", "title": "BLIPs: Bayesian Learned Interatomic Potentials", "abstract": "Machine Learning Interatomic Potentials (MLIPs) are becoming a central tool in simulation-based chemistry. However, like most deep learning models, MLIPs struggle to make accurate predictions on out-of-distribution data or when trained in a data-scarce regime, both common scenarios in simulation-based chemistry. Moreover, MLIPs do not provide uncertainty estimates by construction, which are fundamental to guide active learning pipelines and to ensure the accuracy of simulation results compared to quantum calculations. To address this shortcoming, we propose BLIPs: Bayesian Learned Interatomic Potentials. BLIP is a scalable, architecture-agnostic variational Bayesian framework for training or fine-tuning MLIPs, built on an adaptive version of Variational Dropout. BLIP delivers well-calibrated uncertainty estimates and minimal computational overhead for energy and forces prediction at inference time, while integrating seamlessly with (equivariant) message-passing architectures. Empirical results on simulation-based computational chemistry tasks demonstrate improved predictive accuracy with respect to standard MLIPs, and trustworthy uncertainty estimates, especially in data-scarse or heavy out-of-distribution regimes. Moreover, fine-tuning pretrained MLIPs with BLIP yields consistent performance gains and calibrated uncertainties.", "date": "2025-08-19", "categories": []}
{"id": "2512.04562", "title": "LeMat-GenBench: A Unified Evaluation Framework for Crystal Generative Models", "abstract": "Generative machine learning (ML) models hold great promise for accelerating materials discovery through the inverse design of inorganic crystals, enabling an unprecedented exploration of chemical space. Yet, the lack of standardized evaluation frameworks makes it challenging to evaluate, compare, and further develop these ML models meaningfully. In this work, we introduce LeMat-GenBench, a unified benchmark for generative models of crystalline materials, supported by a set of evaluation metrics designed to better inform model development and downstream applications. We release both an open-source evaluation suite and a public leaderboard on Hugging Face, and benchmark 12 recent generative models. Results reveal that an increase in stability leads to a decrease in novelty and diversity on average, with no model excelling across all dimensions. Altogether, LeMat-GenBench establishes a reproducible and extensible foundation for fair model comparison and aims to guide the development of more reliable, discovery-oriented generative models for crystalline materials.", "date": "2025-12-04", "categories": []}
{"id": "2601.08970", "title": "Machine Learning-Driven Creep Law Discovery Across Alloy Compositional Space", "abstract": "Hihg-temperature creep characterization of structural alloys traditionally relies on serial uniaxial tests, which are highly inefficient for exploring the large search space of alloy compositions and for material discovery. Here, we introduce a machine-learning-assisted, high-throughput framework for creep law identification based on a dimple array bulge instrument (DABI) configuration, which enables parallel creep testing of 25 dimples, each fabricated from a different alloy, in a single experiment. Full-field surface displacements of dimples undergoing time-dependent creep-induced bulging under inert gas pressure are measured by 3D digital image correlation. We train a recurrent neural network (RNN) as a surrogate model, mapping creep parameters and loading conditions to the time-dependent deformation response of DABI. Coupling this surrogate with a particle swarm optimization scheme enables rapid and global inverse identification with sparsity regularization of creep parameters from experiment displacement-time histories. In addition, we propose a phenomenological creep law with a time-dependent stress exponent that captures the sigmoidal primary creep observed in wrought INCONEL 625 and extracts its temperature dependence from DABI test at multiple temperatures. Furthermore, we employ a general creep law combining several conventional forms together with regularized inversion to identify the creep laws for 47 additional Fe-, Ni-, and Co-rich alloys and to automatically select the dominant functional form for each alloy. This workflow combined with DABI experiment provides a quantitative, high-throughput creep characterization platform that is compatible with data mining, composition-property modeling, and nonlinear structural optimization with creep behavior across a large alloy design space.", "date": "2026-01-13", "categories": []}
{"id": "2601.07966", "title": "DataScribe: An AI-Native, Policy-Aligned Web Platform for Multi-Objective Materials Design and Discovery", "abstract": "The acceleration of materials discovery requires digital platforms that go beyond data repositories to embed learning, optimization, and decision-making directly into research workflows. We introduce DataScribe, an AI-native, cloud-based materials discovery platform that unifies heterogeneous experimental and computational data through ontology-backed ingestion and machine-actionable knowledge graphs. The platform integrates FAIR-compliant metadata capture, schema and unit harmonization, uncertainty-aware surrogate modeling, and native multi-objective multi-fidelity Bayesian optimization, enabling closed-loop propose-measure-learn workflows across experimental and computational pipelines. DataScribe functions as an application-layer intelligence stack, coupling data governance, optimization, and explainability rather than treating them as downstream add-ons. We validate the platform through case studies in electrochemical materials and high-entropy alloys, demonstrating end-to-end data fusion, real-time optimization, and reproducible exploration of multi-objective trade spaces. By embedding optimization engines, machine learning, and unified access to public and private scientific data directly within the data infrastructure, and by supporting open, free use for academic and non-profit researchers, DataScribe functions as a general-purpose application-layer backbone for laboratories of any scale, including self-driving laboratories and geographically distributed materials acceleration platforms, with built-in support for performance, sustainability, and supply-chain-aware objectives.", "date": "2026-01-12", "categories": []}
{"id": "2601.08185", "title": "Autonomous Materials Exploration by Integrating Automated Phase Identification and AI-Assisted Human Reasoning", "abstract": "Autonomous experimentation holds the potential to accelerate materials development by combining artificial intelligence (AI) with modular robotic platforms to explore extensive combinatorial chemical and processing spaces. Such self-driving laboratories can not only increase the throughput of repetitive experiments, but also incorporate human domain expertise to drive the search towards user-defined objectives, including improved materials performance metrics. We present an autonomous materials synthesis extension to SARA, the Scientific Autonomous Reasoning Agent, utilizing phase information provided by an automated probabilistic phase labeling algorithm to expedite the search for targeted phase regions. By incorporating human input into an expanded SARA-H (SARA with human-in-the-loop) framework, we enhance the efficiency of the underlying reasoning process. Using synthetic benchmarks, we demonstrate the efficiency of our AI implementation and show that the human input can contribute to significant improvement in sampling efficiency. We conduct experimental active learning campaigns using robotic processing of thin-film samples of several oxide material systems, including Bi$_2$O$_3$, SnO$_x$, and Bi-Ti-O, using lateral-gradient laser spike annealing to synthesize and kinetically trap metastable phases. We showcase the utility of human-in-the-loop autonomous experimentation for the Bi-Ti-O system, where we identify extensive processing domains that stabilize $\\delta$-Bi$_2$O$_3$ and Bi$_2$Ti$_2$O$_7$, explore dwell-dependent ternary oxide phase behavior, and provide evidence confirming predictions that cationic substitutional doping of TiO$_2$ with Bi inhibits the unfavorable transformation of the metastable anatase to the ground-state rutile phase. The autonomous methods we have developed enable the discovery of new materials and new understanding of materials synthesis and properties.", "date": "2026-01-13", "categories": []}
{"id": "2601.04606", "title": "Crystal Generation using the Fully Differentiable Pipeline and Latent Space Optimization", "abstract": "We present a materials generation framework that couples a symmetry-conditioned variational autoencoder (CVAE) with a differentiable SO(3) power spectrum objective to steer candidates toward a specified local environment under the crystallographic constraints. In particular, we implement a fully differentiable pipeline to enable batch-wise optimization on both direct and latent crystallographic representations. Using the GPU acceleration, this implementation achieves about fivefold speed compared to our previous CPU workflow, while yielding comparable outcomes. In addition, we introduce the optimization strategy that alternatively performs optimization on the direct and latent crystal representations. This dual-level relaxation approach can effectively escape local minima defined by different objective gradients, thus increasing the success rate of generating complex structures satisfying the target local environments. This framework can be extended to systems consisting of multi-components and multi-environments, providing a scalable route to generate material structures with the target local environment.", "date": "2026-01-08", "categories": []}
{"id": "2601.07583", "title": "Machine learning nonequilibrium phase transitions in charge-density wave insulators", "abstract": "Nonequilibrium electronic forces play a central role in voltage-driven phase transitions but are notoriously expensive to evaluate in dynamical simulations. Here we develop a machine learning framework for adiabatic lattice dynamics coupled to nonequilibrium electrons, and demonstrate it for a gating induced insulator to metal transition out of a charge density wave state in the Holstein model. Although exact electronic forces can be obtained from nonequilibrium Green's function (NEGF) calculations, their high computational cost renders long time dynamical simulations prohibitively expensive. By exploiting the locality of the electronic response, we train a neural network to directly predict instantaneous local electronic forces from the lattice configuration, thereby bypassing repeated NEGF calculations during time evolution. When combined with Brownian dynamics, the resulting machine learning force field quantitatively reproduces domain wall motion and nonequilibrium phase transition dynamics obtained from full NEGF simulations, while achieving orders of magnitude gains in computational efficiency. Our results establish direct force learning as an efficient and accurate approach for simulating nonequilibrium lattice dynamics in driven quantum materials.", "date": "2026-01-12", "categories": []}
{"id": "2505.23064", "title": "Machine Learning Framework for Characterizing Processing-Structure Relationship in Block Copolymer Thin Films", "abstract": "The morphology of block copolymers (BCPs) critically influences material properties and applications. This work introduces a machine learning (ML)-enabled, high-throughput framework for analyzing grazing incidence small-angle X-ray scattering (GISAXS) data and atomic force microscopy (AFM) images to characterize BCP thin film morphology. A convolutional neural network was trained to classify AFM images by morphology type, achieving 97% testing accuracy. Classified images were then analyzed to extract 2D grain size measurements from the samples in a high-throughput manner. ML models were developed to predict morphological features based on processing parameters such as solvent ratio, additive type, and additive ratio. GISAXS-based properties were predicted with strong performances ($R^2$ > 0.75), while AFM-based property predictions were less accurate ($R^2$ < 0.60), likely due to the localized nature of AFM measurements compared to the bulk information captured by GISAXS. Beyond model performance, interpretability was addressed using Shapley Additive exPlanations (SHAP). SHAP analysis revealed that the additive ratio had the largest impact on morphological predictions, where additive provides the BCP chains with increased volume to rearrange into thermodynamically favorable morphologies. This interpretability helps validate model predictions and offers insight into parameter importance. Altogether, the presented framework combining high-throughput characterization and interpretable ML offers an approach to exploring and optimizing BCP thin film morphology across a broad processing landscape.", "date": "2025-05-29", "categories": []}
{"id": "2508.16012", "title": "FIRE-GNN: Force-informed, Relaxed Equivariance Graph Neural Network for Rapid and Accurate Prediction of Surface Properties", "abstract": "The work function and cleavage energy of a surface are critical properties that determine the viability of materials in electronic emission applications, semiconductor devices, and heterogeneous catalysis. While first principles calculations are accurate in predicting these properties, their computational expense combined with the vast search space of surfaces make a comprehensive screening approach with density functional theory (DFT) infeasible. Here, we introduce FIRE-GNN (Force-Informed, Relaxed Equivariance Graph Neural Network), which integrates surface-normal symmetry breaking and machine learning interatomic potential (MLIP)-derived force information, achieving a twofold reduction in mean absolute error (down to 0.065 eV) over the previous state-of-the-art for work function prediction. We additionally benchmark recent invariant and equivariant architectures, analyze the impact of symmetry breaking, and evaluate out-of-distribution generalization, demonstrating that FIRE-GNN consistently outperforms competing models for work function predictions. This model enables accurate and rapid predictions of the work function and cleavage energy across a vast chemical space and facilitates the discovery of materials with tuned surface properties", "date": "2025-08-22", "categories": []}
{"id": "2601.05528", "title": "Autonomous Probe Microscopy with Robust Bag-of-Features Multi-Objective Bayesian Optimization: Pareto-Front Mapping of Nanoscale Structure-Property Trade-Offs", "abstract": "Combinatorial materials libraries are an efficient route to generate large families of candidate compositions, but their impact is often limited by the speed and depth of characterization and by the difficulty of extracting actionable structure-property relations from complex characterization data. Here we develop an autonomous scanning probe microscopy (SPM) framework that integrates automated atomic force and magnetic force microscopy (AFM/MFM) to rapidly explore magnetic and structural properties across combinatorial spread libraries. To enable automated exploration of systems without a clear optimization target, we introduce a combination of a static physics-informed bag-of-features (BoF) representation of measured surface morphology and magnetic structure with multi-objective Bayesian optimization (MOBO) to discover the relative significance and robustness of features. The resulting closed-loop workflow selectively samples the compositional gradient and reconstructs feature landscapes consistent with dense grid \"ground truth\" measurements. The resulting Pareto structure reveals where multiple nanoscale objectives are simultaneously optimized, where trade-offs between roughness, coherence, and magnetic contrast are unavoidable, and how families of compositions cluster into distinct functional regimes, thereby turning multi-feature imaging data into interpretable maps of competing structure-property trends. While demonstrated for Au-Co-Ni and AFM/MFM, the approach is general and can be extended to other combinatorial systems, imaging modalities, and feature sets, illustrating how feature-based MOBO and autonomous SPM can transform microscopy images from static data products into active feedback for real-time, multi-objective materials discovery.", "date": "2026-01-09", "categories": []}
{"id": "2601.03774", "title": "Scalable Machine Learning Force Fields for Macromolecular Systems Through Long-Range Aware Message Passing", "abstract": "Machine learning force fields (MLFFs) have revolutionized molecular simulations by providing quantum mechanical accuracy at the speed of molecular mechanical computations. However, a fundamental reliance of these models on fixed-cutoff architectures limits their applicability to macromolecular systems where long-range interactions dominate. We demonstrate that this locality constraint causes force prediction errors to scale monotonically with system size, revealing a critical architectural bottleneck. To overcome this, we establish the systematically designed MolLR25 ({Mol}ecules with {L}ong-{R}ange effect) benchmark up to 1200 atoms, generated using high-fidelity DFT, and introduce E2Former-LSR, an equivariant transformer that explicitly integrates long-range attention blocks. E2Former-LSR exhibits stable error scaling, achieves superior fidelity in capturing non-covalent decay, and maintains precision on complex protein conformations. Crucially, its efficient design provides up to 30% speedup compared to purely local models. This work validates the necessity of non-local architectures for generalizable MLFFs, enabling high-fidelity molecular dynamics for large-scale chemical and biological systems.", "date": "2026-01-07", "categories": []}
{"id": "2601.03801", "title": "Physically Consistent Machine Learning for Melting Temperature Prediction of Refractory High-Entropy Alloys", "abstract": "Predicting the melting temperature (Tm) of multi-component and high-entropy alloys (HEAs) is critical for high-temperature applications but computationally expensive using traditional CALPHAD or DFT methods. In this work, we develop a gradient-boosted decision tree (XGBoost) model to predict Tm for complex alloys based on elemental properties. To ensure physical consistency, we address the issue of data leakage by excluding temperature-dependent thermodynamic descriptors (such as Gibbs free energy of mixing) and instead rely on physically motivated elemental features. The optimized model achieves a coefficient of determination (R2) of 0.948 and a Mean Squared Error (MSE) of 9928 which is about 5% relative error for HEAs on a validation set of approximately 1300 compositions. Crucially, we validate the model using the Valence Electron Concentration (VEC) rule. Without explicit constraints during training, the model successfully captures the known stability transition between BCC and FCC phases at a VEC of approximately 6.87. These results demonstrate that data-driven models, when properly feature-engineered, can capture fundamental metallurgical principles for rapid alloy screening.", "date": "2026-01-07", "categories": []}
{"id": "2503.18975", "title": "Machine Learning - Driven Materials Discovery: Unlocking Next-Generation Functional Materials - A review", "abstract": "The rapid advancement of machine learning and artificial intelligence (AI)-driven techniques is revolutionizing materials discovery, property prediction, and material design by minimizing human intervention and accelerating scientific progress. This review provides a comprehensive overview of smart, machine learning (ML)-driven approaches, emphasizing their role in predicting material properties, discovering novel compounds, and optimizing material structures. Key methodologies in this field include deep learning, graph neural networks, Bayesian optimization, and automated generative models (GANs, VAEs). These approaches enable the autonomous design of materials with tailored functionalities. By leveraging AutoML frameworks (AutoGluon, TPOT, and H2O.ai), researchers can automate the model selection, hyperparameter tuning, and feature engineering, significantly improving the efficiency of materials informatics. Furthermore, the integration of AI-driven robotic laboratories and high-throughput computing has established a fully automated pipeline for rapid synthesis and experimental validation, drastically reducing the time and cost of material discovery. This review highlights real-world applications of automated ML-driven approaches in predicting mechanical, thermal, electrical, and optical properties of materials, demonstrating successful cases in superconductors, catalysts, photovoltaics, and energy storage systems. We also address key challenges, such as data quality, interpretability, and the integration of AutoML with quantum computing, which are essential for future advancements. Ultimately, combining AI with automated experimentation and computational modeling is transforming the way materials are discovered and optimized. This synergy paves the way for new innovations in energy, electronics, and nanotechnology.", "date": "2025-03-22", "categories": []}
{"id": "2601.00855", "title": "Physically-Constrained Autoencoder-Assisted Bayesian Optimization for Refinement of High-Dimensional Defect-Sensitive Single Crystalline Structure", "abstract": "Physical properties and functionalities of materials are dictated by global crystal structures as well as local defects. To establish a structure-property relationship, not only the crystallographic symmetry but also quantitative knowledge about defects are required. Here we present a hybrid Machine Learning framework that integrates a physically-constrained variational-autoencoder (pcVAE) with different Bayesian Optimization (BO) methods to systematically accelerate and improve crystal structure refinement with resolution of defects. We chose the pyrochlore structured Ho2Ti2O7 as a model system and employed the GSAS2 package for benchmarking crystallographic parameters from Rietveld refinement. However, the function space of these material systems is highly nonlinear, which limits optimizers like traditional Rietveld refinement, into trapping at local minima. Also, these naive methods don't provide an extensive learning about the overall function space, which is essential for large space, large time consuming explorations to identify various potential regions of interest. Thus, we present the approach of exploring the high Dimensional structure parameters of defect sensitive systems via pretrained pcVAE assisted BO and Sparse Axis Aligned BO. The pcVAE projects high-Dimensional diffraction data consisting of thousands of independently measured diffraction orders into a lowD latent space while enforcing scaling invariance and physical relevance. Then via BO methods, we aim to minimize the L2 norm based chisq errors in the real and latent spaces separately between experimental and simulated diffraction patterns, thereby steering the refinement towards potential optimum crystal structure parameters. We investigated and compared the results among different pcVAE assisted BO, non pcVAE assisted BO, and Rietveld refinement.", "date": "2025-12-29", "categories": []}
{"id": "2509.07579", "title": "Homogenization with Guaranteed Bounds via Primal-Dual Physically Informed Neural Networks", "abstract": "Physics-informed neural networks (PINNs) have shown promise in solving partial differential equations (PDEs) relevant to multiscale modeling, but they often fail when applied to materials with discontinuous coefficients, such as media with piecewise constant properties. This paper introduces a dual formulation for the PINN framework to improve the reliability of the homogenization of periodic thermo-conductive composites, for both strong and variational (weak) formulations. The dual approach facilitates the derivation of guaranteed upper and lower error bounds, enabling more robust detection of PINN failure. We compare standard PINNs applied to smoothed material approximations with variational PINNs (VPINNs) using both spectral and neural network-based test functions. Our results indicate that while strong-form PINNs may outperform VPINNs in controlled settings, they are sensitive to material discontinuities and may fail without clear diagnostics. In contrast, VPINNs accommodate piecewise constant material parameters directly but require careful selection of test functions to avoid instability. Dual formulation serves as a reliable indicator of convergence quality, and its integration into PINN frameworks enhances their applicability to homogenization problems in micromechanics.", "date": "2025-09-09", "categories": []}
{"id": "2508.16614", "title": "CrystalDiT: A Diffusion Transformer for Crystal Generation", "abstract": "We present CrystalDiT, a diffusion transformer for crystal structure generation that achieves state-of-the-art performance by challenging the trend of architectural complexity. Instead of intricate, multi-stream designs, CrystalDiT employs a unified transformer that imposes a powerful inductive bias: treating lattice and atomic properties as a single, interdependent system. Combined with a periodic table-based atomic representation and a balanced training strategy, our approach achieves 8.78% SUN (Stable, Unique, Novel) rate on MP-20, substantially outperforming recent methods including FlowMM (4.21%) and MatterGen (3.66%). Notably, CrystalDiT generates 63.28% unique and novel structures while maintaining comparable stability rates, demonstrating that architectural simplicity can be more effective than complexity for materials discovery. Our results suggest that in data-limited scientific domains, carefully designed simple architectures outperform sophisticated alternatives that are prone to overfitting.", "date": "2025-08-13", "categories": []}
{"id": "2512.17703", "title": "Revisiting the Broken Symmetry Phase of Solid Hydrogen: A Neural Network Variational Monte Carlo Study", "abstract": "The crystal structure of high-pressure solid hydrogen remains a fundamental open problem. Although the research frontier has mostly shifted toward ultra-high pressure phases above 400 GPa, we show that even the broken symmetry phase observed around 130~GPa requires revisiting due to its intricate coupling of electronic and nuclear degrees of freedom. Here, we develop a first principle quantum Monte Carlo framework based on a deep neural network wave function that treats both electrons and nuclei quantum mechanically within the constant pressure ensemble. Our calculations reveal an unreported ground-state structure candidate for the broken symmetry phase with $Cmcm$ space group symmetry, and we test its stability up to 96 atoms. The predicted structure quantitatively matches the experimental equation of state and X-ray diffraction patterns. Furthermore, our group-theoretical analysis shows that the $Cmcm$ structure is compatible with existing Raman and infrared spectroscopic data. Crucially, static density functional theory calculation reveals the $Cmcm$ structure as a dynamically unstable saddle point on the Born-Oppenheimer potential energy surface, demonstrating that a full quantum many-body treatment of the problem is necessary. These results shed new light on the phase diagram of high-pressure hydrogen and call for further experimental verifications.", "date": "2025-12-19", "categories": []}
{"id": "2512.22397", "title": "Integrating Wide and Deep Neural Networks with Squeeze-and-Excitation Blocks for Multi-Target Property Prediction in Additively Manufactured Fiber Reinforced Composites", "abstract": "Continuous fiber-reinforced composite manufactured by additive manufacturing (CFRC-AM) offers opportunities for printing lightweight materials with high specific strength. However, their performance is sensitive to the interaction of process and material parameters, making exhaustive experimental testing impractical. In this study, we introduce a data-efficient, multi-input, multi-target learning approach that integrates Latin Hypercube Sampling (LHS)-guided experimentation with a squeeze-and-excitation wide and deep neural network (SE-WDNN) to jointly predict multiple mechanical and manufacturing properties of CFRC-AMs based on different manufacturing parameters. We printed and tested 155 specimens selected from a design space of 4,320 combinations using a Markforged Mark Two 3D printer. The processed data formed the input-output set for our proposed model. We compared the results with those from commonly used machine learning models, including feedforward neural networks, Kolmogorov-Arnold networks, XGBoost, CatBoost, and random forests. Our model achieved the lowest overall test error (MAPE = 12.33%) and showed statistically significant improvements over the baseline wide and deep neural network for several target variables (paired t-tests, p <= 0.05). SHapley Additive exPlanations (SHAP) analysis revealed that reinforcement strategy was the major influence on mechanical performance. Overall, this study demonstrates that the integration of LHS and SE-WDNN enables interpretable and sample-efficient multi-target predictions, guiding parameter selection in CFRC-AM with a balance between mechanical behavior and manufacturing metrics.", "date": "2025-12-26", "categories": []}
{"id": "2512.22529", "title": "Multi-AI Agent Framework Reveals the \"Oxide Gatekeeper\" in Aluminum Nanoparticle Oxidation", "abstract": "Aluminum nanoparticles (ANPs) are among the most energy-dense solid fuels, yet the atomic mechanisms governing their transition from passivated particles to explosive reactants remain elusive. This stems from a fundamental computational bottleneck: ab initio methods offer quantum accuracy but are restricted to small spatiotemporal scales (< 500 atoms, picoseconds), while empirical force fields lack the reactive fidelity required for complex combustion environments. Herein, we bridge this gap by employing a \"human-in-the-loop\" closed-loop framework where self-auditing AI Agents validate the evolution of a machine learning potential (MLP). By acting as scientific sentinels that visualize hidden model artifacts for human decision-making, this collaborative cycle ensures quantum mechanical accuracy while exhibiting near-linear scalability to million-atom systems and accessing nanosecond timescales (energy RMSE: 1.2 meV/atom, force RMSE: 0.126 eV/Angstrom). Strikingly, our simulations reveal a temperature-regulated dual-mode oxidation mechanism: at moderate temperatures, the oxide shell acts as a dynamic \"gatekeeper,\" regulating oxidation through a \"breathing mode\" of transient nanochannels; above a critical threshold, a \"rupture mode\" unleashes catastrophic shell failure and explosive combustion. Importantly, we resolve a decades-old controversy by demonstrating that aluminum cation outward diffusion, rather than oxygen transport, dominates mass transfer across all temperature regimes, with diffusion coefficients consistently exceeding those of oxygen by 2-3 orders of magnitude. These discoveries establish a unified atomic-scale framework for energetic nanomaterial design, enabling the precision engineering of ignition sensitivity and energy release rates through intelligent computational design.", "date": "2025-12-27", "categories": []}
{"id": "2512.19732", "title": "Leakage-Aware Bandgap Prediction on the JARVIS-DFT Dataset: A Phase-Wise Feature Analysis", "abstract": "In this study, we perform a systematic analysis of the JARVIS-DFT bandgap dataset and identify and remove descriptors that may inadvertently encode band-structure information, such as effective masses. This process yields a curated, leakage-controlled subset of 2280 materials. Using this dataset, a three-phase modeling framework is implemented that incrementally incorporates basic physical descriptors, engineered features, and compositional attributes. The results show that tree-based models achieve R2 values of approximately 0.88 to 0.90 across all phases, indicating that expanding the descriptor space does not substantially improve predictive accuracy when leakage is controlled. SHAP analysis consistently identifies the dielectric tensor components as the dominant contributors. This work provides a curated dataset and baseline performance metrics for future leakage-aware bandgap prediction studies.", "date": "2025-12-17", "categories": []}
{"id": "2512.18029", "title": "Long-range electrostatics for machine learning interatomic potentials is easier than we thought", "abstract": "The lack of long-range electrostatics is a key limitation of modern machine learning interatomic potentials (MLIPs), hindering reliable applications to interfaces, charge-transfer reactions, polar and ionic materials, and biomolecules. In this Perspective, we distill two design principles behind the Latent Ewald Summation (LES) framework, which can capture long-range interactions, charges, and electrical response just by learning from standard energy and force training data: (i) use a Coulomb functional form with environment-dependent charges to capture electrostatic interactions, and (ii) avoid explicit training on ambiguous density functional theory (DFT) partial charges. When both principles are satisfied, substantial flexibility remains: essentially any short-range MLIP can be augmented; charge equilibration schemes can be added when desired; dipoles and Born effective charges can be inferred or finetuned; and charge/spin-state embeddings or tensorial targets can be further incorporated. We also discuss current limitations and open challenges. Together, these minimal, physics-guided design rules suggest that incorporating long-range electrostatics into MLIPs is simpler and perhaps more broadly applicable than is commonly assumed.", "date": "2025-12-19", "categories": []}
{"id": "2512.18251", "title": "CrystalFormer-CSP: Thinking Fast and Slow for Crystal Structure Prediction", "abstract": "Crystal structure prediction is a fundamental problem in materials science. We present CrystalFormer-CSP, an efficient framework that unifies data-driven heuristic and physics-driven optimization approaches to predict stable crystal structures for given chemical compositions. The approach combines pretrained generative models for space-group-informed structure generation and a universal machine learning force field for energy minimization. Reinforcement fine-tuning can be employed to further boost the accuracy of the framework. We demonstrate the effectiveness of CrystalFormer-CSP on benchmark problems and showcase its usage via web interface and language model integration.", "date": "2025-12-20", "categories": []}
{"id": "2506.14850", "title": "Beyond Force Metrics: Pre-Training MLFFs for Stable MD Simulations", "abstract": "Machine-learning force fields (MLFFs) have emerged as a promising solution for speeding up ab initio molecular dynamics (MD) simulations, where accurate force predictions are critical but often computationally expensive. In this work, we employ GemNet-T, a graph neural network model, as an MLFF and investigate two training strategies: (1) direct training on MD17 (10K samples) without pre-training, and (2) pre-training on the large-scale OC20 dataset followed by fine-tuning on MD17 (10K). While both approaches achieve low force mean absolute errors (MAEs), reaching 5 meV/A per atom, we find that lower force errors do not necessarily guarantee stable MD simulations. Notably, the pre-trained GemNet-T model yields significantly improved simulation stability, sustaining trajectories up to three times longer than the model trained from scratch. By analyzing local properties of the learned force fields, we find that pre-training produces more structured latent representations, smoother force responses to local geometric changes, and more consistent force differences between nearby configurations, all of which contribute to more stable and reliable MD simulations. These findings underscore the value of pre-training on large, diverse datasets to capture complex molecular interactions and highlight that force MAE alone is not always a sufficient metric of MD simulation stability.", "date": "2025-06-17", "categories": []}
{"id": "2512.16151", "title": "Artificial Intelligence-Enabled Holistic Design of Catalysts Tailored for Semiconducting Carbon Nanotube Growth", "abstract": "Catalyst design is crucial for materials synthesis, especially for complex reaction networks. Strategies like collaborative catalytic systems and multifunctional catalysts are effective but face challenges at the nanoscale. Carbon nanotube synthesis contains complicated nanoscale catalytic reactions, thus achieving high-density, high-quality semiconducting CNTs demands innovative catalyst design. In this work, we present a holistic framework integrating machine learning into traditional catalyst design for semiconducting CNT synthesis. It combines knowledge-based insights with data-driven techniques. Three key components, including open-access electronic structure databases for precise physicochemical descriptors, pre-trained natural language processing-based embedding model for higher-level abstractions, and physical - driven predictive models based on experiment data, are utilized. Through this framework, a new method for selective semiconducting CNT synthesis via catalyst - mediated electron injection, tuned by light during growth, is proposed. 54 candidate catalysts are screened, and three with high potential are identified. High-throughput experiments validate the predictions, with semiconducting selectivity exceeding 91% and the FeTiO3 catalyst reaching 98.6%. This approach not only addresses semiconducting CNT synthesis but also offers a generalizable methodology for global catalyst design and nanomaterials synthesis, advancing materials science in precise control.", "date": "2025-12-18", "categories": []}
{"id": "2512.16607", "title": "Riemannian Stochastic Interpolants for Amorphous Particle Systems", "abstract": "Modern generative models hold great promise for accelerating diverse tasks involving the simulation of physical systems, but they must be adapted to the specific constraints of each domain. Significant progress has been made for biomolecules and crystalline materials. Here, we address amorphous materials (glasses), which are disordered particle systems lacking atomic periodicity. Sampling equilibrium configurations of glass-forming materials is a notoriously slow and difficult task. This obstacle could be overcome by developing a generative framework capable of producing equilibrium configurations with well-defined likelihoods. In this work, we address this challenge by leveraging an equivariant Riemannian stochastic interpolation framework which combines Riemannian stochastic interpolant and equivariant flow matching. Our method rigorously incorporates periodic boundary conditions and the symmetries of multi-component particle systems, adapting an equivariant graph neural network to operate directly on the torus. Our numerical experiments on model amorphous systems demonstrate that enforcing geometric and symmetry constraints significantly improves generative performance.", "date": "2025-12-18", "categories": []}
{"id": "2512.15228", "title": "Accelerating High-Throughput Catalyst Screening by Direct Generation of Equilibrium Adsorption Structures", "abstract": "The adsorption energy serves as a crucial descriptor for the large-scale screening of catalysts. Nevertheless, the limited distribution of training data for the extensively utilised machine learning interatomic potential (MLIP), predominantly sourced from near-equilibrium structures, results in unreliable adsorption structures and consequent adsorption energy predictions. In this context, we present DBCata, a deep generative model that integrates a periodic Brownian-bridge framework with an equivariant graph neural network to establish a low-dimensional transition manifold between unrelaxed and DFT-relaxed structures, without requiring explicit energy or force information. Upon training, DBCata effectively generates high-fidelity adsorption geometries, achieving an interatomic distance mean absolute error (DMAE) of 0.035 \\text{\\AA} on the Catalysis-Hub dataset, which is nearly three times superior to that of the current state-of-the-art machine learning potential models. Moreover, the corresponding DFT accuracy can be improved within 0.1 eV in 94\\% of instances by identifying and refining anomalous predictions through a hybrid chemical-heuristic and self-supervised outlier detection approach. We demonstrate that the remarkable performance of DBCata facilitates accelerated high-throughput computational screening for efficient alloy catalysts in the oxygen reduction reaction, highlighting the potential of DBCata as a powerful tool for catalyst design and optimisation.", "date": "2025-12-17", "categories": []}
{"id": "2512.11935", "title": "AGAPI-Agents: An Open-Access Agentic AI Platform for Accelerated Materials Design on AtomGPT.org", "abstract": "Artificial intelligence is reshaping scientific discovery, yet its use in materials research remains limited by fragmented computational ecosystems, reproducibility challenges, and dependence on commercial large language models (LLMs). Here we introduce AGAPI (AtomGPT.org API), an open-access agentic AI platform that integrates more than eight open-source LLMs with over twenty materials-science API endpoints, unifying databases, simulation tools, and machine-learning models through a common orchestration framework. AGAPI employs an Agent-Planner-Executor-Summarizer architecture that autonomously constructs and executes multi-step workflows spanning materials data retrieval, graph neural network property prediction, machine-learning force-field optimization, tight-binding calculations, diffraction analysis, and inverse design. We demonstrate AGAPI through end-to-end workflows, including heterostructure construction, powder X-ray diffraction analysis, and semiconductor defect engineering requiring up to ten sequential operations. In addition, we evaluate AGAPI using 30+ example prompts as test cases and compare agentic predictions with and without tool access against experimental data. With more than 1,000 active users, AGAPI provides a scalable and transparent foundation for reproducible, AI-accelerated materials discovery. AGAPI-Agents codebase is available at https://github.com/atomgptlab/agapi.", "date": "2025-12-12", "categories": []}
{"id": "2411.08464", "title": "A Generation Framework with Strict Constraints for Crystal Materials Design", "abstract": "The design of crystal materials plays a critical role in areas such as new energy development, biomedical engineering, and semiconductors. Recent advances in data-driven methods have enabled the generation of diverse crystal structures. However, most existing approaches still rely on random sampling without strict constraints, requiring multiple post-processing steps to identify stable candidates with the desired physical and chemical properties. In this work, we present a new constrained generation framework that takes multiple constraints as input and enables the generation of crystal structures with specific chemical and properties. In this framework, intermediate constraints, such as symmetry information and composition ratio, are generated by a constraint generator based on large language models (LLMs), which considers the target properties. These constraints are then used by a subsequent crystal structure generator to ensure that the structure generation process is under control. Our method generates crystal structures with a probability of meeting the target properties that is more than twice that of existing approaches. Furthermore, nearly 100% of the generated crystals strictly adhere to predefined chemical composition, eliminating the risks of supply chain during production.", "date": "2024-11-13", "categories": []}
{"id": "2512.09514", "title": "Transport Novelty Distance: A Distributional Metric for Evaluating Material Generative Models", "abstract": "Recent advances in generative machine learning have opened new possibilities for the discovery and design of novel materials. However, as these models become more sophisticated, the need for rigorous and meaningful evaluation metrics has grown. Existing evaluation approaches often fail to capture both the quality and novelty of generated structures, limiting our ability to assess true generative performance. In this paper, we introduce the Transport Novelty Distance (TNovD) to judge generative models used for materials discovery jointly by the quality and novelty of the generated materials. Based on ideas from Optimal Transport theory, TNovD uses a coupling between the features of the training and generated sets, which is refined into a quality and memorization regime by a threshold. The features are generated from crystal structures using a graph neural network that is trained to distinguish between materials, their augmented counterparts, and differently sized supercells using contrastive learning. We evaluate our proposed metric on typical toy experiments relevant for crystal structure prediction, including memorization, noise injection and lattice deformations. Additionally, we validate the TNovD on the MP20 validation set and the WBM substitution dataset, demonstrating that it is capable of detecting both memorized and low-quality material data. We also benchmark the performance of several popular material generative models. While introduced for materials, our TNovD framework is domain-agnostic and can be adapted for other areas, such as images and molecules.", "date": "2025-12-10", "categories": []}
{"id": "2503.20541", "title": "Fast, Modular, and Differentiable Framework for Machine Learning-Enhanced Molecular Simulations", "abstract": "We present an end-to-end differentiable molecular simulation framework (DIMOS) for molecular dynamics and Monte Carlo simulations. DIMOS easily integrates machine-learning-based interatomic potentials and implements classical force fields including an efficient implementation of particle-mesh Ewald. Thanks to its modularity, both classical and machine-learning-based approaches can be easily combined into a hybrid description of the system (ML/MM). By supporting key molecular dynamics features such as efficient neighborlists and constraint algorithms for larger time steps, the framework makes steps in bridging the gap between hand-optimized simulation engines and the flexibility of a \\verb|PyTorch| implementation. We show that due to improved linear instead of quadratic scaling as function of system size DIMOS is able to obtain speed-up factors of up to $170\\times$ for classical force field simulations against another fully differentiable simulation framework. The advantage of differentiability is demonstrated by an end-to-end optimization of the proposal distribution in a Markov Chain Monte Carlo simulation based on Hamiltonian Monte Carlo (HMC). Using these optimized simulation parameters a $3\\times$ acceleration is observed in comparison to ad-hoc chosen simulation parameters. The code is available at https://github.com/nec-research/DIMOS.", "date": "2025-03-26", "categories": []}
{"id": "2512.05989", "title": "A self-driving lab for solution-processed electrochromic thin films", "abstract": "Solution-processed electrochromic materials offer high potential for energy-efficient smart windows and displays. Their performance varies with material choice and processing conditions. Electrochromic thin film electrodes require a smooth, defect-free coating for optimal contrast between bleached and colored states. The complexity of optimizing the spin-coated electrochromic thin layer poses challenges for rapid development. This study demonstrates the use of self-driving laboratories to accelerate the development of electrochromic coatings by coupling automation with machine learning. Our system combines automated data acquisition, image processing, spectral analysis, and Bayesian optimization to explore processing parameters efficiently. This approach not only increases throughput but also enables a pointed search for optimal processing parameters. The approach can be applied to various solution-processed materials, highlighting the potential of self-driving labs in enhancing materials discovery and process optimization.", "date": "2025-11-28", "categories": []}
{"id": "2512.06301", "title": "Chemistry Integrated Language Model using Hierarchical Molecular Representation for Polymer Informatics", "abstract": "Machine learning has transformed material discovery for inorganic compounds and small molecules, yet polymers remain largely inaccessible to these methods. While data scarcity is often cited as the primary bottleneck, we demonstrate that strategic molecular representations can overcome this limitation. We introduce CI-LLM (Chemically Informed Language Model), a framework combining HAPPY (Hierarchically Abstracted rePeat unit of PolYmer), which encodes chemical substructures as tokens, with numerical descriptors within transformer architectures. For property prediction, De$^3$BERTa, our descriptor-enriched encoder, achieves 3.5x faster inference than SMILES-based models with improved accuracy ($R^2$ score gains of 0.9-4.1 percent across four properties), while providing interpretable structure-property insights at the subgroup level. For inverse design, our GPT-based generator produces polymers with targeted properties, achieving 100 percent scaffold retention and successful multi-property optimization for negatively correlated objectives. This comprehensive framework demonstrates both forward prediction and inverse design capabilities, showcasing how strategic molecular representation advances machine learning applications in polymer science.", "date": "2025-12-06", "categories": []}
{"id": "2512.07289", "title": "Equivariant Diffusion for Crystal Structure Prediction", "abstract": "In addressing the challenge of Crystal Structure Prediction (CSP), symmetry-aware deep learning models, particularly diffusion models, have been extensively studied, which treat CSP as a conditional generation task. However, ensuring permutation, rotation, and periodic translation equivariance during diffusion process remains incompletely addressed. In this work, we propose EquiCSP, a novel equivariant diffusion-based generative model. We not only address the overlooked issue of lattice permutation equivariance in existing models, but also develop a unique noising algorithm that rigorously maintains periodic translation equivariance throughout both training and inference processes. Our experiments indicate that EquiCSP significantly surpasses existing models in terms of generating accurate structures and demonstrates faster convergence during the training process.", "date": "2025-12-08", "categories": []}
{"id": "2512.07486", "title": "Materium: An Autoregressive Approach for Material Generation", "abstract": "We present Materium: an autoregressive transformer for generating crystal structures that converts 3D material representations into token sequences. These sequences include elements with oxidation states, fractional coordinates and lattice parameters. Unlike diffusion approaches, which refine atomic positions iteratively through many denoising steps, Materium places atoms at precise fractional coordinates, enabling fast, scalable generation. With this design, the model can be trained in a few hours on a single GPU and generate samples much faster on GPUs and CPUs than diffusion-based approaches. The model was trained and evaluated using multiple properties as conditions, including fundamental properties, such as density and space group, as well as more practical targets, such as band gap and magnetic density. In both single and combined conditions, the model performs consistently well, producing candidates that align with the requested inputs.", "date": "2025-12-08", "categories": []}
{"id": "2512.05367", "title": "Enhancing Dimensionality Prediction in Hybrid Metal Halides via Feature Engineering and Class-Imbalance Mitigation", "abstract": "We present a machine learning framework for predicting the structural dimensionality of hybrid metal halides (HMHs), including organic-inorganic perovskites, using a combination of chemically-informed feature engineering and advanced class-imbalance handling techniques. The dataset, consisting of 494 HMH structures, is highly imbalanced across dimensionality classes (0D, 1D, 2D, 3D), posing significant challenges to predictive modeling. This dataset was later augmented to 1336 via the Synthetic Minority Oversampling Technique (SMOTE) to mitigate the effects of the class imbalance. We developed interaction-based descriptors and integrated them into a multi-stage workflow that combines feature selection, model stacking, and performance optimization to improve dimensionality prediction accuracy. Our approach significantly improves F1-scores for underrepresented classes, achieving robust cross-validation performance across all dimensionalities.", "date": "2025-12-05", "categories": []}
{"id": "2512.05818", "title": "Machine-learning-enabled interpretation of tribological deformation patterns in large-scale MD data", "abstract": "Molecular dynamics (MD) simulations have become indispensable for exploring tribological deformation patterns at the atomic scale. However, transforming the resulting high-dimensional data into interpretable deformation pattern maps remains a resource-intensive and largely manual process. In this work, we introduce a data-driven workflow that automates this interpretation step using unsupervised and supervised learning. Grain-orientation-colored computational tomograph pictures obtained from CuNi alloy simulations were first compressed through an autoencoder to a 32-dimensional global feature vector. Despite this strong compression, the reconstructed images retained the essential microstructural motifs: grain boundaries, stacking faults, twins, and partial lattice rotations, while omitting only the finest defects. The learned representations were then combined with simulation metadata (composition, load, time, temperature, and spatial position) to train a CNN-MLP model to predict the dominant deformation pattern. The resulting model achieves a prediction accuracy of approximately 96% on validation data. A refined evaluation strategy, in which an entire spatial region containing distinct grains was excluded from training, provides a more robust measure of generalization. The approach demonstrates that essential tribological deformation signatures can be automatically identified and classified from structural images using Machine Learning. This proof of concept constitutes a first step towards fully automated, data-driven construction of tribological mechanism maps and, ultimately, toward predictive modeling frameworks that may reduce the need for large-scale MD simulation campaigns.", "date": "2025-12-05", "categories": []}
{"id": "2509.12178", "title": "All that structure matches does not glitter", "abstract": "Generative models for materials, especially inorganic crystals, hold potential to transform the theoretical prediction of novel compounds and structures. Advancement in this field depends on robust benchmarks and minimal, information-rich datasets that enable meaningful model evaluation. This paper critically examines common datasets and reported metrics for a crystal structure prediction task$\\unicode{x2014}$generating the most likely structures given the chemical composition of a material. We focus on three key issues: First, materials datasets should contain unique crystal structures; for example, we show that the widely-utilized carbon-24 dataset only contains $\\approx$40% unique structures. Second, materials datasets should not be split randomly if polymorphs of many different compositions are numerous, which we find to be the case for the perov-5 and MP-20 datasets. Third, benchmarks can mislead if used uncritically, e.g., reporting a match rate metric without considering the structural variety exhibited by identical building blocks. To address these oft-overlooked issues, we introduce several fixes. We provide revised versions of the carbon-24 dataset: one with duplicates removed, one deduplicated and split by number of atoms $N$, one with enantiomorphs, and two containing only identical structures but with different unit cells. We also propose new splits for datasets with polymorphs, ensuring that polymorphs are grouped within each split subset, setting a more sensible standard for benchmarking model performance. Finally, we present METRe and cRMSE, new model evaluation metrics that can correct existing issues with the match rate metric.", "date": "2025-09-15", "categories": []}
{"id": "2512.03050", "title": "Physics-Informed Machine Learning for Steel Development: A Computational Framework and CCT Diagram Modelling", "abstract": "Machine learning (ML) has emerged as a powerful tool for accelerating the computational design and production of materials. In materials science, ML has primarily supported large-scale discovery of novel compounds using first-principles data and digital twin applications for optimizing manufacturing processes. However, applying general-purpose ML frameworks to complex industrial materials such as steel remains a challenge. A key obstacle is accurately capturing the intricate relationship between chemical composition, processing parameters, and the resulting microstructure and properties. To address this, we introduce a computational framework that combines physical insights with ML to develop a physics-informed continuous cooling transformation (CCT) model for steels. Our model, trained on a dataset of 4,100 diagrams, is validated against literature and experimental data. It demonstrates high computational efficiency, generating complete CCT diagrams with 100 cooling curves in under 5 seconds. It also shows strong generalizability across alloy steels, achieving phase classification F1 scores above 88% for all phases. For phase transition temperature regression, it attains mean absolute errors (MAE) below 20 {\\deg}C across all phases except bainite, which shows a slightly higher MAE of 27 {\\deg}C. This framework can be extended with additional generic and customized ML models to establish a universal digital twin platform for heat treatment. Integration with complementary simulation tools and targeted experiments will further support accelerated materials design workflows.", "date": "2025-11-21", "categories": []}
{"id": "2511.22307", "title": "Enhanced Conditional Generation of Double Perovskite by Knowledge-Guided Language Model Feedback", "abstract": "Double perovskites (DPs) are promising candidates for sustainable energy technologies due to their compositional tunability and compatibility with low-energy fabrication, yet their vast design space poses a major challenge for conditional materials discovery. This work introduces a multi-agent, text gradient-driven framework that performs DP composition generation under natural-language conditions by integrating three complementary feedback sources: LLM-based self-evaluation, DP-specific domain knowledge-informed feedback, and ML surrogate-based feedback. Analogous to how knowledge-informed machine learning improves the reliability of conventional data-driven models, our framework incorporates domain-informed text gradients to guide the generative process toward physically meaningful regions of the DP composition space. Systematic comparison of three incremental configurations, (i) pure LLM generation, (ii) LLM generation with LLM reasoning-based feedback, and (iii) LLM generation with domain knowledge-guided feedback, shows that iterative guidance from knowledge-informed gradients improves stability-condition satisfaction without additional training data, achieving over 98% compositional validity and up to 54% stable or metastable candidates, surpassing both the LLM-only baseline (43%) and prior GAN-based results (27%). Analyses of ML-based gradients further reveal that they enhance performance in in-distribution (ID) regions but become unreliable in out-of-distribution (OOD) regimes. Overall, this work provides the first systematic analysis of multi-agent, knowledge-guided text gradients for DP discovery and establishes a generalizable blueprint for MAS-driven generative materials design aimed at advancing sustainable technologies.", "date": "2025-11-27", "categories": []}
{"id": "2512.02057", "title": "Opening the Black Box: An Explainable, Few-shot AI4E Framework Informed by Physics and Expert Knowledge for Materials Engineering", "abstract": "The industrial adoption of Artificial Intelligence for Engineering (AI4E) faces two fundamental bottlenecks: scarce high-quality data and the lack of interpretability in black-box models-particularly critical in safety-sensitive sectors like aerospace. We present an explainable, few-shot AI4E framework that is systematically informed by physics and expert knowledge throughout its architecture. Starting from only 32 experimental samples in an aerial K439B superalloy castings repair welding case, we first augment physically plausible synthetic data through a three-stage protocol: differentiated noise injection calibrated to process variabilities, enforcement of hard physical constraints, and preservation of inter-parameter relationships. We then employ a nested optimization strategy for constitutive model discovery, where symbolic regression explores equation structures while differential evolution optimizes parameters, followed by intensive parameter refinement using hybrid global-local optimization. The resulting interpretable constitutive equation achieves 88% accuracy in predicting hot-cracking tendency. This equation not only provides quantitative predictions but also delivers explicit physical insight, revealing how thermal, geometric, and metallurgical mechanisms couple to drive cracking-thereby advancing engineers' cognitive understanding of the process. Furthermore, the constitutive equation serves as a multi-functional tool for process optimization and high-fidelity virtual data generation, enabling accuracy improvements in other data-driven models. Our approach provides a general blueprint for developing trustworthy AI systems that embed engineering domain knowledge directly into their architecture, enabling reliable adoption in high-stakes industrial applications where data is limited but physical understanding is available.", "date": "2025-11-28", "categories": []}
{"id": "2512.02947", "title": "Representation of Inorganic Synthesis Reactions and Prediction: Graphical Framework and Datasets", "abstract": "While machine learning has enabled the rapid prediction of inorganic materials with novel properties, the challenge of determining how to synthesize these materials remains largely unsolved. Previous work has largely focused on predicting precursors or reaction conditions, but only rarely on full synthesis pathways. We introduce the ActionGraph, a directed acyclic graph framework that encodes both the chemical and procedural structure, in terms of synthesis operations, of inorganic synthesis reactions. Using 13,017 text-mined solid-state synthesis reactions from the Materials Project, we show that incorporating PCA-reduced ActionGraph adjacency matrices into a $k$-nearest neighbors retrieval model significantly improves synthesis pathway prediction. While the ActionGraph framework only results in a 1.34% and 2.76% increase in precursor and operation F1 scores (average over varying numbers of PCA components) respectively, the operation length matching accuracy rises 3.4 times (from 15.8% to 53.3%). We observe an interesting trade-off where precursor prediction performance peaks at 10-11 PCA components while operation prediction continues improving up to 30 components. This suggests composition information dominates precursor selection while structural information is critical for operation sequencing. Overall, the ActionGraph framework demonstrates strong potential, and with further adoption, its full range of benefits can be effectively realized.", "date": "2025-12-02", "categories": []}
{"id": "2512.01160", "title": "From Regression to Classification: Exploring the Benefits of Categorical Representations of Energy in MLIPs", "abstract": "Density Functional Theory (DFT) is a widely used computational method for estimating the energy and behavior of molecules. Machine Learning Interatomic Potentials (MLIPs) are models trained to approximate DFT-level energies and forces at dramatically lower computational cost. Many modern MLIPs rely on a scalar regression formulation; given information about a molecule, they predict a single energy value and corresponding forces while minimizing absolute error with DFT's calculations. In this work, we explore a multi-class classification formulation that predicts a categorical distribution over energy/force values, providing richer supervision through multiple targets. Most importantly, this approach offers a principled way to quantify model uncertainty.\n  In particular, our method predicts a histogram of the energy/force distribution, converts scalar targets into histograms, and trains the model using cross-entropy loss. Our results demonstrate that this categorical formulation can achieve absolute error performance comparable to regression baselines. Furthermore, this representation enables the quantification of epistemic uncertainty through the entropy of the predicted distribution, offering a measure of model confidence absent in scalar regression approaches.", "date": "2025-12-01", "categories": []}
{"id": "2512.01519", "title": "QuantumCanvas: A Multimodal Benchmark for Visual Learning of Atomic Interactions", "abstract": "Despite rapid advances in molecular and materials machine learning, most models still lack physical transferability: they fit correlations across whole molecules or crystals rather than learning the quantum interactions between atomic pairs. Yet bonding, charge redistribution, orbital hybridization, and electronic coupling all emerge from these two-body interactions that define local quantum fields in many-body systems. We introduce QuantumCanvas, a large-scale multimodal benchmark that treats two-body quantum systems as foundational units of matter. The dataset spans 2,850 element-element pairs, each annotated with 18 electronic, thermodynamic, and geometric properties and paired with ten-channel image representations derived from l- and m-resolved orbital densities, angular field transforms, co-occupancy maps, and charge-density projections. These physically grounded images encode spatial, angular, and electrostatic symmetries without explicit coordinates, providing an interpretable visual modality for quantum learning. Benchmarking eight architectures across 18 targets, we report mean absolute errors of 0.201 eV on energy gap using GATv2, 0.265 eV on HOMO and 0.274 eV on LUMO using EGNN. For energy-related quantities, DimeNet attains 2.27 eV total-energy MAE and 0.132 eV repulsive-energy MAE, while a multimodal fusion model achieves a 2.15 eV Mermin free-energy MAE. Pretraining on QuantumCanvas further improves convergence stability and generalization when fine-tuned on larger datasets such as QM9, MD17, and CrysMTM. By unifying orbital physics with vision-based representation learning, QuantumCanvas provides a principled and interpretable basis for learning transferable quantum interactions through coupled visual and numerical modalities. Dataset and model implementations are available at https://github.com/KurbanIntelligenceLab/QuantumCanvas.", "date": "2025-12-01", "categories": []}
{"id": "2512.01735", "title": "Automating modeling in mechanics: LLMs as designers of physics-constrained neural networks for constitutive modeling of materials", "abstract": "Large language model (LLM)-based agentic frameworks increasingly adopt the paradigm of dynamically generating task-specific agents. We suggest that not only agents but also specialized software modules for scientific and engineering tasks can be generated on demand. We demonstrate this concept in the field of solid mechanics. There, so-called constitutive models are required to describe the relationship between mechanical stress and body deformation. Constitutive models are essential for both the scientific understanding and industrial application of materials. However, even recent data-driven methods of constitutive modeling, such as constitutive artificial neural networks (CANNs), still require substantial expert knowledge and human labor. We present a framework in which an LLM generates a CANN on demand, tailored to a given material class and dataset provided by the user. The framework covers LLM-based architecture selection, integration of physical constraints, and complete code generation. Evaluation on three benchmark problems demonstrates that LLM-generated CANNs achieve accuracy comparable to or greater than manually engineered counterparts, while also exhibiting reliable generalization to unseen loading scenarios and extrapolation to large deformations. These findings indicate that LLM-based generation of physics-constrained neural networks can substantially reduce the expertise required for constitutive modeling and represent a step toward practical end-to-end automation.", "date": "2025-12-01", "categories": []}
{"id": "2512.01756", "title": "Mofasa: A Step Change in Metal-Organic Framework Generation", "abstract": "Mofasa is an all-atom latent diffusion model with state-of-the-art performance for generating Metal-Organic Frameworks (MOFs). These are highly porous crystalline materials used to harvest water from desert air, capture carbon dioxide, store toxic gases and catalyse chemical reactions. In recognition of their value, the development of MOFs recently received a Nobel Prize in Chemistry.\n  In many ways, MOFs are well-suited for exploiting generative models in chemistry: they are rationally-designable materials with a large combinatorial design space and strong structure-property couplings. And yet, to date, a high performance generative model has been lacking. To fill this gap, we introduce Mofasa, a general-purpose latent diffusion model that jointly samples positions, atom-types and lattice vectors for systems as large as 500 atoms. Mofasa avoids handcrafted assembly algorithms common in the literature, unlocking the simultaneous discovery of metal nodes, linkers and topologies.\n  To help the scientific community build on our work, we release MofasaDB, an annotated library of hundreds of thousands of sampled MOF structures, along with a user-friendly web interface for search and discovery: https://mofux.ai/ .", "date": "2025-12-01", "categories": []}
{"id": "2510.11148", "title": "Enhanced Sampling for Efficient Learning of Coarse-Grained Machine Learning Potentials", "abstract": "Coarse-graining (CG) enables molecular dynamics (MD) simulations of larger systems and longer timescales that are otherwise infeasible with atomistic models. Machine learning potentials (MLPs), with their capacity to capture many-body interactions, can provide accurate approximations of the potential of mean force (PMF) in CG models. Current CG MLPs are typically trained in a bottom-up manner via force matching, which in practice relies on configurations sampled from the unbiased equilibrium Boltzmann distribution to ensure thermodynamic consistency. This convention poses two key limitations: first, sufficiently long atomistic trajectories are needed to reach convergence; and second, even once equilibrated, transition regions remain poorly sampled. To address these issues, we employ enhanced sampling to bias along CG degrees of freedom for data generation, and then recompute the forces with respect to the unbiased potential. This strategy simultaneously shortens the simulation time required to produce equilibrated data and enriches sampling in transition regions, while preserving the correct PMF. We demonstrate its effectiveness on the M\\\"uller-Brown potential and capped alanine, achieving notable improvements. Our findings support the use of enhanced sampling for force matching as a promising direction to improve the accuracy and reliability of CG MLPs.", "date": "2025-10-13", "categories": []}
{"id": "2511.17741", "title": "Diffusion Models are Molecular Dynamics Simulators", "abstract": "We prove that a denoising diffusion sampler equipped with a sequential bias across the batch dimension is exactly an Euler-Maruyama integrator for overdamped Langevin dynamics. Each reverse denoising step, with its associated spring stiffness, can be interpreted as one step of a stochastic differential equation with an effective time step set jointly by the noise schedule and that stiffness. The learned score then plays the role of the drift, equivalently the gradient of a learned energy, yielding a precise correspondence between diffusion sampling and Langevin time evolution.\n  This equivalence recasts molecular dynamics (MD) in terms of diffusion models. Accuracy is no longer tied to a fixed, extremely small MD time step; instead, it is controlled by two scalable knobs: model capacity, which governs how well the drift is approximated, and the number of denoising steps, which sets the integrator resolution. In practice, this leads to a fully data-driven MD framework that learns forces from uncorrelated equilibrium snapshots, requires no hand-engineered force fields, uses no trajectory data for training, and still preserves the Boltzmann distribution associated with the learned energy.\n  We derive trajectory-level, information-theoretic error bounds that cleanly separate discretization error from score-model error, clarify how temperature enters through the effective spring, and show that the resulting sampler generates molecular trajectories with MD-like temporal correlations, even though the model is trained only on static configurations.", "date": "2025-11-21", "categories": []}
{"id": "2511.20327", "title": "MXtalTools: A Toolkit for Machine Learning on Molecular Crystals", "abstract": "We present MXtalTools, a flexible Python package for the data-driven modelling of molecular crystals, facilitating machine learning studies of the molecular solid state. MXtalTools comprises several classes of utilities: (1) synthesis, collation, and curation of molecule and crystal datasets, (2) integrated workflows for model training and inference, (3) crystal parameterization and representation, (4) crystal structure sampling and optimization, (5) end-to-end differentiable crystal sampling, construction and analysis. Our modular functions can be integrated into existing workflows or combined and used to build novel modelling pipelines. MXtalTools leverages CUDA acceleration to enable high-throughput crystal modelling. The Python code is available open-source on our GitHub page, with detailed documentation on ReadTheDocs.", "date": "2025-11-25", "categories": []}
{"id": "2408.12984", "title": "PDDFormer: Pairwise Distance Distribution Graph Transformer for Crystal Material Property Prediction", "abstract": "Crystal structures can be simplified as a periodic point set that repeats across three-dimensional space along an underlying lattice. Traditionally, crystal representation methods characterize the structure using descriptors such as lattice parameters, symmetry, and space groups. However, in reality, atoms in materials always vibrate above absolute zero, causing their positions to fluctuate continuously. This dynamic behavior disrupts the fundamental periodicity of the lattice, making crystal graphs based on static lattice parameters and conventional descriptors discontinuous under slight perturbations. Chemists proposed the pairwise distance distribution (PDD) method to address this problem. However, the completeness of PDD requires defining a large number of neighboring atoms, leading to high computational costs. Additionally, PDD does not account for atomic information, making it challenging to apply it directly to crystal material property prediction tasks. To tackle these challenges, we introduce the atom-Weighted Pairwise Distance Distribution (WPDD) and Unit cell Pairwise Distance Distribution (UPDD) and apply them to the construction of multi-edge crystal graphs. We demonstrate the continuity and general completeness of crystal graphs under slight atomic position perturbations. Moreover, by modeling PDD as global information and integrating it into matrix-based message passing, we significantly reduce computational costs. Comprehensive evaluation results show that WPDDFormer achieves state-of-the-art predictive accuracy across tasks on benchmark datasets such as the Materials Project and JARVIS-DFT.", "date": "2024-08-23", "categories": []}
{"id": "2507.19438", "title": "Gradient-based grand canonical optimization enabled by graph neural networks with fractional atomic existence", "abstract": "Machine learning interatomic potentials have become an indispensable tool for materials science, enabling the study of larger systems and longer timescales. State-of-the-art models are generally graph neural networks that employ message passing to iteratively update atomic embeddings that are ultimately used for predicting properties. In this work we extend the message passing formalism with the inclusion of a continuous variable that accounts for fractional atomic existence. This allows us to calculate the gradient of the Gibbs free energy with respect to both the Cartesian coordinates of atoms and their existence. Using this we propose a gradient-based grand canonical optimization method and document its capabilities for a Cu(110) surface oxide.", "date": "2025-07-25", "categories": []}
{"id": "2508.00578", "title": "Learning Potential Energy Surfaces of Hydrogen Atom Transfer Reactions in Peptides", "abstract": "Hydrogen atom transfer (HAT) reactions are essential in many biological processes, such as radical migration in damaged proteins, but their mechanistic pathways remain incompletely understood. Simulating HAT is challenging due to the need for quantum chemical accuracy at biologically relevant scales; thus, neither classical force fields nor DFT-based molecular dynamics are applicable. Machine-learned potentials offer an alternative, able to learn potential energy surfaces (PESs) with near-quantum accuracy. However, training these models to generalize across diverse HAT configurations, especially at radical positions in proteins, requires tailored data generation and careful model selection. Here, we systematically generate HAT configurations in peptides to build large datasets using semiempirical methods and DFT. We benchmark three graph neural network architectures (SchNet, Allegro, and MACE) on their ability to learn HAT PESs and indirectly predict reaction barriers from energy predictions. MACE consistently outperforms the others in energy, force, and barrier prediction, achieving a mean absolute error of 1.13 kcal/mol on out-of-distribution DFT barrier predictions. Using molecular dynamics, we show our MACE potential is stable, reactive, and generalizes beyond training data to model HAT barriers in collagen I. This accuracy enables integration of ML potentials into large-scale collagen simulations to compute reaction rates from predicted barriers, advancing mechanistic understanding of HAT and radical migration in peptides. We analyze scaling laws, model transferability, and cost-performance trade-offs, and outline strategies for improvement by combining ML potentials with transition state search algorithms and active learning. Our approach is generalizable to other biomolecular systems, enabling quantum-accurate simulations of chemical reactivity in complex environments.", "date": "2025-08-01", "categories": []}
{"id": "2511.19263", "title": "Solar-GECO: Perovskite Solar Cell Property Prediction with Geometric-Aware Co-Attention", "abstract": "Perovskite solar cells are promising candidates for next-generation photovoltaics. However, their performance as multi-scale devices is determined by complex interactions between their constituent layers. This creates a vast combinatorial space of possible materials and device architectures, making the conventional experimental-based screening process slow and expensive. Machine learning models try to address this problem, but they only focus on individual material properties or neglect the important geometric information of the perovskite crystal. To address this problem, we propose to predict perovskite solar cell power conversion efficiency with a geometric-aware co-attention (Solar-GECO) model. Solar-GECO combines a geometric graph neural network (GNN) - that directly encodes the atomic structure of the perovskite absorber - with language model embeddings that process the textual strings representing the chemical compounds of the transport layers and other device components. Solar-GECO also integrates a co-attention module to capture intra-layer dependencies and inter-layer interactions, while a probabilistic regression head predicts both power conversion efficiency (PCE) and its associated uncertainty. Solar-GECO achieves state-of-the-art performance, significantly outperforming several baselines, reducing the mean absolute error (MAE) for PCE prediction from 3.066 to 2.936 compared to semantic GNN (the previous state-of-the-art model). Solar-GECO demonstrates that integrating geometric and textual information provides a more powerful and accurate framework for PCE prediction.", "date": "2025-11-24", "categories": []}
{"id": "2511.19335", "title": "High-throughput validation of phase formability and simulation accuracy of Cantor alloys", "abstract": "High-throughput methods enable accelerated discovery of novel materials in complex systems such as high-entropy alloys, which exhibit intricate phase stability across vast compositional spaces. Computational approaches, including Density Functional Theory (DFT) and calculation of phase diagrams (CALPHAD), facilitate screening of phase formability as a function of composition and temperature. However, the integration of computational predictions with experimental validation remains challenging in high-throughput studies. In this work, we introduce a quantitative confidence metric to assess the agreement between predictions and experimental observations, providing a quantitative measure of the confidence of machine learning models trained on either DFT or CALPHAD input in accounting for experimental evidence. The experimental dataset was generated via high-throughput in-situ synchrotron X-ray diffraction on compositionally varied FeNiMnCr alloy libraries, heated from room temperature to ~1000 {\\deg}C. Agreement between the observed and predicted phases was evaluated using either temperature-independent phase classification or a model that incorporates a temperature-dependent probability of phase formation. This integrated approach demonstrates where strong overall agreement between computation and experiment exists, while also identifying key discrepancies, particularly in FCC/BCC predictions at Mn-rich regions to inform future model refinement.", "date": "2025-11-24", "categories": []}
{"id": "2405.11783", "title": "Property-guided Inverse Design of Metal-Organic Frameworks Using Quantum Natural Language Processing", "abstract": "In this study, we explore the potential of using quantum natural language processing (QNLP) to inverse design metal-organic frameworks (MOFs) with targeted properties. Specifically, by analyzing 450 hypothetical MOF structures consisting of 3 topologies, 10 metal nodes and 15 organic ligands, we categorize these structures into four distinct classes for pore volume and $CO_{2}$ Henry's constant values. We then compare various QNLP models (i.e. the bag-of-words, DisCoCat (Distributional Compositional Categorical), and sequence-based models) to identify the most effective approach to process the MOF dataset. Using a classical simulator provided by the IBM Qiskit, the bag-of-words model is identified to be the optimum model, achieving validation accuracies of 88.6% and 78.0% for binary classification tasks on pore volume and $CO_{2}$ Henry's constant, respectively. Further, we developed multi-class classification models tailored to the probabilistic nature of quantum circuits, with average test accuracies of 92% and 80% across different classes for pore volume and $CO_{2}$ Henry's constant datasets. Finally, the performance of generating MOF with target properties showed accuracies of 93.5% for pore volume and 87% for $CO_{2}$ Henry's constant, respectively. Although our investigation covers only a fraction of the vast MOF search space, it marks a promising first step towards using quantum computing for materials design, offering a new perspective through which to explore the complex landscape of MOFs.", "date": "2024-05-20", "categories": []}
{"id": "2511.16135", "title": "CoSP: Reconfigurable Multi-State Metamaterial Inverse Design via Contrastive Pretrained Large Language Model", "abstract": "Metamaterials, known for their ability to manipulate light at subwavelength scales, face significant design challenges due to their complex and sophisticated structures. Consequently, deep learning has emerged as a powerful tool to streamline their design process. Reconfigurable multi-state metamaterials (RMMs) with adjustable parameters can switch their optical characteristics between different states upon external stimulation, leading to numerous applications. However, existing deep learning-based inverse design methods fall short in considering reconfigurability with multi-state switching. To address this challenge, we propose CoSP, an intelligent inverse design method based on contrastive pretrained large language model (LLM). By performing contrastive pretraining on multi-state spectrum, a well-trained spectrum encoder capable of understanding the spectrum is obtained, and it subsequently interacts with a pretrained LLM. This approach allows the model to preserve its linguistic capabilities while also comprehending Maxwell's Equations, enabling it to describe material structures with target optical properties in natural language. Our experiments demonstrate that CoSP can design corresponding thin-film metamaterial structures for arbitrary multi-state, multi-band optical responses, showing great potentials in the intelligent design of RMMs for versatile applications.", "date": "2025-11-20", "categories": []}
{"id": "2508.06996", "title": "Explainable AI for Curie Temperature Prediction in Magnetic Materials", "abstract": "We explore machine learning techniques for predicting Curie temperatures of magnetic materials using the NEMAD database. By augmenting the dataset with composition-based and domain-aware descriptors, we evaluate the performance of several machine learning models. We find that the Extra Trees Regressor delivers the best performance reaching an R^2 score of up to 0.85 $\\pm$ 0.01 (cross-validated) for a balanced dataset. We employ the k-means clustering algorithm to gain insights into the performance of chemically distinct material groups. Furthermore, we perform the SHAP analysis to identify key physicochemical drivers of Curie behavior, such as average atomic number and magnetic moment. By employing explainable AI techniques, this analysis offers insights into the model's predictive behavior, thereby advancing scientific interpretability.", "date": "2025-08-09", "categories": []}
{"id": "2511.15222", "title": "Why Physics Still Matters: Improving Machine Learning Prediction of Material Properties with Phonon-Informed Datasets", "abstract": "Machine learning (ML) methods have become powerful tools for predicting material properties with near first-principles accuracy and vastly reduced computational cost. However, the performance of ML models critically depends on the quality, size, and diversity of the training dataset. In materials science, this dependence is particularly important for learning from low-symmetry atomistic configurations that capture thermal excitations, structural defects, and chemical disorder, features that are ubiquitous in real materials but underrepresented in most datasets. The absence of systematic strategies for generating representative training data may therefore limit the predictive power of ML models in technologically critical fields such as energy conversion and photonics. In this work, we assess the effectiveness of graph neural network (GNN) models trained on two fundamentally different types of datasets: one composed of randomly generated atomic configurations and another constructed using physically informed sampling based on lattice vibrations. As a case study, we address the challenging task of predicting electronic and mechanical properties of a prototypical family of optoelectronic materials under realistic finite-temperature conditions. We find that the phonons-informed model consistently outperforms the randomly trained counterpart, despite relying on fewer data points. Explainability analyses further reveal that high-performing models assign greater weight to chemically meaningful bonds that control property variations, underscoring the importance of physically guided data generation. Overall, this work demonstrates that larger datasets do not necessarily yield better GNN predictive models and introduces a simple and general strategy for efficiently constructing high-quality training data in materials informatics.", "date": "2025-11-19", "categories": []}
{"id": "2412.18067", "title": "Automated Materials Discovery Platform Realized: Scanning Probe Microscopy of Combinatorial Libraries", "abstract": "Combinatorial materials libraries provide a powerful platform for mapping how physical properties evolve across binary and ternary cross-sections of multicomponent phase diagrams. While synthesis of such libraries has advanced since the 1960s and been accelerated by laboratory automation, their broader utility depends on rapid, quantitative measurements of composition-dependent structures and functionalities. Scanning probe microscopies (SPM), including piezoresponse force microscopy (PFM), offer unique potential for providing these functionally relevant, spatially resolved readouts. Here, we demonstrate a fully automated SPM framework for exploring ferroelectric properties across combinatorial libraries, focusing on binary Sm-doped BiFeO3 (SmBFO) and ternary Al$_{1-x-y}$Sc$_x$B$_y$N (Al,Sc,B)N systems. In SmBFO, automated exploration identifies the known morphotropic phase boundary with enhanced ferroelectric response and reveals a previously unreported double-peak fine structure. In the (Al,Sc,B)N library, ferroelectric behavior emerges at the phase-stability boundary, correlating with variations in morphology and defect concentration. By integrating automated SPM with wavelength-dispersive spectroscopy (WDS) and photoluminescence mapping, we resolve the composition-morphology-defect-property relationships underlying ferroelectric response and demonstrate a pathway toward a multi-tool, high-throughput characterization platform. Finally, we implement Gaussian-process-based single- and multi-objective Bayesian optimization to enable autonomous exploration, highlighting the Pareto front as a powerful framework for balancing competing physical rewards and accelerating data-driven physics discovery.", "date": "2024-12-24", "categories": []}
{"id": "2511.08955", "title": "MicroEvoEval: A Systematic Evaluation Framework for Image-Based Microstructure Evolution Prediction", "abstract": "Simulating microstructure evolution (MicroEvo) is vital for materials design but demands high numerical accuracy, efficiency, and physical fidelity. Although recent studies on deep learning (DL) offer a promising alternative to traditional solvers, the field lacks standardized benchmarks. Existing studies are flawed due to a lack of comparing specialized MicroEvo DL models with state-of-the-art spatio-temporal architectures, an overemphasis on numerical accuracy over physical fidelity, and a failure to analyze error propagation over time. To address these gaps, we introduce MicroEvoEval, the first comprehensive benchmark for image-based microstructure evolution prediction. We evaluate 14 models, encompassing both domain-specific and general-purpose architectures, across four representative MicroEvo tasks with datasets specifically structured for both short- and long-term assessment. Our multi-faceted evaluation framework goes beyond numerical accuracy and computational cost, incorporating a curated set of structure-preserving metrics to assess physical fidelity. Our extensive evaluations yield several key insights. Notably, we find that modern architectures (e.g., VMamba), not only achieve superior long-term stability and physical fidelity but also operate with an order-of-magnitude greater computational efficiency. The results highlight the necessity of holistic evaluation and identify these modern architectures as a highly promising direction for developing efficient and reliable surrogate models in data-driven materials science.", "date": "2025-11-12", "categories": []}
{"id": "2511.14268", "title": "Statistically controllable microstructure reconstruction framework for heterogeneous materials using sliced-Wasserstein metric and neural networks", "abstract": "Heterogeneous porous materials play a crucial role in various engineering systems. Microstructure characterization and reconstruction provide effective means for modeling these materials, which are critical for conducting physical property simulations, structure-property linkage studies, and enhancing their performance across different applications. To achieve superior controllability and applicability with small sample sizes, we propose a statistically controllable microstructure reconstruction framework that integrates neural networks with sliced-Wasserstein metric. Specifically, our approach leverages local pattern distribution for microstructure characterization and employs a controlled sampling strategy to generate target distributions that satisfy given conditional parameters. A neural network-based model establishes the mapping from the input distribution to the target local pattern distribution, enabling microstructure reconstruction. Combinations of sliced-Wasserstein metric and gradient optimization techniques minimize the distance between these distributions, leading to a stable and reliable model. Our method can perform stochastic and controllable reconstruction tasks even with small sample sizes. Additionally, it can generate large-size (e.g. 512 and 1024) 3D microstructures using a chunking strategy. By introducing spatial location masks, our method excels at generating spatially heterogeneous and complex microstructures. We conducted experiments on stochastic reconstruction, controllable reconstruction, heterogeneous reconstruction, and large-size microstructure reconstruction across various materials. Comparative analysis through visualization, statistical measures, and physical property simulations demonstrates the effectiveness, providing new insights and possibilities for research on structure-property linkage and material inverse design.", "date": "2025-11-18", "categories": []}
{"id": "2507.19536", "title": "Graph Learning Metallic Glass Discovery from Wikipedia", "abstract": "Synthesizing new materials efficiently is highly demanded in various research fields. However, this process is usually slow and expensive, especially for metallic glasses, whose formation strongly depends on the optimal combinations of multiple elements to resist crystallization. This constraint renders only several thousands of candidates explored in the vast material space since 1960. Recently, data-driven approaches armed by advanced machine learning techniques provided alternative routes for intelligent materials design. Due to data scarcity and immature material encoding, the conventional tabular data is usually mined by statistical learning algorithms, giving limited model predictability and generalizability. Here, we propose sophisticated data learning from material network representations. The node elements are encoded from the Wikipedia by a language model. Graph neural networks with versatile architectures are designed to serve as recommendation systems to explore hidden relationships among materials. By employing Wikipedia embeddings from different languages, we assess the capability of natural languages in materials design. Our study proposes a new paradigm to harvesting new amorphous materials and beyond with artificial intelligence.", "date": "2025-07-22", "categories": []}
{"id": "2511.11626", "title": "Omics-scale polymer computational database transferable to real-world artificial intelligence applications", "abstract": "Developing large-scale foundational datasets is a critical milestone in advancing artificial intelligence (AI)-driven scientific innovation. However, unlike AI-mature fields such as natural language processing, materials science, particularly polymer research, has significantly lagged in developing extensive open datasets. This lag is primarily due to the high costs of polymer synthesis and property measurements, along with the vastness and complexity of the chemical space. This study presents PolyOmics, an omics-scale computational database generated through fully automated molecular dynamics simulation pipelines that provide diverse physical properties for over $10^5$ polymeric materials. The PolyOmics database is collaboratively developed by approximately 260 researchers from 48 institutions to bridge the gap between academia and industry. Machine learning models pretrained on PolyOmics can be efficiently fine-tuned for a wide range of real-world downstream tasks, even when only limited experimental data are available. Notably, the generalisation capability of these simulation-to-real transfer models improve significantly as the size of the PolyOmics database increases, exhibiting power-law scaling. The emergence of scaling laws supports the \"more is better\" principle, highlighting the significance of ultralarge-scale computational materials data for improving real-world prediction performance. This unprecedented omics-scale database reveals vast unexplored regions of polymer materials, providing a foundation for AI-driven polymer science.", "date": "2025-11-07", "categories": []}
{"id": "2511.11630", "title": "Predicting Grain Growth in Polycrystalline Materials Using Deep Learning Time Series Models", "abstract": "Grain Growth strongly influences the mechanical behavior of materials, making its prediction a key objective in microstructural engineering. In this study, several deep learning approaches were evaluated, including recurrent neural networks (RNN), long short-term memory (LSTM), temporal convolutional networks (TCN), and transformers, to forecast grain size distributions during grain growth. Unlike full-field simulations, which are computationally demanding, the present work relies on mean-field statistical descriptors extracted from high-fidelity simulations. A dataset of 120 grain growth sequences was processed into normalized grain size distributions as a function of time. The models were trained to predict future distributions from a short temporal history using a recursive forecasting strategy. Among the tested models, the LSTM network achieved the highest accuracy (above 90\\%) and the most stable performance, maintaining physically consistent predictions over extended horizons while reducing computation time from about 20 minutes per sequence to only a few seconds, whereas the other architectures tended to diverge when forecasting further in time. These results highlight the potential of low-dimensional descriptors and LSTM-based forecasting for efficient and accurate microstructure prediction, with direct implications for digital twin development and process optimization.", "date": "2025-11-07", "categories": []}
{"id": "2511.11697", "title": "Benchmarking GNNs for OOD Materials Property Prediction with Uncertainty Quantification", "abstract": "We present MatUQ, a benchmark framework for evaluating graph neural networks (GNNs) on out-of-distribution (OOD) materials property prediction with uncertainty quantification (UQ). MatUQ comprises 1,375 OOD prediction tasks constructed from six materials datasets using five OFM-based and a newly proposed structure-aware splitting strategy, SOAP-LOCO, which captures local atomic environments more effectively. We evaluate 12 representative GNN models under a unified uncertainty-aware training protocol that combines Monte Carlo Dropout and Deep Evidential Regression (DER), and introduce a novel uncertainty metric, D-EviU, which shows the strongest correlation with prediction errors in most tasks. Our experiments yield two key findings. First, the uncertainty-aware training approach significantly improves model prediction accuracy, reducing errors by an average of 70.6\\% across challenging OOD scenarios. Second, the benchmark reveals that no single model dominates universally: earlier models such as SchNet and ALIGNN remain competitive, while newer models like CrystalFramer and SODNet demonstrate superior performance on specific material properties. These results provide practical insights for selecting reliable models under distribution shifts in materials discovery.", "date": "2025-11-12", "categories": []}
{"id": "2511.12036", "title": "Preference Learning from Physics-Based Feedback: Tuning Language Models to Design BCC/B2 Superalloys", "abstract": "We apply preference learning to the task of language model-guided design of novel structural alloys. In contrast to prior work that focuses on generating stable inorganic crystals, our approach targets the synthesizeability of a specific structural class: BCC/B2 superalloys, an underexplored family of materials with potential applications in extreme environments. Using three open-weight models (LLaMA-3.1, Gemma-2, and OLMo-2), we demonstrate that language models can be optimized for multiple design objectives using a single, unified reward signal through Direct Preference Optimization (DPO). Unlike prior approaches that rely on heuristic or human-in-the-loop feedback (costly), our reward signal is derived from thermodynamic phase calculations, offering a scientifically grounded criterion for model tuning. To our knowledge, this is the first demonstration of preference-tuning a language model using physics-grounded feedback for structural alloy design. The resulting framework is general and extensible, providing a path forward for intelligent design-space exploration across a range of physical science domains.", "date": "2025-11-15", "categories": []}
{"id": "2511.12171", "title": "FGM optimization in complex domains using Gaussian process regression based profile generation algorithm", "abstract": "This manuscript addresses the challenge of designing functionally graded materials (FGMs) for arbitrary-shaped domains. Towards this goal, the present work proposes a generic volume fraction profile generation algorithm based on Gaussian Process Regression (GPR). The proposed algorithm can handle complex-shaped domains and generate smooth FGM profiles while adhering to the specified volume fraction values at boundaries/part of boundaries. The resulting design space from GPR comprises diverse profiles, enhancing the potential for discovering optimal configurations. Further, the algorithm allows the user to control the smoothness of the underlying profiles and the size of the design space through a length scale parameter. Further, the proposed profile generation scheme is coupled with the genetic algorithm to find the optimum FGM profiles for a given application. To make the genetic algorithm consistent with the GPR profile generation scheme, the standard simulated binary crossover operator in the genetic algorithm has been modified with a projection operator. We present numerous thermoelastic optimization examples to demonstrate the efficacy of the proposed profile generation algorithm and optimization framework.", "date": "2025-11-15", "categories": []}
{"id": "2511.12995", "title": "Revealing the dynamic responses of Pb under shock loading based on DFT-accuracy machine learning potential", "abstract": "Lead (Pb) is a typical low-melting-point ductile metal and serves as an important model material in the study of dynamic responses. Under shock-wave loading, its dynamic mechanical behavior comprises two key phenomena: plastic deformation and shock induced phase transitions. The underlying mechanisms of these processes are still poorly understood. Revealing these mechanisms remains challenging for experimental approaches. Non-equilibrium molecular dynamics (NEMD) simulations are an alternative theoretical tool for studying dynamic responses, as they capture atomic-scale mechanisms such as defect evolution and deformation pathways. However, due to the limited accuracy of empirical interatomic potentials, the reliability of previous NEMD studies is questioned. Using our newly developed machine learning potential for Pb-Sn alloys, we revisited the microstructure evolution in response to shock loading under various shock orientations. The results reveal that shock loading along the [001] orientation of Pb exhibits a fast, reversible, and massive phase transition and stacking fault evolution. The behavior of Pb differs from previous studies by the absence of twinning during plastic deformation. Loading along the [011] orientation leads to slow, irreversible plastic deformation, and a localized FCC-BCC phase transition in the Pitsch orientation relationship. This study provides crucial theoretical insights into the dynamic mechanical response of Pb, offering a theoretical input for understanding the microstructure-performance relationship under extreme conditions.", "date": "2025-11-17", "categories": []}
{"id": "2508.21663", "title": "Surface Stability Modeling with Universal Machine Learning Interatomic Potentials: A Comprehensive Cleavage Energy Benchmarking Study", "abstract": "Machine learning interatomic potentials (MLIPs) have revolutionized computational materials science by bridging the gap between quantum mechanical accuracy and classical simulation efficiency, enabling unprecedented exploration of materials properties across the periodic table. Despite their remarkable success in predicting bulk properties, no systematic evaluation has assessed how well these universal MLIPs (uMLIPs) can predict cleavage energies, a critical property governing fracture, catalysis, surface stability, and interfacial phenomena. Here, we present a comprehensive benchmark of 19 state-of-the-art uMLIPs for cleavage energy prediction using our previously established density functional theory (DFT) database of 36,718 slab structures spanning elemental, binary, and ternary metallic compounds. We evaluate diverse architectural paradigms, analyzing their performance across chemical compositions, crystal systems, thickness, and surface orientations. Our results reveal that training data composition dominates architectural sophistication: models trained on the Open Materials 2024 (OMat24) dataset, which emphasizes non-equilibrium configurations, achieve mean absolute percentage errors below 6% and correctly identify the thermodynamically most stable surface terminations in 87% of cases, without any explicit surface energy training. In contrast, architecturally identical models trained on equilibrium-only datasets show five-fold higher errors, while models trained on surface-adsorbate data fail catastrophically with a 17-fold degradation. Remarkably, simpler architectures trained on appropriate data achieve comparable accuracy to complex transformers while offering 10-100x computational speedup. These findings show that the community should focus on strategic training data generation that captures the relevant physical phenomena.", "date": "2025-08-29", "categories": []}
{"id": "2511.11361", "title": "Toward Multi-Fidelity Machine Learning Force Field for Cathode Materials", "abstract": "Machine learning force fields (MLFFs), which employ neural networks to map atomic structures to system energies, effectively combine the high accuracy of first-principles calculation with the computational efficiency of empirical force fields. They are widely used in computational materials simulations. However, the development and application of MLFFs for lithium-ion battery cathode materials remain relatively limited. This is primarily due to the complex electronic structure characteristics of cathode materials and the resulting scarcity of high-quality computational datasets available for force field training. In this work, we develop a multi-fidelity machine learning force field framework to enhance the data efficiency of computational results, which can simultaneously utilize both low-fidelity non-magnetic and high-fidelity magnetic computational datasets of cathode materials for training. Tests conducted on the lithium manganese iron phosphate (LMFP) cathode material system demonstrate the effectiveness of this multi-fidelity approach. This work helps to achieve high-accuracy MLFF training for cathode materials at a lower training dataset cost, and offers new perspectives for applying MLFFs to computational simulations of cathode materials.", "date": "2025-11-14", "categories": []}
{"id": "2503.08207", "title": "Are Foundational Atomistic Models Reliable for Finite-Temperature Molecular Dynamics?", "abstract": "Machine learning force fields have emerged as promising tools for molecular dynamics (MD) simulations, potentially offering quantum-mechanical accuracy with the efficiency of classical MD. Inspired by foundational large language models, recent years have seen considerable progress in developing foundational atomistic models, sometimes referred to as universal force fields, designed to cover most elements in the periodic table. This Perspective adopts a practitioner's viewpoint to ask a critical question: Are these foundational atomistic models reliable for one of their most compelling applications, in particular simulating finite-temperature dynamics? Instead of a broad benchmark, we use the canonical ferroelectric-paraelectric phase transition in PbTiO$_3$ as a focused case study to evaluate prominent foundational atomistic models. Our findings suggest a potential disconnect between static accuracy and dynamic reliability. While 0 K properties are often well-reproduced, we observed that the models can struggle to consistently capture the correct phase transition, sometimes exhibiting simulation instabilities. We believe these challenges may stem from inherent biases in training data and a limited description of anharmonicity. These observed shortcomings, though demonstrated on a single system, appear to point to broader, systemic challenges that can be addressed with targeted fine-tuning. This Perspective serves not to rank models, but to initiate a crucial discussion on the practical readiness of foundational atomistic models and to explore future directions for their improvement.", "date": "2025-03-11", "categories": []}
{"id": "2508.20138", "title": "MicroLad: 2D-to-3D Microstructure Reconstruction and Generation via Latent Diffusion and Score Distillation", "abstract": "A major obstacle to establishing reliable structure-property (SP) linkages in materials engineering is the scarcity of diverse 3D microstructure datasets. Limited dataset availability and insufficient control over the analysis and design space restrict the variety of achievable microstructure morphologies, hindering progress in solving the inverse (property-to-structure) design problem. To address these challenges, we introduce MicroLad, a latent diffusion framework specifically designed for reconstructing 3D microstructures from 2D data. Trained on 2D images and employing multi-plane denoising diffusion sampling in the latent space, the framework reliably generates stable and coherent 3D volumes that remain statistically consistent with the original data. While this reconstruction capability enables dimensionality expansion (2D-to-3D) for generating statistically equivalent 3D samples from 2D data, effective exploration of microstructure design requires methods to guide the generation process toward specific objectives. To achieve this, MicroLad integrates score distillation sampling (SDS), which combines a differentiable score loss with microstructural descriptor-matching and property-alignment terms. This approach updates encoded 2D slices of the 3D volume in the latent space, enabling robust inverse-controlled 2D-to-3D microstructure generation. Consequently, the method facilitates exploration of an expanded 3D microstructure analysis and design space in terms of both microstructural descriptors and material properties.", "date": "2025-08-27", "categories": []}
{"id": "2511.09906", "title": "Beyond empirical models: Discovering new constitutive laws in solids with graph-based equation discovery", "abstract": "Constitutive models are fundamental to solid mechanics and materials science, underpinning the quantitative description and prediction of material responses under diverse loading conditions. Traditional phenomenological models, which are derived through empirical fitting, often lack generalizability and rely heavily on expert intuition and predefined functional forms. In this work, we propose a graph-based equation discovery framework for the automated discovery of constitutive laws directly from multisource experimental data. This framework expresses equations as directed graphs, where nodes represent operators and variables, edges denote computational relations, and edge features encode parametric dependencies. This enables the generation and optimization of free-form symbolic expressions with undetermined material-specific parameters. Through the proposed framework, we have discovered new constitutive models for strain-rate effects in alloy steel materials and the deformation behavior of lithium metal. Compared with conventional empirical models, these new models exhibit compact analytical structures and achieve higher accuracy. The proposed graph-based equation discovery framework provides a generalizable and interpretable approach for data-driven scientific modelling, particularly in contexts where traditional empirical formulations are inadequate for representing complex physical phenomena.", "date": "2025-11-13", "categories": []}
{"id": "2511.10108", "title": "MATAI: A Generalist Machine Learning Framework for Property Prediction and Inverse Design of Advanced Alloys", "abstract": "The discovery of advanced metallic alloys is hindered by vast composition spaces, competing property objectives, and real-world constraints on manufacturability. Here we introduce MATAI, a generalist machine learning framework for property prediction and inverse design of as-cast alloys. MATAI integrates a curated alloy database, deep neural network-based property predictors, a constraint-aware optimization engine, and an iterative AI-experiment feedback loop. The framework estimates key mechanical propertie, sincluding density, yield strength, ultimate tensile strength, and elongation, directly from composition, using multi-task learning and physics-informed inductive biases. Alloy design is framed as a constrained optimization problem and solved using a bi-level approach that combines local search with symbolic constraint programming. We demonstrate MATAI's capabilities on the Ti-based alloy system, a canonical class of lightweight structural materials, where it rapidly identifies candidates that simultaneously achieve lower density (<4.45 g/cm3), higher strength (>1000 MPa) and appreciable ductility (>5%) through only seven iterations. Experimental validation confirms that MATAI-designed alloys outperform commercial references such as TC4, highlighting the framework's potential to accelerate the discovery of lightweight, high-performance materials under real-world design constraints.", "date": "2025-11-13", "categories": []}
{"id": "2404.12445", "title": "Adaptive Catalyst Discovery Using Multicriteria Bayesian Optimization with Representation Learning", "abstract": "High-performance catalysts are crucial for sustainable energy conversion and human health. However, the discovery of catalysts faces challenges due to the absence of efficient approaches to navigating vast and high-dimensional structure and composition spaces. In this study, we propose a high-throughput computational catalyst screening approach integrating density functional theory (DFT) and Bayesian Optimization (BO). Within the BO framework, we propose an uncertainty-aware atomistic machine learning model, UPNet, which enables automated representation learning directly from high-dimensional catalyst structures and achieves principled uncertainty quantification. Utilizing a constrained expected improvement acquisition function, our BO framework simultaneously considers multiple evaluation criteria. Using the proposed methods, we explore catalyst discovery for the CO2 reduction reaction. The results demonstrate that our approach achieves high prediction accuracy, facilitates interpretable feature extraction, and enables multicriteria design optimization, leading to significant reduction of computing power and time (10x reduction of required DFT calculations) in high-performance catalyst discovery.", "date": "2024-04-18", "categories": []}
{"id": "2511.08424", "title": "Identification of Empirical Constitutive Models for Age-Hardenable Aluminium Alloy and High-Chromium Martensitic Steel Using Symbolic Regression", "abstract": "Process-structure-property relationships are fundamental in materials science and engineering and are key to the development of new and improved materials. Symbolic regression serves as a powerful tool for uncovering mathematical models that describe these relationships. It can automatically generate equations to predict material behaviour under specific manufacturing conditions and optimize performance characteristics such as strength and elasticity.\n  The present work illustrates how symbolic regression can derive constitutive models that describe the behaviour of various metallic alloys during plastic deformation. Constitutive modelling is a mathematical framework for understanding the relationship between stress and strain in materials under different loading conditions. In this study, two materials (age-hardenable aluminium alloy and high-chromium martensitic steel) and two different testing methods (compression and tension) are considered to obtain the required stress-strain data. The results highlight the benefits of using symbolic regression while also discussing potential challenges.", "date": "2025-11-11", "categories": []}
{"id": "2511.05577", "title": "Fine-Tuning Vision-Language Models for Multimodal Polymer Property Prediction", "abstract": "Vision-Language Models (VLMs) have shown strong performance in tasks like visual question answering and multimodal text generation, but their effectiveness in scientific domains such as materials science remains limited. While some machine learning methods have addressed specific challenges in this field, there is still a lack of foundation models designed for broad tasks like polymer property prediction using multimodal data. In this work, we present a multimodal polymer dataset to fine-tune VLMs through instruction-tuning pairs and assess the impact of multimodality on prediction performance. Our fine-tuned models, using LoRA, outperform unimodal and baseline approaches, demonstrating the benefits of multimodal learning. Additionally, this approach reduces the need to train separate models for different properties, lowering deployment and maintenance costs.", "date": "2025-11-04", "categories": []}
{"id": "2511.06976", "title": "Rethinking Crystal Symmetry Prediction: A Decoupled Perspective", "abstract": "Efficiently and accurately determining the symmetry is a crucial step in the structural analysis of crystalline materials. Existing methods usually mindlessly apply deep learning models while ignoring the underlying chemical rules. More importantly, experiments show that they face a serious sub-property confusion SPC problem. To address the above challenges, from a decoupled perspective, we introduce the XRDecoupler framework, a problem-solving arsenal specifically designed to tackle the SPC problem. Imitating the thinking process of chemists, we innovatively incorporate multidimensional crystal symmetry information as superclass guidance to ensure that the model's prediction process aligns with chemical intuition. We further design a hierarchical PXRD pattern learning model and a multi-objective optimization approach to achieve high-quality representation and balanced optimization. Comprehensive evaluations on three mainstream databases (e.g., CCDC, CoREMOF, and InorganicData) demonstrate that XRDecoupler excels in performance, interpretability, and generalization.", "date": "2025-11-10", "categories": []}
{"id": "2511.07158", "title": "Guiding Generative Models to Uncover Diverse and Novel Crystals via Reinforcement Learning", "abstract": "Discovering functional crystalline materials entails navigating an immense combinatorial design space. While recent advances in generative artificial intelligence have enabled the sampling of chemically plausible compositions and structures, a fundamental challenge remains: the objective misalignment between likelihood-based sampling in generative modelling and targeted focus on underexplored regions where novel compounds reside. Here, we introduce a reinforcement learning framework that guides latent denoising diffusion models toward diverse and novel, yet thermodynamically viable crystalline compounds. Our approach integrates group relative policy optimisation with verifiable, multi-objective rewards that jointly balance creativity, stability, and diversity. Beyond de novo generation, we demonstrate enhanced property-guided design that preserves chemical validity, while targeting desired functional properties. This approach establishes a modular foundation for controllable AI-driven inverse design that addresses the novelty-validity trade-off across scientific discovery applications of generative models.", "date": "2025-11-10", "categories": []}
{"id": "2404.09515", "title": "Revealing the structure-property relationships of copper alloys with FAGC", "abstract": "Cu-Cr-Zr alloys play a crucial role in electronic devices and the electric power industry, where their electrical conductivity and hardness are of great importance. However, due to the scarcity of available samples, there has been a lack of effective studies exploring the relationship between the microstructural images of Cu-Cr-Zr alloys and their key properties. In this paper, the FAGC feature augmentation method is employed to enhance the microstructural images of Cu-Cr-Zr alloys within a feature space known as the pre-shape space. Pseudo-labels are then constructed to expand the number of training samples. These features are then input into various machine learning models to construct performance prediction models for the alloy. Finally, we validate the impact of different machine learning methods and the number of augmented features on prediction accuracy through experiments. Experimental results demonstrate that our method achieves superior performance in predicting electrical conductivity (\\(R^2=0.978\\)) and hardness (\\(R^2=0.998\\)) when using the decision tree classifier with 100 augmented samples. Further analysis reveals that regions with reduced image noise, such as fewer grain or phase boundaries, exhibit higher contributions to electrical conductivity. These findings highlight the potential of the FAGC method in overcoming the challenges of limited image data in materials science, offering a powerful tool for establishing detailed and quantitative relationships between complex microstructures and material properties.", "date": "2024-04-15", "categories": []}
{"id": "2511.03735", "title": "Friction on Demand: A Generative Framework for the Inverse Design of Metainterfaces", "abstract": "Designing frictional interfaces to exhibit prescribed macroscopic behavior is a challenging inverse problem, made difficult by the non-uniqueness of solutions and the computational cost of contact simulations. Traditional approaches rely on heuristic search over low-dimensional parameterizations, which limits their applicability to more complex or nonlinear friction laws. We introduce a generative modeling framework using Variational Autoencoders (VAEs) to infer surface topographies from target friction laws. Trained on a synthetic dataset composed of 200 million samples constructed from a parameterized contact mechanics model, the proposed method enables efficient, simulation-free generation of candidate topographies. We examine the potential and limitations of generative modeling for this inverse design task, focusing on balancing accuracy, throughput, and diversity in the generated solutions. Our results highlight trade-offs and outline practical considerations when balancing these objectives. This approach paves the way for near-real-time control of frictional behavior through tailored surface topographies.", "date": "2025-10-27", "categories": []}
{"id": "2505.11722", "title": "Explainable Machine Learning for Oxygen Diffusion in Perovskites and Pyrochlores", "abstract": "Explainable machine learning can help to discover new physical relationships for material properties. To understand the material properties that govern the activation energy for oxygen diffusion in perovskites and pyrochlores, we build a database of experimental activation energies and apply a grouping algorithm to the material property features. These features are then used to fit seven different machine learning models. An ensemble consensus determines that the most important features for predicting the activation energy are the ionicity of the A-site bond and the partial pressure of oxygen for perovskites. For pyrochlores, the two most important features are the A-site $s$ valence electron count and the B-site electronegativity. The most important features are all constructed using the weighted averages of elemental metal properties, despite weighted averages of the constituent binary oxides being included in our feature set. This is surprising because the material properties of the constituent oxides are more similar to the experimentally measured properties of perovskites and pyrochlores than the features of the metals that are chosen. The easy-to-measure features identified in this work enable rapid screening for new materials with fast oxide-ion diffusivity.", "date": "2025-05-16", "categories": []}
{"id": "2511.03012", "title": "Heterogeneous Metamaterials Design via Multiscale Neural Implicit Representation", "abstract": "Metamaterials are engineered materials composed of specially designed unit cells that exhibit extraordinary properties beyond those of natural materials. Complex engineering tasks often require heterogeneous unit cells to accommodate spatially varying property requirements. However, designing heterogeneous metamaterials poses significant challenges due to the enormous design space and strict compatibility requirements between neighboring cells. Traditional concurrent multiscale design methods require solving an expensive optimization problem for each unit cell and often suffer from discontinuities at cell boundaries. On the other hand, data-driven approaches that assemble structures from a fixed library of microstructures are limited by the dataset and require additional post-processing to ensure seamless connections. In this work, we propose a neural network-based metamaterial design framework that learns a continuous two-scale representation of the structure, thereby jointly addressing these challenges. Central to our framework is a multiscale neural representation in which the neural network takes both global (macroscale) and local (microscale) coordinates as inputs, outputting an implicit field that represents multiscale structures with compatible unit cell geometries across the domain, without the need for a predefined dataset. We use a compatibility loss term during training to enforce connectivity between adjacent unit cells. Once trained, the network can produce metamaterial designs at arbitrarily high resolution, hence enabling infinite upsampling for fabrication or simulation. We demonstrate the effectiveness of the proposed approach on mechanical metamaterial design, negative Poisson's ratio, and mechanical cloaking problems with potential applications in robotics, bioengineering, and aerospace.", "date": "2025-11-04", "categories": []}
{"id": "2511.01913", "title": "Delta-learned force fields for nonbonded interactions: Addressing the strength mismatch between covalent-nonbonded interaction for global models", "abstract": "Noncovalent interactions--vdW dispersion, hydrogen/halogen bonding, ion-$\\pi$, and $\\pi$-stacking--govern structure, dynamics, and emergent phenomena in materials and molecular systems, yet accurately learning them alongside covalent forces remains a core challenge for machine-learned force fields (MLFFs). This challenge is acute for global models that use Coulomb-matrix (CM) descriptors compared under Euclidean/Frobenius metrics in multifragment settings. We show that the mismatch between predominantly covalent force labels and the CM's overrepresentation of intermolecular features biases single-model training and degrades force-field fidelity. To address this, we introduce \\textit{$\\Delta$-sGDML}, a scale-aware formulation within the sGDML framework that explicitly decouples intra- and intermolecular physics by training fragment-specific models alongside a dedicated binding model, then composing them at inference. Across benzene dimers, host-guest complexes (C$_{60}$@buckycatcher, NO$_3^-$@i-corona[6]arene), benzene-water, and benzene-Na$^+$, \\mbox{$\\Delta$-sGDML} delivers consistent gains over a single global model, with fragment-resolved force-error reductions up to \\textbf{75\\%}, without loss of energy accuracy. Furthermore, molecular-dynamics simulations further confirm that the $\\Delta$-model yields a reliable force field for C$_{60}$@buckycatcher, producing stable trajectories across a wide range of temperatures (10-400~K), unlike the single global model, which loses stability above $\\sim$200~K. The method offers a practical route to homogenize per-fragment errors and recover reliable noncovalent physics in global MLFFs.", "date": "2025-11-01", "categories": []}
{"id": "2511.02290", "title": "From data to design: Random forest regression model for predicting mechanical properties of alloy steel", "abstract": "This study investigates the application of Random Forest Regression for predicting mechanical properties of alloy steel-Elongation, Tensile Strength, and Yield Strength-from material composition features including Iron (Fe), Chromium (Cr), Nickel (Ni), Manganese (Mn), Silicon (Si), Copper (Cu), Carbon (C), and deformation percentage during cold rolling. Utilizing a dataset comprising these features, we trained and evaluated the Random Forest model, achieving high predictive performance as evidenced by R2 scores and Mean Squared Errors (MSE). The results demonstrate the model's efficacy in providing accurate predictions, which is validated through various performance metrics including residual plots and learning curves. The findings underscore the potential of ensemble learning techniques in enhancing material property predictions, with implications for industrial applications in material science.", "date": "2025-11-04", "categories": []}
{"id": "2507.16336", "title": "Constructing material network representations for intelligent amorphous alloys design", "abstract": "Designing high-performance amorphous alloys is demanding for various applications. But this process intensively relies on empirical laws and unlimited attempts. The high-cost and low-efficiency nature of the traditional strategies prevents effective sampling in the enormous material space. Here, we propose material networks to accelerate the discovery of binary and ternary amorphous alloys. The network topologies reveal hidden material candidates that were obscured by traditional tabular data representations. By scrutinizing the amorphous alloys synthesized in different years, we construct dynamical material networks to track the history of the alloy discovery. We find that some innovative materials designed in the past were encoded in the networks, demonstrating their predictive power in guiding new alloy design. These material networks show physical similarities with several real-world networks in our daily lives. Our findings pave a new way for intelligent materials design, especially for complex alloys.", "date": "2025-07-22", "categories": []}
{"id": "2511.00204", "title": "Transfer learning discovery of molecular modulators for perovskite solar cells", "abstract": "The discovery of effective molecular modulators is essential for advancing perovskite solar cells (PSCs), but the research process is hindered by the vastness of chemical space and the time-consuming and expensive trial-and-error experimental screening. Concurrently, machine learning (ML) offers significant potential for accelerating materials discovery. However, applying ML to PSCs remains a major challenge due to data scarcity and limitations of traditional quantitative structure-property relationship (QSPR) models. Here, we apply a chemical informed transfer learning framework based on pre-trained deep neural networks, which achieves high accuracy in predicting the molecular modulator's effect on the power conversion efficiency (PCE) of PSCs. This framework is established through systematical benchmarking of diverse molecular representations, enabling lowcost and high-throughput virtual screening over 79,043 commercially available molecules. Furthermore, we leverage interpretability techniques to visualize the learned chemical representation and experimentally characterize the resulting modulator-perovskite interactions. The top molecular modulators identified by the framework are subsequently validated experimentally, delivering a remarkably improved champion PCE of 26.91% in PSCs.", "date": "2025-10-31", "categories": []}
{"id": "2411.14390", "title": "Persistent Homology for Structural Characterization in Disordered Systems", "abstract": "We propose a unified framework based on persistent homology (PH) to characterize both local and global structures in disordered systems. It can simultaneously generate local and global descriptors using the same algorithm and data structure, and has shown to be highly effective and interpretable in predicting particle rearrangements and classifying global phases. We also demonstrated that using a single variable enables a linear SVM to achieve nearly perfect three-phase classification. Inspired by this discovery, we define a non-parametric metric, the Separation Index (SI), which not only achieves this classification without sacrificing significant performance but also establishes a connection between particle environments and the global phase structure. Our methods provide an effective framework for understanding and analyzing the properties of disordered materials, with broad potential applications in materials science and even wider studies of complex systems.", "date": "2024-11-21", "categories": []}
{"id": "2510.23371", "title": "Towards a Generalizable AI for Materials Discovery: Validation through Immersion Coolant Screening", "abstract": "Artificial intelligence (AI) has emerged as a powerful accelerator of materials discovery, yet most existing models remain problem-specific, requiring additional data collection and retraining for each new property. Here we introduce and validate GATE (Geometrically Aligned Transfer Encoder) -- a generalizable AI framework that jointly learns 34 physicochemical properties spanning thermal, electrical, mechanical, and optical domains. By aligning these properties within a shared geometric space, GATE captures cross-property correlations that reduce disjoint-property bias -- a key factor causing false positives in multi-criteria screening. To demonstrate its generalizable utility, GATE -- without any problem-specific model reconfiguration -- applied to the discovery of immersion cooling fluids for data centers, a stringent real-world challenge defined by the Open Compute Project (OCP). Screening billions of candidates, GATE identified 92,861 molecules as promising for practical deployment. Four were experimentally or literarily validated, showing strong agreement with wet-lab measurements and performance comparable to or exceeding a commercial coolant. These results establish GATE as a generalizable AI platform readily applicable across diverse materials discovery tasks.", "date": "2025-10-27", "categories": []}
{"id": "2510.25788", "title": "SHA-256 Infused Embedding-Driven Generative Modeling of High-Energy Molecules in Low-Data Regimes", "abstract": "High-energy materials (HEMs) are critical for propulsion and defense domains, yet their discovery remains constrained by experimental data and restricted access to testing facilities. This work presents a novel approach toward high-energy molecules by combining Long Short-Term Memory (LSTM) networks for molecular generation and Attentive Graph Neural Networks (GNN) for property predictions. We propose a transformative embedding space construction strategy that integrates fixed SHA-256 embeddings with partially trainable representations. Unlike conventional regularization techniques, this changes the representational basis itself, reshaping the molecular input space before learning begins. Without recourse to pretraining, the generator achieves 67.5% validity and 37.5% novelty. The generated library exhibits a mean Tanimoto coefficient of 0.214 relative to training set signifying the ability of framework to generate a diverse chemical space. We identified 37 new super explosives higher than 9 km/s predicted detonation velocity.", "date": "2025-10-28", "categories": []}
{"id": "2508.01822", "title": "Deep Learning-Driven Prediction of Microstructure Evolution via Latent Space Interpolation", "abstract": "Phase-field models accurately simulate microstructure evolution, but their dependence on solving complex differential equations makes them computationally expensive. This work achieves a significant acceleration via a novel deep learning-based framework, utilizing a Conditional Variational Autoencoder (CVAE) coupled with Cubic Spline Interpolation and Spherical Linear Interpolation (SLERP). We demonstrate the method for binary spinodal decomposition by predicting microstructure evolution for intermediate alloy compositions from a limited set of training compositions. First, using microstructures from phase-field simulations of binary spinodal decomposition, we train the CVAE, which learns compact latent representations that encode essential morphological features. Next, we use cubic spline interpolation in the latent space to predict microstructures for any unknown composition. Finally, SLERP ensures smooth morphological evolution with time that closely resembles coarsening. The predicted microstructures exhibit high visual and statistical similarity to phase-field simulations. This framework offers a scalable and efficient surrogate model for microstructure evolution, enabling accelerated materials design and composition optimization.", "date": "2025-08-03", "categories": []}
{"id": "2504.07297", "title": "Data Fusion of Deep Learned Molecular Embeddings for Property Prediction", "abstract": "Data-driven approaches such as deep learning can result in predictive models for material properties with exceptional accuracy and efficiency. However, in many applications, data is sparse, severely limiting their accuracy and applicability. To improve predictions, techniques such as transfer learning and multitask learning have been used. The performance of multitask learning models depends on the strength of the underlying correlations between tasks and the completeness of the data set. Standard multitask models tend to underperform when trained on sparse data sets with weakly correlated properties. To address this gap, we fuse deep-learned embeddings generated by independent pretrained single-task models, resulting in a multitask model that inherits rich, property-specific representations. By reusing (rather than retraining) these embeddings, the resulting fused model outperforms standard multitask models and can be extended with fewer trainable parameters. We demonstrate this technique on a widely used benchmark data set of quantum chemistry data for small molecules as well as a newly compiled sparse data set of experimental data collected from literature and our own quantum chemistry and thermochemical calculations.", "date": "2025-04-09", "categories": []}
{"id": "2412.11569", "title": "The dark side of the forces: assessing non-conservative force models for atomistic machine learning", "abstract": "The use of machine learning to estimate the energy of a group of atoms, and the forces that drive them to more stable configurations, has revolutionized the fields of computational chemistry and materials discovery. In this domain, rigorous enforcement of symmetry and conservation laws has traditionally been considered essential. For this reason, interatomic forces are usually computed as the derivatives of the potential energy, ensuring energy conservation. Several recent works have questioned this physically constrained approach, suggesting that directly predicting the forces yields a better trade-off between accuracy and computational efficiency, and that energy conservation can be learned during training. This work investigates the applicability of such non-conservative models in microscopic simulations. We identify and demonstrate several fundamental issues, from ill-defined convergence of geometry optimization to instability in various types of molecular dynamics. Given the difficulty in monitoring and correcting the lack of energy conservation, direct forces should be used with great care. We show that the best approach to exploit the acceleration they afford is to use them in conjunction with conservative forces. A model can be pre-trained efficiently on direct forces, then fine-tuned using backpropagation. At evaluation time, both force types can be used together to avoid unphysical effects while still benefitting almost entirely from the computational efficiency of direct forces.", "date": "2024-12-16", "categories": []}
{"id": "2504.20940", "title": "Energy-Based Coarse-Graining in Molecular Dynamics: A Flow-Based Framework without Data", "abstract": "Coarse-grained (CG) models provide an effective route to reducing the complexity of molecular simulations (MD), but conventional approaches depend heavily on long all-atom MD trajectories to adequately sample configurational space. This data dependence limits accuracy and generalizability, as unvisited configurations remain excluded from the resulting CG models. We introduce a fully data-free, generative framework for CG that directly targets the all-atom Boltzmann distribution. The model defines a structured latent space comprising slow collective variables, associated with multimodal marginal densities capturing metastable states, and fast variables, represented through simple, unimodal conditional distributions. A learnable, bijective map from latent space to atomistic coordinates enables the automatic and accurate reconstruction of molecular structures. Training relies solely on the interatomic potential and minimizes the reverse Kullback-Leibler (KL) divergence via an energy-based objective. To stabilize optimization and ensure mode coverage, we employ an adaptive tempering scheme that promotes the exploration of diverse configurations. Once trained, the model can generate independent, one-shot equilibrium samples at full atomic resolution. Validation on two synthetic systems, a double-well potential and a Gaussian mixture model, as well as on the benchmark alanine dipeptide, demonstrates that the method captures all relevant modes of the Boltzmann distribution, reconstructs atomic configurations, and automatically learns physically meaningful CG representations. These results suggest a promising, data-free alternative to traditional CG techniques, offering both a principled approach to addressing the long-standing \"chicken-and-egg\" challenge in coarse-graining and an effective solution to the back-mapping problem by enabling accurate reconstruction of all-atom configurations.", "date": "2025-04-29", "categories": []}
{"id": "2510.22424", "title": "Reinforcement learning-guided optimization of critical current in high-temperature superconductors", "abstract": "High-temperature superconductors are essential for next-generation energy and quantum technologies, yet their performance is often limited by the critical current density ($J_c$), which is strongly influenced by microstructural defects. Optimizing $J_c$ through defect engineering is challenging due to the complex interplay of defect type, density, and spatial correlation. Here we present an integrated workflow that combines reinforcement learning (RL) with time-dependent Ginzburg-Landau (TDGL) simulations to autonomously identify optimal defect configurations that maximize $J_c$. In our framework, TDGL simulations generate current-voltage characteristics to evaluate $J_c$, which serves as the reward signal that guides the RL agent to iteratively refine defect configurations. We find that the agent discovers optimal defect densities and correlations in two-dimensional thin-film geometries, enhancing vortex pinning and $J_c$ relative to the pristine thin-film, approaching 60\\% of theoretical depairing limit with up to 15-fold enhancement compared to random initialization. This RL-driven approach provides a scalable strategy for defect engineering, with broad implications for advancing HTS applications in fusion magnets, particle accelerators, and other high-field technologies.", "date": "2025-10-25", "categories": []}
{"id": "2510.23040", "title": "LLM Meets Diffusion: A Hybrid Framework for Crystal Material Generation", "abstract": "Recent advances in generative modeling have shown significant promise in designing novel periodic crystal structures. Existing approaches typically rely on either large language models (LLMs) or equivariant denoising models, each with complementary strengths: LLMs excel at handling discrete atomic types but often struggle with continuous features such as atomic positions and lattice parameters, while denoising models are effective at modeling continuous variables but encounter difficulties in generating accurate atomic compositions. To bridge this gap, we propose CrysLLMGen, a hybrid framework that integrates an LLM with a diffusion model to leverage their complementary strengths for crystal material generation. During sampling, CrysLLMGen first employs a fine-tuned LLM to produce an intermediate representation of atom types, atomic coordinates, and lattice structure. While retaining the predicted atom types, it passes the atomic coordinates and lattice structure to a pre-trained equivariant diffusion model for refinement. Our framework outperforms state-of-the-art generative models across several benchmark tasks and datasets. Specifically, CrysLLMGen not only achieves a balanced performance in terms of structural and compositional validity but also generates more stable and novel materials compared to LLM-based and denoisingbased models Furthermore, CrysLLMGen exhibits strong conditional generation capabilities, effectively producing materials that satisfy user-defined constraints. Code is available at https://github.com/kdmsit/crysllmgen", "date": "2025-10-27", "categories": []}
{"id": "2505.10994", "title": "Space Group Equivariant Crystal Diffusion", "abstract": "Accelerating inverse design of crystalline materials with generative models has significant implications for a range of technologies. Unlike other atomic systems, 3D crystals are invariant to discrete groups of isometries called the space groups. Crucially, these space group symmetries are known to heavily influence materials properties. We propose SGEquiDiff, a crystal generative model which naturally handles space group constraints with space group invariant likelihoods. SGEquiD-iff consists of an SE(3)-invariant, telescoping discrete sampler of crystal lattices; permutation-invariant, transformer-based autoregressive sampling of Wyckoff positions, elements, and numbers of symmetrically unique atoms; and space group equivariant diffusion of atomic coordinates. We show that space group equivariant vector fields automatically live in the tangent spaces of the Wyckoff positions. SGEquiDiff achieves state-of-the-art performance on standard benchmark datasets as assessed by quantitative proxy metrics and quantum mechanical calculations. Our code is available at https://github.com/rees-c/sgequidiff.", "date": "2025-05-16", "categories": []}
{"id": "2510.20976", "title": "L^2M^3OF: A Large Language Multimodal Model for Metal-Organic Frameworks", "abstract": "Large language models have demonstrated remarkable reasoning capabilities across diverse natural language tasks. However, comparable breakthroughs in scientific discovery are more limited, because understanding complex physical phenomena demands multifaceted representations far beyond language alone. A compelling example is the design of functional materials such as MOFs-critical for a range of impactful applications like carbon capture and hydrogen storage. Navigating their vast and intricate design space in language-based representations interpretable by LLMs is challenging due to the numerous possible three-dimensional atomic arrangements and strict reticular rules of coordination geometry and topology. Despite promising early results in LLM-assisted discovery for simpler materials systems, MOF design remains heavily reliant on tacit human expertise rarely codified in textual information alone. To overcome this barrier, we introduce L2M3OF, the first multimodal LLM for MOFs. L2M3OF integrates crystal representation learning with language understanding to process structural, textual, and knowledge modalities jointly. L2M3OF employs a pre-trained crystal encoder with a lightweight projection layer to compress structural information into a token space, enabling efficient alignment with language instructions. To facilitate training and evaluation, we curate a structure-property-knowledge database of crystalline materials and benchmark L2M3OF against state-of-the-art closed-source LLMs such as GPT-5, Gemini-2.5-Pro and DeepSeek-R1. Experiments show that L2M3OF outperforms leading text-based closed-source LLMs in property prediction and knowledge generation tasks, despite using far fewer parameters. These results highlight the importance of multimodal approaches for porous material understanding and establish L2M3OF as a foundation for next-generation AI systems in materials discovery.", "date": "2025-10-23", "categories": []}
{"id": "2501.03456", "title": "Text to Band Gap: Pre-trained Language Models as Encoders for Semiconductor Band Gap Prediction", "abstract": "We investigate transformer-based language models, including RoBERTa, T5, Llama-3, and MatSciBERT, for predicting the band gaps of semiconductor materials directly from textual descriptions. The inputs encode key material features, such as chemical composition, crystal system, space group, and other structural and electronic properties. Unlike shallow machine learning models, which require extensive feature engineering, or Graph Neural Networks, which rely on graph representations derived from atomic coordinates, pretrained language models can process textual inputs directly, eliminating the need for manual feature preprocessing or structure-based encoding. Material descriptions were constructed in two formats: structured strings with a consistent template and natural language narratives generated via the ChatGPT API. Each model was augmented with a custom regression head and finetuned for band gap prediction task. Language models of different architectures and parameter sizes were all able to predict band gaps from human-readable text with strong accuracy, achieving MAEs in the range of 0.25-0.33 eV, highlighting the success of this approach for scientific regression tasks. Finetuned Llama-3, with 1.2 billion parameters, achieved the highest accuracy (MAE 0.248 eV, R2 0.891). MatSciBERT, pretrained on materials science literature, reached comparable performance (MAE 0.288 eV, R2 0.871) with significantly fewer parameters (110 million), emphasizing the importance of domain-specific pretraining. Attention analysis shows that both models selectively focus on compositional and spin-related features while de-emphasizing geometric features, reflecting the difficulty of capturing spatial information from text. These results establish that pretrained language models can effectively extract complex feature-property relationships from textual material descriptions.", "date": "2025-01-07", "categories": []}
{"id": "2411.06565", "title": "Foundation Model for Composite Microstructures: Reconstruction, Stiffness, and Nonlinear Behavior Prediction", "abstract": "We present the Material Masked Autoencoder (MMAE), a self-supervised Vision Transformer pretrained on a large corpus of short-fiber composite images via masked image reconstruction. The pretrained MMAE learns latent representations that capture essential microstructural features and are broadly transferable across tasks. We demonstrate two key applications: (i) predicting homogenized stiffness components through fine-tuning on limited data, and (ii) inferring physically interpretable parameters by coupling MMAE with an interaction-based material network (IMN), thereby enabling extrapolation of nonlinear stress-strain responses. These results highlight the promise of microstructure foundation models and lay the groundwork for future extensions to more complex systems, such as 3D composites and experimental datasets.", "date": "2024-11-10", "categories": []}
{"id": "2501.13373", "title": "Advancing Carbon Capture using AI: Design of permeable membrane and estimation of parameters for Carbon Capture using linear regression and membrane-based equations", "abstract": "This study focuses on membrane-based systems for CO$_2$ separation, addressing the urgent need for efficient carbon capture solutions to mitigate climate change. Linear regression models, based on membrane equations, were utilized to estimate key parameters, including porosity ($\\epsilon$) of 0.4805, Kozeny constant (K) of 2.9084, specific surface area ($\\sigma$) of 105.3272 m$^2$/m$^3$, mean pressure (Pm) of 6.2166 MPa, viscosity ($\\mu$) of 0.1997 Ns/m$^2$, and gas flux (Jg) of 3.2559 kg m$^{-2}$ s$^{-1}$. These parameters were derived from the analysis of synthetic datasets using linear regression. The study also provides insights into the performance of the membrane, with a flow rate (Q) of 9.8778 $\\times$ 10$^{-4}$ m$^3$/s, an injection pressure (P$_1$) of 2.8219 MPa, and an exit pressure (P$_2$) of 2.5762 MPa. The permeability value of 0.045 for CO$_2$ indicates the potential for efficient separation. Optimizing membrane properties to selectively block CO$_2$ while allowing other gases to pass is crucial for improving carbon capture efficiency. By integrating these technologies into industrial processes, significant reductions in greenhouse gas emissions can be achieved, fostering a circular carbon economy and contributing to global climate goals. This study also explores how artificial intelligence (AI) can aid in designing membranes for carbon capture, addressing the global climate change challenge and supporting the Sustainable Development Goals (SDGs) set by the United Nations.", "date": "2025-01-23", "categories": []}
{"id": "2505.18817", "title": "High-order Equivariant Flow Matching for Density Functional Theory Hamiltonian Prediction", "abstract": "Density functional theory (DFT) is a fundamental method for simulating quantum chemical properties, but it remains expensive due to the iterative self-consistent field (SCF) process required to solve the Kohn-Sham equations. Recently, deep learning methods are gaining attention as a way to bypass this step by directly predicting the Hamiltonian. However, they rely on deterministic regression and do not consider the highly structured nature of Hamiltonians. In this work, we propose QHFlow, a high-order equivariant flow matching framework that generates Hamiltonian matrices conditioned on molecular geometry. Flow matching models continuous-time trajectories between simple priors and complex targets, learning the structured distributions over Hamiltonians instead of direct regression. To further incorporate symmetry, we use a neural architecture that predicts SE(3)-equivariant vector fields, improving accuracy and generalization across diverse geometries. To further enhance physical fidelity, we additionally introduce a fine-tuning scheme to align predicted orbital energies with the target. QHFlow achieves state-of-the-art performance, reducing Hamiltonian error by 71% on MD17 and 53% on QH9. Moreover, we further show that QHFlow accelerates the DFT process without trading off the solution quality when initializing SCF iterations with the predicted Hamiltonian, significantly reducing the number of iterations and runtime.", "date": "2025-05-24", "categories": []}
{"id": "2510.19251", "title": "Synthesizability Prediction of Crystalline Structures with a Hierarchical Transformer and Uncertainty Quantification", "abstract": "Predicting which hypothetical inorganic crystals can be experimentally realized remains a central challenge in accelerating materials discovery. SyntheFormer is a positive-unlabeled framework that learns synthesizability directly from crystal structure, combining a Fourier-transformed crystal periodicity (FTCP) representation with hierarchical feature extraction, Random-Forest feature selection, and a compact deep MLP classifier. The model is trained on historical data from 2011 through 2018 and evaluated prospectively on future years from 2019 to 2025, where the positive class constitutes only 1.02 per cent of samples. Under this temporally separated evaluation, SyntheFormer achieves a test area under the ROC curve of 0.735 and, with dual-threshold calibration, attains high-recall screening with 97.6 per cent recall at 94.2 per cent coverage, which minimizes missed opportunities while preserving discriminative power. Crucially, the model recovers experimentally confirmed metastable compounds that lie far from the convex hull and simultaneously assigns low scores to many thermodynamically stable yet unsynthesized candidates, demonstrating that stability alone is insufficient to predict experimental attainability. By aligning structure-aware representation with uncertainty-aware decision rules, SyntheFormer provides a practical route to prioritize synthesis targets and focus laboratory effort on the most promising new inorganic materials.", "date": "2025-10-22", "categories": []}
{"id": "2410.01464", "title": "Flow Matching for Accelerated Simulation of Atomic Transport in Crystalline Materials", "abstract": "Atomic transport underpins the performance of materials in technologies such as energy storage and electronics, yet its simulation remains computationally demanding. In particular, modeling ionic diffusion in solid-state electrolytes (SSEs) requires methods that can overcome the scale limitations of traditional ab initio molecular dynamics (AIMD). We introduce LiFlow, a generative framework to accelerate MD simulations for crystalline materials that formulates the task as conditional generation of atomic displacements. The model uses flow matching, with a Propagator submodel to generate atomic displacements and a Corrector to locally correct unphysical geometries, and incorporates an adaptive prior based on the Maxwell-Boltzmann distribution to account for chemical and thermal conditions. We benchmark LiFlow on a dataset comprising 25-ps trajectories of lithium diffusion across 4,186 SSE candidates at four temperatures. The model obtains a consistent Spearman rank correlation of 0.7-0.8 for lithium mean squared displacement (MSD) predictions on unseen compositions. Furthermore, LiFlow generalizes from short training trajectories to larger supercells and longer simulations while maintaining high accuracy. With speed-ups of up to 600,000$\\times$ compared to first-principles methods, LiFlow enables scalable simulations at significantly larger length and time scales.", "date": "2024-10-02", "categories": []}
{"id": "2510.16165", "title": "AtomBench: A Benchmark for Generative Atomic Structure Models using GPT, Diffusion, and Flow Architectures", "abstract": "Generative models have become significant assets in the exploration and identification of new materials, enabling the rapid proposal of candidate crystal structures that satisfy target properties. Despite the increasing adoption of diverse architectures, a rigorous comparative evaluation of their performance on materials datasets is lacking. In this work, we present a systematic benchmark of three representative generative models- AtomGPT (a transformer-based model), Crystal Diffusion Variational Autoencoder (CDVAE), and FlowMM (a Riemannian flow matching model). These models were trained to reconstruct crystal structures from subsets of two publicly available superconductivity datasets- JARVIS Supercon 3D and DS A/B from the Alexandria database. Performance was assessed using the Kullback-Leibler (KL) divergence between predicted and reference distributions of lattice parameters, as well as the mean absolute error (MAE) of individual lattice constants. For the computed KLD and MAE scores, CDVAE performs most favorably, followed by AtomGPT, and then FlowMM. All benchmarking code and model configurations will be made publicly available at https://github.com/atomgptlab/atombench_inverse.", "date": "2025-10-17", "categories": []}
{"id": "2411.09896", "title": "Revealing the Evolution of Order in Materials Microstructures Using Multi-Modal Computer Vision", "abstract": "The development of high-performance materials for microelectronics, energy storage, and extreme environments depends on our ability to describe and direct property-defining microstructural order. Our present understanding is typically derived from laborious manual analysis of imaging and spectroscopy data, which is difficult to scale, challenging to reproduce, and lacks the ability to reveal latent associations needed for mechanistic models. Here, we demonstrate a multi-modal machine learning (ML) approach to describe order from electron microscopy analysis of the complex oxide La$_{1-x}$Sr$_x$FeO$_3$. We construct a hybrid pipeline based on fully and semi-supervised classification, allowing us to evaluate both the characteristics of each data modality and the value each modality adds to the ensemble. We observe distinct differences in the performance of uni- and multi-modal models, from which we draw general lessons in describing crystal order using computer vision.", "date": "2024-11-15", "categories": []}
{"id": "2506.17719", "title": "Clarifying the Ti-V Phase Diagram Using First-Principles Calculations and Bayesian Learning", "abstract": "Conflicting experiments disagree on whether the titanium-vanadium (Ti-V) binary alloy exhibits a body-centred cubic (BCC) miscibility gap or remains completely soluble. A leading hypothesis attributes the miscibility gap to oxygen contamination during alloy preparation. To resolve this disagreement, we use an ab initio + machine-learning workflow that couples an actively-trained Moment Tensor Potential with Bayesian inference of free energy surface. This workflow enables construction of the Ti-V phase diagram across the full composition range with systematically reduced statistical and finite-size errors. The resulting diagram reproduces all experimental features, demonstrating the robustness of our approach, and clearly favors the variant with a BCC miscibility gap terminating at T = 980 K and c = 0.67. Because our simulations model a perfectly oxygen-free Ti-V system, the observed gap cannot originate from impurity effects, in contrast to recent CALPHAD reassessments.", "date": "2025-06-21", "categories": []}
{"id": "2508.20875", "title": "LeMat-Traj: A Scalable and Unified Dataset of Materials Trajectories for Atomistic Modeling", "abstract": "The development of accurate machine learning interatomic potentials (MLIPs) is limited by the fragmented availability and inconsistent formatting of quantum mechanical trajectory datasets derived from Density Functional Theory (DFT). These datasets are expensive to generate yet difficult to combine due to variations in format, metadata, and accessibility. To address this, we introduce LeMat-Traj, a curated dataset comprising over 120 million atomic configurations aggregated from large-scale repositories, including the Materials Project, Alexandria, and OQMD. LeMat-Traj standardizes data representation, harmonizes results and filters for high-quality configurations across widely used DFT functionals (PBE, PBESol, SCAN, r2SCAN). It significantly lowers the barrier for training transferrable and accurate MLIPs. LeMat-Traj spans both relaxed low-energy states and high-energy, high-force structures, complementing molecular dynamics and active learning datasets. By fine-tuning models pre-trained on high-force data with LeMat-Traj, we achieve a significant reduction in force prediction errors on relaxation tasks. We also present LeMaterial-Fetcher, a modular and extensible open-source library developed for this work, designed to provide a reproducible framework for the community to easily incorporate new data sources and ensure the continued evolution of large-scale materials datasets. LeMat-Traj and LeMaterial-Fetcher are publicly available at https://huggingface.co/datasets/LeMaterial/LeMat-Traj and https://github.com/LeMaterial/lematerial-fetcher.", "date": "2025-08-28", "categories": []}
{"id": "2509.00090", "title": "Migration as a Probe: A Generalizable Benchmark Framework for Specialist vs. Generalist Machine-Learned Force Fields", "abstract": "Machine-learned force fields (MLFFs), especially pre-trained foundation models, are transforming computational materials science by enabling ab initio-level accuracy at molecular dynamics scales. Yet their rapid rise raises a key question: should researchers train specialist models from scratch, fine-tune generalist foundation models, or use hybrid approaches? The trade-offs in data efficiency, accuracy, cost, and robustness to out-of-distribution failure remain unclear. We introduce a benchmarking framework using defect migration pathways, evaluated through nudged elastic band trajectories, as diagnostic probes that test both interpolation and extrapolation. Using Cr-doped Sb2Te3 as a representative two-dimensional material, we benchmark multiple training paradigms within the MACE architecture across equilibrium, kinetic (atomic migration), and mechanical (interlayer sliding) tasks. Fine-tuned models substantially outperform from-scratch and zero-shot approaches for kinetic properties but show partial loss of long-range physics. Representational analysis reveals distinct, non-overlapping latent encodings, indicating that different training strategies learn different aspects of system physics. This framework provides practical guidelines for MLFF development and establishes migration-based probes as efficient diagnostics linking performance to learned representations, guiding future uncertainty-aware active learning.", "date": "2025-08-27", "categories": []}
{"id": "2505.14606", "title": "Electrostatics from Laplacian Eigenbasis for Neural Network Interatomic Potentials", "abstract": "In this work, we introduce Phi-Module, a universal plugin module that enforces Poisson's equation within the message-passing framework to learn electrostatic interactions in a self-supervised manner. Specifically, each atom-wise representation is encouraged to satisfy a discretized Poisson's equation, making it possible to acquire a potential {\\phi} and corresponding charges \\r{ho} linked to the learnable Laplacian eigenbasis coefficients of a given molecular graph. We then derive an electrostatic energy term, crucial for improved total energy predictions. This approach integrates seamlessly into any existing neural potential with insignificant computational overhead. Our results underscore how embedding a first-principles constraint in neural interatomic potentials can significantly improve performance while remaining hyperparameter-friendly, memory-efficient, and lightweight in training. Code will be available at https://github.com/dunnolab/phi-module.", "date": "2025-05-20", "categories": []}
{"id": "2503.02407", "title": "Wyckoff Transformer: Generation of Symmetric Crystals", "abstract": "Crystal symmetry plays a fundamental role in determining its physical, chemical, and electronic properties such as electrical and thermal conductivity, optical and polarization behavior, and mechanical strength. Almost all known crystalline materials have internal symmetry. However, this is often inadequately addressed by existing generative models, making the consistent generation of stable and symmetrically valid crystal structures a significant challenge. We introduce WyFormer, a generative model that directly tackles this by formally conditioning on space group symmetry. It achieves this by using Wyckoff positions as the basis for an elegant, compressed, and discrete structure representation. To model the distribution, we develop a permutation-invariant autoregressive model based on the Transformer encoder and an absence of positional encoding. Extensive experimentation demonstrates WyFormer's compelling combination of attributes: it achieves best-in-class symmetry-conditioned generation, incorporates a physics-motivated inductive bias, produces structures with competitive stability, predicts material properties with competitive accuracy even without atomic coordinates, and exhibits unparalleled inference speed.", "date": "2025-03-04", "categories": []}
{"id": "2312.02796", "title": "Materials Expert-Artificial Intelligence for Materials Discovery", "abstract": "The advent of material databases provides an unprecedented opportunity to uncover predictive descriptors for emergent material properties from vast data space. However, common reliance on high-throughput ab initio data necessarily inherits limitations of such data: mismatch with experiments. On the other hand, experimental decisions are often guided by an expert's intuition honed from experiences that are rarely articulated. We propose using machine learning to \"bottle\" such operational intuition into quantifiable descriptors using expertly curated measurement-based data. We introduce \"Materials Expert-Artificial Intelligence\" (ME-AI) to encapsulate and articulate this human intuition. As a first step towards such a program, we focus on the topological semimetal (TSM) among square-net materials as the property inspired by the expert-identified descriptor based on structural information: the tolerance factor. We start by curating a dataset encompassing 12 primary features of 879 square-net materials, using experimental data whenever possible. We then use Dirichlet-based Gaussian process regression using a specialized kernel to reveal composite descriptors for square-net topological semimetals. The ME-AI learned descriptors independently reproduce expert intuition and expand upon it. Specifically, new descriptors point to hypervalency as a critical chemical feature predicting TSM within square-net compounds. Our success with a carefully defined problem points to the \"machine bottling human insight\" approach as promising for machine learning-aided material discovery.", "date": "2023-12-05", "categories": []}
{"id": "2508.16067", "title": "Training a Foundation Model for Materials on a Budget", "abstract": "Foundation models for materials modeling are advancing quickly, but their training remains expensive, often placing state-of-the-art methods out of reach for many research groups. We introduce Nequix, a compact E(3)-equivariant potential that pairs a simplified NequIP design with modern training practices, including equivariant root-mean-square layer normalization and the Muon optimizer, to retain accuracy while substantially reducing compute requirements. Nequix has 700K parameters and was trained in 100 A100 GPU-hours. On the Matbench-Discovery and MDR Phonon benchmarks, Nequix ranks third overall while requiring a 20 times lower training cost than most other methods, and it delivers two orders of magnitude faster inference speed than the current top-ranked model. We release model weights and fully reproducible codebase at https://github.com/atomicarchitects/nequix.", "date": "2025-08-22", "categories": []}
{"id": "2510.07474", "title": "Surrogate Modeling for the Design of Optimal Lattice Structures using Tensor Completion", "abstract": "When designing new materials, it is often necessary to design a material with specific desired properties. Unfortunately, as new design variables are added, the search space grows exponentially, which makes synthesizing and validating the properties of each material very impractical and time-consuming. In this work, we focus on the design of optimal lattice structures with regard to mechanical performance. Computational approaches, including the use of machine learning (ML) methods, have shown improved success in accelerating materials design. However, these ML methods are still lacking in scenarios when training data (i.e. experimentally validated materials) come from a non-uniformly random sampling across the design space. For example, an experimentalist might synthesize and validate certain materials more frequently because of convenience. For this reason, we suggest the use of tensor completion as a surrogate model to accelerate the design of materials in these atypical supervised learning scenarios. In our experiments, we show that tensor completion is superior to classic ML methods such as Gaussian Process and XGBoost with biased sampling of the search space, with around 5\\% increased $R^2$. Furthermore, tensor completion still gives comparable performance with a uniformly random sampling of the entire search space.", "date": "2025-10-08", "categories": []}
{"id": "2502.20933", "title": "MatLLMSearch: Crystal Structure Discovery with Evolution-Guided Large Language Models", "abstract": "Crystal structure generation is fundamental to materials science, enabling the discovery of novel materials with desired properties. While existing approaches leverage Large Language Models (LLMs) through extensive fine-tuning on materials databases, we show that pre-trained LLMs can inherently generate novel and stable crystal structures without additional fine-tuning. Our framework employs LLMs as intelligent proposal agents within an evolutionary pipeline that guides them to perform implicit crossover and mutation operations while maintaining chemical validity. We demonstrate that MatLLMSearch achieves a 78.38% metastable rate validated by machine learning interatomic potentials and 31.7% DFT-verified stability, outperforming specialized models such as CrystalTextLLM. Beyond crystal structure generation, we further demonstrate that our framework adapts to diverse materials design tasks, including crystal structure prediction and multi-objective optimization of properties such as deformation energy and bulk modulus, all without fine-tuning. These results establish our framework as a versatile and effective framework for consistent high-quality materials discovery, offering training-free generation of novel stable structures with reduced overhead and broader accessibility.", "date": "2025-02-28", "categories": []}
{"id": "2408.03508", "title": "On-Demand Growth of Semiconductor Heterostructures Guided by Physics-Informed Machine Learning", "abstract": "Developing tailored semiconductor heterostructures on demand represents a critical capability for addressing the escalating performance demands in electronic and optoelectronic devices. However, traditional fabrication methods remain constrained by simulation-based design and iterative trial-and-error optimization. Here, we introduce SemiEpi, a self-driving platform designed for molecular beam epitaxy (MBE) to perform multi-step semiconductor heterostructure growth through in-situ monitoring and on-the-fly feedback control. By integrating standard MBE reactors, physics-informed machine learning (ML) models, and parameter initialization, SemiEpi identifies optimal initial conditions and proposes experiments for heterostructure growth, eliminating the need for extensive expertise in MBE processes. As a proof of concept, we demonstrate the optimization of high-density InAs quantum dot (QD) growth with a target emission wavelength of 1240 nm, showcasing the power of SemiEpi. We achieve a QD density of 5 x 10^10 cm^-2, a 1.6-fold increase in photoluminescence (PL) intensity, and a reduced full width at half maximum (FWHM) of 29.13 meV, leveraging in-situ reflective high-energy electron diffraction monitoring with feedback control for adjusting growth temperatures. Taken together, our results highlight the potential of ML-guided systems to address challenges in multi-step heterostructure growth, facilitate the development of a hardware-independent framework, and enhance process repeatability and stability, even without exhaustive knowledge of growth parameters.", "date": "2024-08-07", "categories": []}
{"id": "2503.22962", "title": "Multimodal machine learning with large language embedding model for polymer property prediction", "abstract": "Contemporary large language models (LLMs), such as GPT-4 and Llama, have harnessed extensive computational power and diverse text corpora to achieve remarkable proficiency in interpreting and generating domain-specific content, including materials science. To leverage the domain knowledge embedded within these models, we propose a simple yet effective multimodal architecture, PolyLLMem, which integrates text embeddings generated by Llama 3 with molecular structure embeddings derived from Uni-Mol, for polymer properties prediction tasks. In our model, Low-rank adaptation (LoRA) layers were also incorporated during the property prediction tasks to refine the embeddings based on our limited polymer dataset, thereby enhancing their chemical relevance for polymer SMILES representation. This balanced fusion of fine-tuned textual and structural information enables PolyLLMem to accurately predict a variety of polymer properties despite the scarcity of training data. Its performance is comparable to, and in some cases exceeds, that of graph-based models, as well as transformer-based models that typically require pretraining on millions of polymer samples. These findings demonstrate that LLM, such as Llama, can effectively capture chemical information encoded in polymer PSMILES, and underscore the efficacy of multimodal fusion of LLM embeddings and molecular structure embeddings in overcoming data scarcity and accelerating the discovery of advanced polymeric materials.", "date": "2025-03-29", "categories": []}
{"id": "2510.03355", "title": "High Cycle S-N curve prediction for Al 7075-T6 alloy using Recurrent Neural Networks (RNNs)", "abstract": "Aluminum is a widely used alloy, which is susceptible to fatigue failure. Characterizing fatigue performance for materials is extremely time and cost demanding, especially for high cycle data. To help mitigate this, a transfer learning based framework has been developed using Long short-term memory networks (LSTMs) in which a source LSTM model is trained based on pure axial fatigue data for Aluminum 7075-T6 alloy which is then transferred to predict high cycle torsional S-N curves. The framework was able to accurately predict Al torsional S-N curves for a much higher cycle range. It is the belief that this framework will help to drastically mitigate the cost of gathering fatigue characteristics for different materials and help prioritize tests with better cost and time constraints.", "date": "2025-10-02", "categories": []}
{"id": "2510.02142", "title": "Catalyst GFlowNet for electrocatalyst design: A hydrogen evolution reaction case study", "abstract": "Efficient and inexpensive energy storage is essential for accelerating the adoption of renewable energy and ensuring a stable supply, despite fluctuations in sources such as wind and solar. Electrocatalysts play a key role in hydrogen energy storage (HES), allowing the energy to be stored as hydrogen. However, the development of affordable and high-performance catalysts for this process remains a significant challenge. We introduce Catalyst GFlowNet, a generative model that leverages machine learning-based predictors of formation and adsorption energy to design crystal surfaces that act as efficient catalysts. We demonstrate the performance of the model through a proof-of-concept application to the hydrogen evolution reaction, a key reaction in HES, for which we successfully identified platinum as the most efficient known catalyst. In future work, we aim to extend this approach to the oxygen evolution reaction, where current optimal catalysts are expensive metal oxides, and open the search space to discover new materials. This generative modeling framework offers a promising pathway for accelerating the search for novel and efficient catalysts.", "date": "2025-10-02", "categories": []}
{"id": "2411.16483", "title": "Graph Transformer Networks for Accurate Band Structure Prediction: An End-to-End Approach", "abstract": "Predicting electronic band structures from crystal structures is crucial for understanding structure-property correlations in materials science. First-principles approaches are accurate but computationally intensive. Recent years, machine learning (ML) has been extensively applied to this field, while existing ML models predominantly focus on band gap predictions or indirect band structure estimation via solving predicted Hamiltonians. An end-to-end model to predict band structure accurately and efficiently is still lacking. Here, we introduce a graph Transformer-based end-to-end approach that directly predicts band structures from crystal structures with high accuracy. Our method leverages the continuity of the k-path and treat continuous bands as a sequence. We demonstrate that our model not only provides accurate band structure predictions but also can derive other properties (such as band gap, band center, and band dispersion) with high accuracy. We verify the model performance on large and diverse datasets.", "date": "2024-11-25", "categories": []}
{"id": "2502.14234", "title": "OBELiX: A Curated Dataset of Crystal Structures and Experimentally Measured Ionic Conductivities for Lithium Solid-State Electrolytes", "abstract": "Solid-state electrolyte batteries are expected to replace liquid electrolyte lithium-ion batteries in the near future thanks to their higher theoretical energy density and improved safety. However, their adoption is currently hindered by their lower effective ionic conductivity, a quantity that governs charge and discharge rates. Identifying highly ion-conductive materials using conventional theoretical calculations and experimental validation is both time-consuming and resource-intensive. While machine learning holds the promise to expedite this process, relevant ionic conductivity and structural data is scarce. Here, we present OBELiX, a database of $\\sim$600 synthesized solid electrolyte materials and their experimentally measured room temperature ionic conductivities gathered from literature and curated by domain experts. Each material is described by their measured composition, space group and lattice parameters. A full-crystal description in the form of a crystallographic information file (CIF) is provided for $\\sim$320 structures for which atomic positions were available. We discuss various statistics and features of the dataset and provide training and testing splits carefully designed to avoid data leakage. Finally, we benchmark seven existing ML models on the task of predicting ionic conductivity and discuss their performance. The goal of this work is to facilitate the use of machine learning for solid-state electrolyte materials discovery.", "date": "2025-02-20", "categories": []}
{"id": "2505.02361", "title": "Learning simple heuristic rules for classifying materials based on chemical composition", "abstract": "In the past decade, there has been a significant interest in the use of machine learning approaches in materials science research. Conventional deep learning approaches that rely on complex, nonlinear models have become increasingly important in computational materials science due to their high predictive accuracy. In contrast to these approaches, we have shown in a recent work that a remarkably simple learned heuristic rule -- based on the concept of topogivity -- can classify whether a material is topological using only its chemical composition. In this paper, we go beyond the topology classification scenario by also studying the use of machine learning to develop simple heuristic rules for classifying whether a material is a metal based on chemical composition. Moreover, we present a framework for incorporating chemistry-informed inductive bias based on the structure of the periodic table. For both the topology classification and the metallicity classification tasks, we empirically characterize the performance of simple heuristic rules fit with and without chemistry-informed inductive bias across a wide range of training set sizes. We find evidence that incorporating chemistry-informed inductive bias can reduce the amount of training data required to reach a given level of test accuracy.", "date": "2025-05-05", "categories": []}
{"id": "2510.00243", "title": "Reward driven discovery of the optimal microstructure representations with invariant variational autoencoders", "abstract": "Microscopy techniques generate vast amounts of complex image data that in principle can be used to discover simpler, interpretable, and parsimonious forms to reveal the underlying physical structures, such as elementary building blocks in molecular systems or order parameters and phases in crystalline materials. Variational Autoencoders (VAEs) provide a powerful means of constructing such low-dimensional representations, but their performance heavily depends on multiple non-myopic design choices, which are often optimized through trial-and-error and empirical analysis. To enable automated and unbiased optimization of VAE workflows, we investigated reward-based strategies for evaluating latent space representations. Using Piezoresponse Force Microscopy data as a model system, we examined multiple policies and reward functions that can serve as a foundation for automated optimization. Our analysis shows that approximating the latent space with Gaussian Mixture Models (GMM) and Bayesian Gaussian Mixture Models (BGMM) provides a strong basis for constructing reward functions capable of estimating model efficiency and guiding the search for optimal parsimonious representations.", "date": "2025-09-30", "categories": []}
{"id": "2311.09891", "title": "Classification-based detection and quantification of cross-domain data bias in materials discovery", "abstract": "It stands to reason that the amount and the quality of data is of key importance for setting up accurate AI-driven models. Among others, a fundamental aspect to consider is the bias introduced during sample selection in database generation. This is particularly relevant when a model is trained on a specialized dataset to predict a property of interest, and then applied to forecast the same property over samples having a completely different genesis. Indeed, the resulting biased model will likely produce unreliable predictions for many of those out-of-the-box samples. Neglecting such an aspect may hinder the AI-based discovery process, even when high quality, sufficiently large and highly reputable data sources are available. In this regard, with superconducting and thermoelectric materials as two prototypical case studies in the field of energy material discovery, we present and validate a new method (based on a classification strategy) capable of detecting, quantifying and circumventing the presence of cross-domain data bias.", "date": "2023-11-16", "categories": []}
{"id": "2411.10125", "title": "Energy-GNoME: A Living Database of Selected Materials for Energy Applications", "abstract": "Artificial Intelligence (AI) in materials science is driving significant advancements in the discovery of advanced materials for energy applications. The recent GNoME protocol identifies over 380,000 novel stable crystals. From this, we identify over 33,000 materials with potential as energy materials forming the Energy-GNoME database. Leveraging Machine Learning (ML) and Deep Learning (DL) tools, our protocol mitigates cross-domain data bias using feature spaces to identify potential candidates for thermoelectric materials, novel battery cathodes, and novel perovskites. Classifiers with both structural and compositional features identify domains of applicability, where we expect enhanced accuracy of the regressors. Such regressors are trained to predict key materials properties like, thermoelectric figure of merit (zT), band gap (Eg), and cathode voltage ($\\Delta V_c$). This method significantly narrows the pool of potential candidates, serving as an efficient guide for experimental and computational chemistry investigations and accelerating the discovery of materials suited for electricity generation, energy storage and conversion.", "date": "2024-11-15", "categories": []}
{"id": "2401.13192", "title": "Generative Design of Crystal Structures by Point Cloud Representations and Diffusion Model", "abstract": "Efficiently generating energetically stable crystal structures has long been a challenge in material design, primarily due to the immense arrangement of atoms in a crystal lattice. To facilitate the discovery of stable material, we present a framework for the generation of synthesizable materials, leveraging a point cloud representation to encode intricate structural information. At the heart of this framework lies the introduction of a diffusion model as its foundational pillar. To gauge the efficacy of our approach, we employ it to reconstruct input structures from our training datasets, rigorously validating its high reconstruction performance. Furthermore, we demonstrate the profound potential of Point Cloud-Based Crystal Diffusion (PCCD) by generating entirely new materials, emphasizing their synthesizability. Our research stands as a noteworthy contribution to the advancement of materials design and synthesis through the cutting-edge avenue of generative design instead of the conventional substitution or experience-based discovery.", "date": "2024-01-24", "categories": []}
{"id": "2412.18414", "title": "Discovery of 2D Materials via Symmetry-Constrained Diffusion Model", "abstract": "Generative model for 2D materials has shown significant promise in accelerating the material discovery process. The stability and performance of these materials are strongly influenced by their underlying symmetry. However, existing generative models for 2D materials often neglect symmetry constraints, which limits both the diversity and quality of the generated structures. Here, we introduce a symmetry-constrained diffusion model (SCDM) that integrates space group symmetry into the generative process. By incorporating Wyckoff positions, the model ensures adherence to symmetry principles, leading to the generation of 2,000 candidate structures. DFT calculations were conducted to evaluate the convex hull energies of these structures after structural relaxation. From the generated samples, 843 materials that met the energy stability criteria (Ehull < 0.6 eV/atom) were identified. Among these, six candidates were selected for further stability analysis, including phonon band structure evaluations and electronic properties investigations, all of which exhibited phonon spectrum stability. To benchmark the performance of SCDM, a symmetry-unconstrained diffusion model was also evaluated via crystal structure prediction model. The results highlight that incorporating symmetry constraints enhances the effectiveness of generated 2D materials, making a contribution to the discovery of 2D materials through generative modeling.", "date": "2024-12-24", "categories": []}
{"id": "2509.22667", "title": "A Comparison of Surrogate Constitutive Models for Viscoplastic Creep Simulation of HT-9 Steel", "abstract": "Mechanistic microstructure-informed constitutive models for the mechanical response of polycrystals are a cornerstone of computational materials science. However, as these models become increasingly more complex - often involving coupled differential equations describing the effect of specific deformation modes - their associated computational costs can become prohibitive, particularly in optimization or uncertainty quantification tasks that require numerous model evaluations. To address this challenge, surrogate constitutive models that balance accuracy and computational efficiency are highly desirable. Data-driven surrogate models, that learn the constitutive relation directly from data, have emerged as a promising solution. In this work, we develop two local surrogate models for the viscoplastic response of a steel: a piecewise response surface method and a mixture of experts model. These surrogates are designed to adapt to complex material behavior, which may vary with material parameters or operating conditions. The surrogate constitutive models are applied to creep simulations of HT-9 steel, an alloy of considerable interest to the nuclear energy sector due to its high tolerance to radiation damage, using training data generated from viscoplastic self-consistent (VPSC) simulations. We define a set of test metrics to numerically assess the accuracy of our surrogate models for predicting viscoplastic material behavior, and show that the mixture of experts model outperforms the piecewise response surface method in terms of accuracy.", "date": "2025-09-05", "categories": []}
{"id": "2509.23822", "title": "Space Group Conditional Flow Matching", "abstract": "Inorganic crystals are periodic, highly-symmetric arrangements of atoms in three-dimensional space. Their structures are constrained by the symmetry operations of a crystallographic \\emph{space group} and restricted to lie in specific affine subspaces known as \\emph{Wyckoff positions}. The frequency an atom appears in the crystal and its rough positioning are determined by its Wyckoff position. Most generative models that predict atomic coordinates overlook these symmetry constraints, leading to unrealistically high populations of proposed crystals exhibiting limited symmetry. We introduce Space Group Conditional Flow Matching, a novel generative framework that samples significantly closer to the target population of highly-symmetric, stable crystals. We achieve this by conditioning the entire generation process on a given space group and set of Wyckoff positions; specifically, we define a conditionally symmetric noise base distribution and a group-conditioned, equivariant, parametric vector field that restricts the motion of atoms to their initial Wyckoff position. Our form of group-conditioned equivariance is achieved using an efficient reformulation of \\emph{group averaging} tailored for symmetric crystals. Importantly, it reduces the computational overhead of symmetrization to a negligible level. We achieve state of the art results on crystal structure prediction and de novo generation benchmarks. We also perform relevant ablations.", "date": "2025-09-28", "categories": []}
{"id": "2509.25186", "title": "Guided Diffusion for the Discovery of New Superconductors", "abstract": "The inverse design of materials with specific desired properties, such as high-temperature superconductivity, represents a formidable challenge in materials science due to the vastness of chemical and structural space. We present a guided diffusion framework to accelerate the discovery of novel superconductors. A DiffCSP foundation model is pretrained on the Alexandria Database and fine-tuned on 7,183 superconductors with first principles derived labels. Employing classifier-free guidance, we sample 200,000 structures, which lead to 34,027 unique candidates. A multistage screening process that combines machine learning and density functional theory (DFT) calculations to assess stability and electronic properties, identifies 773 candidates with DFT-calculated $T_\\mathrm{c}>5$ K. Notably, our generative model demonstrates effective property-driven design. Our computational findings were validated against experimental synthesis and characterization performed as part of this work, which highlighted challenges in sparsely charted chemistries. This end-to-end workflow accelerates superconductor discovery while underscoring the challenge of predicting and synthesizing experimentally realizable materials.", "date": "2025-09-29", "categories": []}
{"id": "2403.15734", "title": "Space Group Informed Transformer for Crystalline Materials Generation", "abstract": "We introduce CrystalFormer, a transformer-based autoregressive model specifically designed for space group-controlled generation of crystalline materials. By explicitly incorporating space group symmetry, CrystalFormer greatly reduces the effective complexity of crystal space, which is essential for data-and compute-efficient generative modeling of crystalline materials. Leveraging the prominent discrete and sequential nature of the Wyckoff positions, CrystalFormer learns to generate crystals by directly predicting the species and coordinates of symmetry-inequivalent atoms in the unit cell. We demonstrate the advantages of CrystalFormer in standard tasks such as symmetric structure initialization and element substitution over widely used conventional approaches. Furthermore, we showcase its plug-and-play application to property-guided materials design, highlighting its flexibility. Our analysis reveals that CrystalFormer ingests sensible solid-state chemistry knowledge and heuristics by compressing the material dataset, thus enabling systematic exploration of crystalline materials space. The simplicity, generality, and adaptability of CrystalFormer position it as a promising architecture to be the foundational model of the entire crystalline materials space, heralding a new era in materials discovery and design.", "date": "2024-03-23", "categories": []}
{"id": "2509.21811", "title": "Scaling Laws for Neural Material Models", "abstract": "Predicting material properties is crucial for designing better batteries, semiconductors, and medical devices. Deep learning helps scientists quickly find promising materials by predicting their energy, forces, and stresses. Companies scale capacities of deep learning models in multiple domains, such as language modeling, and invest many millions of dollars into such models. Our team analyzes how scaling training data (giving models more information to learn from), model sizes (giving models more capacity to learn patterns), and compute (giving models more computational resources) for neural networks affects their performance for material property prediction. In particular, we trained both transformer and EquiformerV2 neural networks to predict material properties. We find empirical scaling laws for these models: we can predict how increasing each of the three hyperparameters (training data, model size, and compute) affects predictive performance. In particular, the loss $L$ can be measured with a power law relationship $L = \\alpha \\cdot N^{-\\beta}$, where $\\alpha$ and $\\beta$ are constants while $N$ is the relevant hyperparameter. We also incorporate command-line arguments for changing training settings such as the amount of epochs, maximum learning rate, and whether mixed precision is enabled. Future work could entail further investigating scaling laws for other neural network models in this domain, such as GemNet and fully connected networks, to assess how they compare to the models we trained.", "date": "2025-09-26", "categories": []}
{"id": "2509.22028", "title": "MCGM: Multi-stage Clustered Global Modeling for Long-range Interactions in Molecules", "abstract": "Geometric graph neural networks (GNNs) excel at capturing molecular geometry, yet their locality-biased message passing hampers the modeling of long-range interactions. Current solutions have fundamental limitations: extending cutoff radii causes computational costs to scale cubically with distance; physics-inspired kernels (e.g., Coulomb, dispersion) are often system-specific and lack generality; Fourier-space methods require careful tuning of multiple parameters (e.g., mesh size, k-space cutoff) with added computational overhead. We introduce Multi-stage Clustered Global Modeling (MCGM), a lightweight, plug-and-play module that endows geometric GNNs with hierarchical global context through efficient clustering operations. MCGM builds a multi-resolution hierarchy of atomic clusters, distills global information via dynamic hierarchical clustering, and propagates this context back through learned transformations, ultimately reinforcing atomic features via residual connections. Seamlessly integrated into four diverse backbone architectures, MCGM reduces OE62 energy prediction error by an average of 26.2%. On AQM, MCGM achieves state-of-the-art accuracy (17.0 meV for energy, 4.9 meV/{\\AA} for forces) while using 20% fewer parameters than Neural P3M. Code will be made available upon acceptance.", "date": "2025-09-26", "categories": []}
{"id": "2509.17094", "title": "DiffSyn: A Generative Diffusion Approach to Materials Synthesis Planning", "abstract": "The synthesis of crystalline materials, such as zeolites, remains a significant challenge due to a high-dimensional synthesis space, intricate structure-synthesis relationships and time-consuming experiments. Considering the one-to-many relationship between structure and synthesis, we propose DiffSyn, a generative diffusion model trained on over 23,000 synthesis recipes spanning 50 years of literature. DiffSyn generates probable synthesis routes conditioned on a desired zeolite structure and an organic template. DiffSyn achieves state-of-the-art performance by capturing the multi-modal nature of structure-synthesis relationships. We apply DiffSyn to differentiate among competing phases and generate optimal synthesis routes. As a proof of concept, we synthesize a UFI material using DiffSyn-generated synthesis routes. These routes, rationalized by density functional theory binding energies, resulted in the successful synthesis of a UFI material with a high Si/Al$_{\\text{ICP}}$ of 19.0, which is expected to improve thermal stability and is higher than that of any previously recorded.", "date": "2025-09-21", "categories": []}
{"id": "2509.20770", "title": "Extrapolating Phase-Field Simulations in Space and Time with Purely Convolutional Architectures", "abstract": "Phase-field models of liquid metal dealloying (LMD) can resolve rich microstructural dynamics but become intractable for large domains or long time horizons. We present a conditionally parameterized, fully convolutional U-Net surrogate that generalizes far beyond its training window in both space and time. The design integrates convolutional self-attention and physics-aware padding, while parameter conditioning enables variable time-step skipping and adaptation to diverse alloy systems. Although trained only on short, small-scale simulations, the surrogate exploits the translational invariance of convolutions to extend predictions to much longer horizons than traditional solvers. It accurately reproduces key LMD physics, with relative errors typically under 5% within the training regime and below 10% when extrapolating to larger domains and later times. The method accelerates computations by up to 16,000 times, cutting weeks of simulation down to seconds, and marks an early step toward scalable, high-fidelity extrapolation of LMD phase-field models.", "date": "2025-09-25", "categories": []}
{"id": "2502.04668", "title": "Machine-Learning Interatomic Potentials for Long-Range Systems", "abstract": "Machine-learning interatomic potentials have emerged as a revolutionary class of force-field models in molecular simulations, delivering quantum-mechanical accuracy at a fraction of the computational cost and enabling the simulation of large-scale systems over extended timescales. However, they often focus on modeling local environments, neglecting crucial long-range interactions. We propose a Sum-of-Gaussians Neural Network (SOG-Net), a lightweight and versatile framework for integrating long-range interactions into machine learning force field. The SOG-Net employs a latent-variable learning network that seamlessly bridges short-range and long-range components, coupled with an efficient Fourier convolution layer that incorporates long-range effects. By learning sum-of-Gaussians multipliers across different convolution layers, the SOG-Net adaptively captures diverse long-range decay behaviors while maintaining close-to-linear computational complexity during training and simulation via non-uniform fast Fourier transforms. The method is demonstrated effective for a broad range of long-range systems.", "date": "2025-02-07", "categories": []}
{"id": "2508.03162", "title": "The Open DAC 2025 Dataset for Sorbent Discovery in Direct Air Capture", "abstract": "Identifying useful sorbent materials for direct air capture (DAC) from humid air remains a challenge. We present the Open DAC 2025 (ODAC25) dataset, a significant expansion and improvement upon ODAC23 (Sriram et al., ACS Central Science, 10 (2024) 923), comprising nearly 60 million DFT single-point calculations for CO$_2$, H$_2$O, N$_2$, and O$_2$ adsorption in 15,000 MOFs. ODAC25 introduces chemical and configurational diversity through functionalized MOFs, high-energy GCMC-derived placements, and synthetically generated frameworks. ODAC25 also significantly improves upon the accuracy of DFT calculations and the treatment of flexible MOFs in ODAC23. Along with the dataset, we release new state-of-the-art machine-learned interatomic potentials trained on ODAC25 and evaluate them on adsorption energy and Henry's law coefficient predictions.", "date": "2025-08-05", "categories": []}
{"id": "2509.16266", "title": "Vibrational Fingerprints of Strained Polymers: A Spectroscopic Pathway to Mechanical State Prediction", "abstract": "The vibrational response of polymer networks under load provides a sensitive probe of molecular deformation and a route to non-destructive diagnostics. Here we show that machine-learned force fields reproduce these spectroscopic fingerprints with quantum-level fidelity in realistic epoxy thermosets. Using MACE-OFF23 molecular dynamics, we capture the experimentally observed redshifts of para-phenylene stretching modes under tensile load, in contrast to the harmonic OPLS-AA model. These shifts correlate with molecular elongation and alignment, consistent with Badger's rule, directly linking vibrational features to local stress. To capture IR intensities, we trained a symmetry-adapted dipole moment model on representative epoxy fragments, enabling validation of strain responses. Together, these approaches provide chemically accurate and computationally accessible predictions of strain-dependent vibrational spectra. Our results establish vibrational fingerprints as predictive markers of mechanical state in polymer networks, pointing to new strategies for stress mapping and structural-health diagnostics in advanced materials.", "date": "2025-09-18", "categories": []}
{"id": "2509.15908", "title": "Interpretable Nanoporous Materials Design with Symmetry-Aware Networks", "abstract": "Nanoporous materials hold promise for diverse sustainable applications, yet their vast chemical space poses challenges for efficient design. Machine learning offers a compelling pathway to accelerate the exploration, but existing models lack either interpretability or fidelity for elucidating the correlation between crystal geometry and property. Here, we report a three-dimensional periodic space sampling method that decomposes large nanoporous structures into local geometrical sites for combined property prediction and site-wise contribution quantification. Trained with a constructed database and retrieved datasets, our model achieves state-of-the-art accuracy and data efficiency for property prediction on gas storage, separation, and electrical conduction. Meanwhile, this approach enables the interpretation of the prediction and allows for accurate identification of significant local sites for targeted properties. Through identifying transferable high-performance sites across diverse nanoporous frameworks, our model paves the way for interpretable, symmetry-aware nanoporous materials design, which is extensible to other materials, like molecular crystals and beyond.", "date": "2025-09-19", "categories": []}
{"id": "2509.18573", "title": "Interaction Topological Transformer for Multiscale Learning in Porous Materials", "abstract": "Porous materials exhibit vast structural diversity and support critical applications in gas storage, separations, and catalysis. However, predictive modeling remains challenging due to the multiscale nature of structure-property relationships, where performance is governed by both local chemical environments and global pore-network topology. These complexities, combined with sparse and unevenly distributed labeled data, hinder generalization across material families. We propose the Interaction Topological Transformer (ITT), a unified data-efficient framework that leverages novel interaction topology to capture materials information across multiple scales and multiple levels, including structural, elemental, atomic, and pairwise-elemental organization. ITT extracts scale-aware features that reflect both compositional and relational structure within complex porous frameworks, and integrates them through a built-in Transformer architecture that supports joint reasoning across scales. Trained using a two-stage strategy, i.e., self-supervised pretraining on 0.6 million unlabeled structures followed by supervised fine-tuning, ITT achieves state-of-the-art, accurate, and transferable predictions for adsorption, transport, and stability properties. This framework provides a principled and scalable path for learning-guided discovery in structurally and chemically diverse porous materials.", "date": "2025-09-23", "categories": []}
{"id": "2509.14408", "title": "Deep Gaussian Process-based Cost-Aware Batch Bayesian Optimization for Complex Materials Design Campaigns", "abstract": "The accelerating pace and expanding scope of materials discovery demand optimization frameworks that efficiently navigate vast, nonlinear design spaces while judiciously allocating limited evaluation resources. We present a cost-aware, batch Bayesian optimization scheme powered by deep Gaussian process (DGP) surrogates and a heterotopic querying strategy. Our DGP surrogate, formed by stacking GP layers, models complex hierarchical relationships among high-dimensional compositional features and captures correlations across multiple target properties, propagating uncertainty through successive layers. We integrate evaluation cost into an upper-confidence-bound acquisition extension, which, together with heterotopic querying, proposes small batches of candidates in parallel, balancing exploration of under-characterized regions with exploitation of high-mean, low-variance predictions across correlated properties. Applied to refractory high-entropy alloys for high-temperature applications, our framework converges to optimal formulations in fewer iterations with cost-aware queries than conventional GP-based BO, highlighting the value of deep, uncertainty-aware, cost-sensitive strategies in materials campaigns.", "date": "2025-09-17", "categories": []}
{"id": "2502.17490", "title": "A generalized dual potential for inelastic Constitutive Artificial Neural Networks: A JAX implementation at finite strains", "abstract": "We present a methodology for designing a generalized dual potential, or pseudo potential, for inelastic Constitutive Artificial Neural Networks (iCANNs). This potential, expressed in terms of stress invariants, inherently satisfies thermodynamic consistency for large deformations. In comparison to our previous work, the new potential captures a broader spectrum of material behaviors, including pressure-sensitive inelasticity.\n  To this end, we revisit the underlying thermodynamic framework of iCANNs for finite strain inelasticity and derive conditions for constructing a convex, zero-valued, and non-negative dual potential. To embed these principles in a neural network, we detail the architecture's design, ensuring a priori compliance with thermodynamics.\n  To evaluate the proposed architecture, we study its performance and limitations discovering visco-elastic material behavior, though the method is not limited to visco-elasticity. In this context, we investigate different aspects in the strategy of discovering inelastic materials. Our results indicate that the novel architecture robustly discovers interpretable models and parameters, while autonomously revealing the degree of inelasticity.\n  The iCANN framework, implemented in JAX, is publicly accessible at https://doi.org/10.5281/zenodo.14894687.", "date": "2025-02-19", "categories": []}
{"id": "2508.12063", "title": "Generalized invariants meet constitutive neural networks: A novel framework for hyperelastic materials", "abstract": "The major challenge in determining a hyperelastic model for a given material is the choice of invariants and the selection how the strain energy function depends functionally on these invariants. Here we introduce a new data-driven framework that simultaneously discovers appropriate invariants and constitutive models for isotropic incompressible hyperelastic materials. Our approach identifies both the most suitable invariants in a class of generalized invariants and the corresponding strain energy function directly from experimental observations. Unlike previous methods that rely on fixed invariant choices or sequential fitting procedures, our method integrates the discovery process into a single neural network architecture. By looking at a continuous family of possible invariants, the model can flexibly adapt to different material behaviors. We demonstrate the effectiveness of this approach using popular benchmark datasets for rubber and brain tissue. For rubber, the method recovers a stretch-dominated formulation consistent with classical models. For brain tissue, it identifies a formulation sensitive to small stretches, capturing the nonlinear shear response characteristic of soft biological matter. Compared to traditional and neural-network-based models, our framework provides improved predictive accuracy and interpretability across a wide range of deformation states. This unified strategy offers a robust tool for automated and physically meaningful model discovery in hyperelasticity.", "date": "2025-08-16", "categories": []}
{"id": "2509.14600", "title": "TICA-Based Free Energy Matching for Machine-Learned Molecular Dynamics", "abstract": "Molecular dynamics (MD) simulations provide atomistic insight into biomolecular systems but are often limited by high computational costs required to access long timescales. Coarse-grained machine learning models offer a promising avenue for accelerating sampling, yet conventional force matching approaches often fail to capture the full thermodynamic landscape as fitting a model on the gradient may not fit the absolute differences between low-energy conformational states. In this work, we incorporate a complementary energy matching term into the loss function. We evaluate our framework on the Chignolin protein using the CGSchNet model, systematically varying the weight of the energy loss term. While energy matching did not yield statistically significant improvements in accuracy, it revealed distinct tendencies in how models generalize the free energy surface. Our results suggest future opportunities to enhance coarse-grained modeling through improved energy estimation techniques and multi-modal loss formulations.", "date": "2025-09-18", "categories": []}
{"id": "2509.15029", "title": "Physics-Informed GCN-LSTM Framework for Long-Term Forecasting of 2D and 3D Microstructure Evolution", "abstract": "This paper presents a physics-informed framework that integrates graph convolutional networks (GCN) with long short-term memory (LSTM) architecture to forecast microstructure evolution over long time horizons in both 2D and 3D with remarkable performance across varied metrics. The proposed framework is composition-aware, trained jointly on datasets with different compositions, and operates in latent graph space, which enables the model to capture compositions and morphological dynamics while remaining computationally efficient. Compressing and encoding phase-field simulation data with convolutional autoencoders and operating in Latent graph space facilitates efficient modeling of microstructural evolution across composition, dimensions, and long-term horizons. The framework captures the spatial and temporal patterns of evolving microstructures while enabling long-range forecasting at reduced computational cost after training.", "date": "2025-09-18", "categories": []}
{"id": "2509.12358", "title": "Unsupervised Atomic Data Mining via Multi-Kernel Graph Autoencoders for Machine Learning Force Fields", "abstract": "Constructing a chemically diverse dataset while avoiding sampling bias is critical to training efficient and generalizable force fields. However, in computational chemistry and materials science, many common dataset generation techniques are prone to oversampling regions of the potential energy surface. Furthermore, these regions can be difficult to identify and isolate from each other or may not align well with human intuition, making it challenging to systematically remove bias in the dataset. While traditional clustering and pruning (down-sampling) approaches can be useful for this, they can often lead to information loss or a failure to properly identify distinct regions of the potential energy surface due to difficulties associated with the high dimensionality of atomic descriptors. In this work, we introduce the Multi-kernel Edge Attention-based Graph Autoencoder (MEAGraph) model, an unsupervised approach for analyzing atomic datasets. MEAGraph combines multiple linear kernel transformations with attention-based message passing to capture geometric sensitivity and enable effective dataset pruning without relying on labels or extensive training. Demonstrated applications on niobium, tantalum, and iron datasets show that MEAGraph efficiently groups similar atomic environments, allowing for the use of basic pruning techniques for removing sampling bias. This approach provides an effective method for representation learning and clustering that can be used for data analysis, outlier detection, and dataset optimization.", "date": "2025-09-15", "categories": []}
{"id": "2509.10532", "title": "Crystal Systems Classification of Phosphate-Based Cathode Materials Using Machine Learning for Lithium-Ion Battery", "abstract": "The physical and chemical characteristics of cathodes used in batteries are derived from the lithium-ion phosphate cathodes crystalline arrangement, which is pivotal to the overall battery performance. Therefore, the correct prediction of the crystal system is essential to estimate the properties of cathodes. This study applies machine learning classification algorithms for predicting the crystal systems, namely monoclinic, orthorhombic, and triclinic, related to Li P (Mn, Fe, Co, Ni, V) O based Phosphate cathodes. The data used in this work is extracted from the Materials Project. Feature evaluation showed that cathode properties depend on the crystal structure, and optimized classification strategies lead to better predictability. Ensemble machine learning algorithms such as Random Forest, Extremely Randomized Trees, and Gradient Boosting Machines have demonstrated the best predictive capabilities for crystal systems in the Monte Carlo cross-validation test. Additionally, sequential forward selection (SFS) is performed to identify the most critical features influencing the prediction accuracy for different machine learning models, with Volume, Band gap, and Sites as input features ensemble machine learning algorithms such as Random Forest (80.69%), Extremely Randomized Tree (78.96%), and Gradient Boosting Machine (80.40%) approaches lead to the maximum accuracy towards crystallographic classification with stability and the predicted materials can be the potential cathode materials for lithium ion batteries.", "date": "2025-09-05", "categories": []}
{"id": "2505.08159", "title": "Self-Optimizing Machine Learning Potential Assisted Automated Workflow for Highly Efficient Complex Systems Material Design", "abstract": "Machine learning interatomic potentials have revolutionized complex materials design by enabling rapid exploration of material configurational spaces via crystal structure prediction with ab initio accuracy. However, critical challenges persist in ensuring robust generalization to unknown structures and minimizing the requirement for substantial expert knowledge and time-consuming manual interventions. Here, we propose an automated crystal structure prediction framework built upon the attention-coupled neural networks potential to address these limitations. The generalizability of the potential is achieved by sampling regions across the local minima of the potential energy surface, where the self-evolving pipeline autonomously refines the potential iteratively while minimizing human intervention. The workflow is validated on Mg-Ca-H ternary and Be-P-N-O quaternary systems by exploring nearly 10 million configurations, demonstrating substantial speedup compared to first-principles calculations. These results underscore the effectiveness of our approach in accelerating the exploration and discovery of complex multi-component functional materials.", "date": "2025-05-13", "categories": []}
{"id": "2509.10210", "title": "Towards Fully Automated Molecular Simulations: Multi-Agent Framework for Simulation Setup and Force Field Extraction", "abstract": "Automated characterization of porous materials has the potential to accelerate materials discovery, but it remains limited by the complexity of simulation setup and force field selection. We propose a multi-agent framework in which LLM-based agents can autonomously understand a characterization task, plan appropriate simulations, assemble relevant force fields, execute them and interpret their results to guide subsequent steps. As a first step toward this vision, we present a multi-agent system for literature-informed force field extraction and automated RASPA simulation setup. Initial evaluations demonstrate high correctness and reproducibility, highlighting this approach's potential to enable fully autonomous, scalable materials characterization.", "date": "2025-09-12", "categories": []}
{"id": "2509.07417", "title": "Benchmarking Universal Interatomic Potentials on Zeolite Structures", "abstract": "Interatomic potentials (IPs) with wide elemental coverage and high accuracy are powerful tools for high-throughput materials discovery. While the past few years witnessed the development of multiple new universal IPs that cover wide ranges of the periodic table, their applicability to target chemical systems should be carefully investigated. We benchmark several universal IPs using equilibrium zeolite structures as testbeds. We select a diverse set of universal IPs encompassing two major categories: (i) universal analytic IPs, including GFN-FF, UFF, and Dreiding; (ii) pretrained universal machine learning IPs (MLIPs), comprising CHGNet, ORB-v3, MatterSim, eSEN-30M-OAM, PFP-v7, and EquiformerV2-lE4-lF100-S2EFS-OC22. We compare them with established tailor-made IPs, SLC, ClayFF, and BSFF using experimental data and density functional theory (DFT) calculations with dispersion correction as the reference. The tested zeolite structures comprise pure silica frameworks and aluminosilicates containing copper species, potassium, and organic cations. We found that GFN-FF is the best among the tested universal analytic IPs, but it does not achieve satisfactory accuracy for highly strained silica rings and aluminosilicate systems. All MLIPs can well reproduce experimental or DFT-level geometries and energetics. Among the universal MLIPs, the eSEN-30M-OAM model shows the most consistent performance across all zeolite structures studied. These findings show that the modern pretrained universal MLIPs are practical tools in zeolite screening workflows involving various compositions.", "date": "2025-09-09", "categories": []}
{"id": "2509.08734", "title": "DEQuify your force field: More efficient simulations using deep equilibrium models", "abstract": "Machine learning force fields show great promise in enabling more accurate molecular dynamics simulations compared to manually derived ones. Much of the progress in recent years was driven by exploiting prior knowledge about physical systems, in particular symmetries under rotation, translation, and reflections. In this paper, we argue that there is another important piece of prior information that, thus fa,r hasn't been explored: Simulating a molecular system is necessarily continuous, and successive states are therefore extremely similar. Our contribution is to show that we can exploit this information by recasting a state-of-the-art equivariant base model as a deep equilibrium model. This allows us to recycle intermediate neural network features from previous time steps, enabling us to improve both accuracy and speed by $10\\%-20\\%$ on the MD17, MD22, and OC20 200k datasets, compared to the non-DEQ base model. The training is also much more memory efficient, allowing us to train more expressive models on larger systems.", "date": "2025-09-10", "categories": []}
{"id": "2503.17949", "title": "Equivariant Machine Learning Interatomic Potentials with Global Charge Redistribution", "abstract": "Machine learning interatomic potentials (MLIPs) provide a computationally efficient alternative to quantum mechanical simulations for predicting material properties. Message-passing graph neural networks, commonly used in these MLIPs, rely on local descriptor-based symmetry functions to model atomic interactions. However, such local descriptor-based approaches struggle with systems exhibiting long-range interactions, charge transfer, and compositional heterogeneity. In this work, we develop a new equivariant MLIP incorporating long-range Coulomb interactions through explicit treatment of electronic degrees of freedom, specifically global charge distribution within the system. This is achieved using a charge equilibration scheme based on predicted atomic electronegativities. We systematically evaluate our model across a range of benchmark periodic and non-periodic datasets, demonstrating that it outperforms both short-range equivariant and long-range invariant MLIPs in energy and force predictions. Our approach enables more accurate and efficient simulations of systems with long-range interactions and charge heterogeneity, expanding the applicability of MLIPs in computational materials science.", "date": "2025-03-23", "categories": []}
{"id": "2509.03547", "title": "Combining feature-based approaches with graph neural networks and symbolic regression for synergistic performance and interpretability", "abstract": "This study introduces MatterVial, an innovative hybrid framework for feature-based machine learning in materials science. MatterVial expands the feature space by integrating latent representations from a diverse suite of pretrained graph neural network (GNN) models including: structure-based (MEGNet), composition-based (ROOST), and equivariant (ORB) graph networks, with computationally efficient, GNN-approximated descriptors and novel features from symbolic regression. Our approach combines the chemical transparency of traditional feature-based models with the predictive power of deep learning architectures. When augmenting the feature-based model MODNet on Matbench tasks, this method yields significant error reductions and elevates its performance to be competitive with, and in several cases superior to, state-of-the-art end-to-end GNNs, with accuracy increases exceeding 40% for multiple tasks. An integrated interpretability module, employing surrogate models and symbolic regression, decodes the latent GNN-derived descriptors into explicit, physically meaningful formulas. This unified framework advances materials informatics by providing a high-performance, transparent tool that aligns with the principles of explainable AI, paving the way for more targeted and autonomous materials discovery.", "date": "2025-09-02", "categories": []}
{"id": "2506.10352", "title": "History-Aware Neural Operator: Robust Data-Driven Constitutive Modeling of Path-Dependent Materials", "abstract": "This study presents an end-to-end learning framework for data-driven modeling of path-dependent inelastic materials using neural operators. The framework is built on the premise that irreversible evolution of material responses, governed by hidden dynamics, can be inferred from observable data.\n  We develop the History-Aware Neural Operator (HANO), an autoregressive model that predicts path-dependent material responses from short segments of recent strain-stress history without relying on hidden state variables, thereby overcoming self-consistency issues commonly encountered in recurrent neural network (RNN)-based models. Built on a Fourier-based neural operator backbone, HANO enables discretization-invariant learning. To enhance its ability to capture both global loading patterns and critical local path dependencies, we embed a hierarchical self-attention mechanism that facilitates multiscale feature extraction.\n  Beyond ensuring self-consistency, HANO mitigates sensitivity to initial hidden states, a commonly overlooked issue that can lead to instability in recurrent models when applied to generalized loading paths. By modeling stress-strain evolution as a continuous operator rather than relying on fixed input-output mappings, HANO naturally accommodates varying path discretizations and exhibits robust performance under complex conditions, including irregular sampling, multi-cycle loading, noisy data, and pre-stressed states. We evaluate HANO on two benchmark problems: elastoplasticity with hardening and progressive anisotropic damage in brittle solids. Results show that HANO consistently outperforms baseline models in predictive accuracy, generalization, and robustness. With its demonstrated capabilities, HANO provides an effective data-driven surrogate for simulating inelastic materials and is well-suited for integration with classical numerical solvers.", "date": "2025-06-12", "categories": []}
{"id": "2507.02436", "title": "Toward a Robust and Generalizable Metamaterial Foundation Model", "abstract": "Advances in material functionalities drive innovations across various fields, where metamaterials-defined by structure rather than composition-are leading the way. Despite the rise of artificial intelligence (AI)-driven design strategies, their impact is limited by task-specific retraining, poor out-of-distribution(OOD) generalization, and the need for separate models for forward and inverse design. To address these limitations, we introduce the Metamaterial Foundation Model (MetaFO), a Bayesian transformer-based foundation model inspired by large language models. MetaFO learns the underlying mechanics of metamaterials, enabling probabilistic, zero-shot predictions across diverse, unseen combinations of material properties and structural responses. It also excels in nonlinear inverse design, even under OOD conditions. By treating metamaterials as an operator that maps material properties to structural responses, MetaFO uncovers intricate structure-property relationships and significantly expands the design space. This scalable and generalizable framework marks a paradigm shift in AI-driven metamaterial discovery, paving the way for next-generation innovations.", "date": "2025-07-03", "categories": []}
{"id": "2508.20143", "title": "CrystalICL: Enabling In-Context Learning for Crystal Generation", "abstract": "Designing crystal materials with desired physicochemical properties remains a fundamental challenge in materials science. While large language models (LLMs) have demonstrated strong in-context learning (ICL) capabilities, existing LLM-based crystal generation approaches are limited to zero-shot scenarios and are unable to benefit from few-shot scenarios. In contrast, human experts typically design new materials by modifying relevant known structures which aligns closely with the few-shot ICL paradigm. Motivated by this, we propose CrystalICL, a novel model designed for few-shot crystal generation. Specifically, we introduce a space-group based crystal tokenization method, which effectively reduces the complexity of modeling crystal symmetry in LLMs. We further introduce a condition-structure aware hybrid instruction tuning framework and a multi-task instruction tuning strategy, enabling the model to better exploit ICL by capturing structure-property relationships from limited data. Extensive experiments on four crystal generation benchmarks demonstrate the superiority of CrystalICL over the leading baseline methods on conditional and unconditional generation tasks.", "date": "2025-08-27", "categories": []}
{"id": "2508.20243", "title": "Linking heterogeneous microstructure informatics with expert characterization knowledge through customized and hybrid vision-language representations for industrial qualification", "abstract": "Rapid and reliable qualification of advanced materials remains a bottleneck in industrial manufacturing, particularly for heterogeneous structures produced via non-conventional additive manufacturing processes. This study introduces a novel framework that links microstructure informatics with a range of expert characterization knowledge using customized and hybrid vision-language representations (VLRs). By integrating deep semantic segmentation with pre-trained multi-modal models (CLIP and FLAVA), we encode both visual microstructural data and textual expert assessments into shared representations. To overcome limitations in general-purpose embeddings, we develop a customized similarity-based representation that incorporates both positive and negative references from expert-annotated images and their associated textual descriptions. This allows zero-shot classification of previously unseen microstructures through a net similarity scoring approach. Validation on an additively manufactured metal matrix composite dataset demonstrates the framework's ability to distinguish between acceptable and defective samples across a range of characterization criteria. Comparative analysis reveals that FLAVA model offers higher visual sensitivity, while the CLIP model provides consistent alignment with the textual criteria. Z-score normalization adjusts raw unimodal and cross-modal similarity scores based on their local dataset-driven distributions, enabling more effective alignment and classification in the hybrid vision-language framework. The proposed method enhances traceability and interpretability in qualification pipelines by enabling human-in-the-loop decision-making without task-specific model retraining. By advancing semantic interoperability between raw data and expert knowledge, this work contributes toward scalable and domain-adaptable qualification strategies in engineering informatics.", "date": "2025-08-27", "categories": []}
{"id": "2508.18806", "title": "Temperature-Aware Recurrent Neural Operator for Temperature-Dependent Anisotropic Plasticity in HCP Materials", "abstract": "Neural network surrogate models for constitutive laws in computational mechanics have been in use for some time. In plasticity, these models often rely on gated recurrent units (GRUs) or long short-term memory (LSTM) cells, which excel at capturing path-dependent phenomena. However, they suffer from long training times and time-resolution-dependent predictions that extrapolate poorly. Moreover, most existing surrogates for macro- or mesoscopic plasticity handle only relatively simple material behavior. To overcome these limitations, we introduce the Temperature-Aware Recurrent Neural Operator (TRNO), a time-resolution-independent neural architecture. We apply the TRNO to model the temperature-dependent plastic response of polycrystalline magnesium, which shows strong plastic anisotropy and thermal sensitivity. The TRNO achieves high predictive accuracy and generalizes effectively across diverse loading cases, temperatures, and time resolutions. It also outperforms conventional GRU and LSTM models in training efficiency and predictive performance. Finally, we demonstrate multiscale simulations with the TRNO, yielding a speedup of at least three orders of magnitude over traditional constitutive models.", "date": "2025-08-26", "categories": []}
{"id": "2505.14745", "title": "Explainable Prediction of the Mechanical Properties of Composites with CNNs", "abstract": "Composites are amongst the most important materials manufactured today, as evidenced by their use in countless applications. In order to establish the suitability of composites in specific applications, finite element (FE) modelling, a numerical method based on partial differential equations, is the industry standard for assessing their mechanical properties. However, FE modelling is exceptionally costly from a computational viewpoint, a limitation which has led to efforts towards applying AI models to this task. However, in these approaches: the chosen model architectures were rudimentary, feed-forward neural networks giving limited accuracy; the studies focused on predicting elastic mechanical properties, without considering material strength limits; and the models lacked transparency, hindering trustworthiness by users. In this paper, we show that convolutional neural networks (CNNs) equipped with methods from explainable AI (XAI) can be successfully deployed to solve this problem. Our approach uses customised CNNs trained on a dataset we generate using transverse tension tests in FE modelling to predict composites' mechanical properties, i.e., Young's modulus and yield strength. We show empirically that our approach achieves high accuracy, outperforming a baseline, ResNet-34, in estimating the mechanical properties. We then use SHAP and Integrated Gradients, two post-hoc XAI methods, to explain the predictions, showing that the CNNs use the critical geometrical features that influence the composites' behaviour, thus allowing engineers to verify that the models are trustworthy by representing the science of composites.", "date": "2025-05-20", "categories": []}
{"id": "2506.12051", "title": "GUST: Quantifying Free-Form Geometric Uncertainty of Metamaterials Using Small Data", "abstract": "This paper introduces GUST (Generative Uncertainty learning via Self-supervised pretraining and Transfer learning), a framework for quantifying free-form geometric uncertainties inherent in the manufacturing of metamaterials. GUST leverages the representational power of deep generative models to learn a high-dimensional conditional distribution of as-fabricated unit cell geometries given nominal designs, thereby enabling uncertainty quantification. To address the scarcity of real-world manufacturing data, GUST employs a two-stage learning process. First, it leverages self-supervised pretraining on a large-scale synthetic dataset to capture the structure variability inherent in metamaterial geometries and an approximated distribution of as-fabricated geometries given nominal designs. Subsequently, GUST employs transfer learning by fine-tuning the pretrained model on limited real-world manufacturing data, allowing it to adapt to specific manufacturing processes and nominal designs. With only 960 unit cells additively manufactured in only two passes, GUST can capture the variability in geometry and effective material properties. In contrast, directly training a generative model on the same amount of real-world data proves insufficient, as demonstrated through both qualitative and quantitative comparisons. This scalable and cost-effective approach significantly reduces data requirements while maintaining the effectiveness in learning complex, real-world geometric uncertainties, offering an affordable method for free-form geometric uncertainty quantification in the manufacturing of metamaterials. The capabilities of GUST hold significant promise for high-precision industries such as aerospace and biomedical engineering, where understanding and mitigating manufacturing uncertainties are critical.", "date": "2025-05-28", "categories": []}
{"id": "2508.00959", "title": "Enhancing material behavior discovery using embedding-oriented Physically-Guided Neural Networks with Internal Variables", "abstract": "Physically Guided Neural Networks with Internal Variables are SciML tools that use only observable data for training and and have the capacity to unravel internal state relations. They incorporate physical knowledge both by prescribing the model architecture and using loss regularization, thus endowing certain specific neurons with a physical meaning as internal state variables. Despite their potential, these models face challenges in scalability when applied to high-dimensional data such as fine-grid spatial fields or time-evolving systems. In this work, we propose some enhancements to the PGNNIV framework that address these scalability limitations through reduced-order modeling techniques. Specifically, we introduce alternatives to the original decoder structure using spectral decomposition, POD, and pretrained autoencoder-based mappings. These surrogate decoders offer varying trade-offs between computational efficiency, accuracy, noise tolerance, and generalization, while improving drastically the scalability. Additionally, we integrate model reuse via transfer learning and fine-tuning strategies to exploit previously acquired knowledge, supporting efficient adaptation to novel materials or configurations, and significantly reducing training time while maintaining or improving model performance. To illustrate these various techniques, we use a representative case governed by the nonlinear diffusion equation, using only observable data. Results demonstrate that the enhanced PGNNIV framework successfully identifies the underlying constitutive state equations while maintaining high predictive accuracy. It also improves robustness to noise, mitigates overfitting, and reduces computational demands. The proposed techniques can be tailored to various scenarios depending on data availability, resources, and specific modeling objectives, overcoming scalability challenges in all the scenarios.", "date": "2025-08-01", "categories": []}
{"id": "2508.16857", "title": "Neural Contrast Expansion for Explainable Structure-Property Prediction and Random Microstructure Design", "abstract": "Effective properties of composite materials are defined as the ensemble average of property-specific PDE solutions over the underlying microstructure distributions. Traditionally, predicting such properties can be done by solving PDEs derived from microstructure samples or building data-driven models that directly map microstructure samples to properties. The former has a higher running cost, but provides explainable sensitivity information that may guide material design; the latter could be more cost-effective if the data overhead is amortized, but its learned sensitivities are often less explainable. With a focus on properties governed by linear self-adjoint PDEs (e.g., Laplace, Helmholtz, and Maxwell curl-curl) defined on bi-phase microstructures, we propose a structure-property model that is both cost-effective and explainable. Our method is built on top of the strong contrast expansion (SCE) formalism, which analytically maps $N$-point correlations of an unbounded random field to its effective properties. Since real-world material samples have finite sizes and analytical PDE kernels are not always available, we propose Neural Contrast Expansion (NCE), an SCE-inspired architecture to learn surrogate PDE kernels from structure-property data. For static conduction and electromagnetic wave propagation cases, we show that NCE models reveal accurate and insightful sensitivity information useful for material design. Compared with other PDE kernel learning methods, our method does not require measurements about the PDE solution fields, but rather only requires macroscopic property measurements that are more accessible in material development contexts.", "date": "2025-08-23", "categories": []}
{"id": "2508.17568", "title": "MetaGen: A DSL, Database, and Benchmark for VLM-Assisted Metamaterial Generation", "abstract": "Metamaterials are micro-architected structures whose geometry imparts highly tunable-often counter-intuitive-bulk properties. Yet their design is difficult because of geometric complexity and a non-trivial mapping from architecture to behaviour. We address these challenges with three complementary contributions. (i) MetaDSL: a compact, semantically rich domain-specific language that captures diverse metamaterial designs in a form that is both human-readable and machine-parsable. (ii) MetaDB: a curated repository of more than 150,000 parameterized MetaDSL programs together with their derivatives-three-dimensional geometry, multi-view renderings, and simulated elastic properties. (iii) MetaBench: benchmark suites that test three core capabilities of vision-language metamaterial assistants-structure reconstruction, property-driven inverse design, and performance prediction. We establish baselines by fine-tuning state-of-the-art vision-language models and deploy an omni-model within an interactive, CAD-like interface. Case studies show that our framework provides a strong first step toward integrated design and understanding of structure-representation-property relationships.", "date": "2025-08-25", "categories": []}
{"id": "2508.16239", "title": "UniEM-3M: A Universal Electron Micrograph Dataset for Microstructural Segmentation and Generation", "abstract": "Quantitative microstructural characterization is fundamental to materials science, where electron micrograph (EM) provides indispensable high-resolution insights. However, progress in deep learning-based EM characterization has been hampered by the scarcity of large-scale, diverse, and expert-annotated datasets, due to acquisition costs, privacy concerns, and annotation complexity. To address this issue, we introduce UniEM-3M, the first large-scale and multimodal EM dataset for instance-level understanding. It comprises 5,091 high-resolution EMs, about 3 million instance segmentation labels, and image-level attribute-disentangled textual descriptions, a subset of which will be made publicly available. Furthermore, we are also releasing a text-to-image diffusion model trained on the entire collection to serve as both a powerful data augmentation tool and a proxy for the complete data distribution. To establish a rigorous benchmark, we evaluate various representative instance segmentation methods on the complete UniEM-3M and present UniEM-Net as a strong baseline model. Quantitative experiments demonstrate that this flow-based model outperforms other advanced methods on this challenging benchmark. Our multifaceted release of a partial dataset, a generative model, and a comprehensive benchmark -- available at huggingface -- will significantly accelerate progress in automated materials analysis.", "date": "2025-08-22", "categories": []}
{"id": "2508.14127", "title": "Comparison of derivative-free and gradient-based minimization for multi-objective compositional design of shape memory alloys", "abstract": "Designing shape memory alloys (SMAs) that meet performance targets while remaining affordable and sustainable is a complex challenge. In this work, we focus on optimizing SMA compositions to achieve a desired martensitic start temperature (Ms) while minimizing cost. To do this, we use machine learning models as surrogate predictors and apply numerical optimization methods to search for suitable alloy combinations. We trained two types of machine learning models, a tree-based ensemble and a neural network, using a dataset of experimentally characterized alloys and physics-informed features. The tree-based model was used with a derivative-free optimizer (COBYLA), while the neural network, which provides gradient information, was paired with a gradient-based optimizer (TRUST-CONSTR). Our results show that while both models predict Ms with similar accuracy, the optimizer paired with the neural network finds better solutions more consistently. COBYLA often converged to suboptimal results, especially when the starting guess was far from the target. The TRUST-CONSTR method showed more stable behavior and was better at reaching alloy compositions that met both objectives. This study demonstrates a practical approach to exploring new SMA compositions by combining physics-informed data, machine learning models, and optimization algorithms. Although the scale of our dataset is smaller than simulation-based efforts, the use of experimental data improves the reliability of the predictions. The approach can be extended to other materials where design trade-offs must be made with limited data.", "date": "2025-08-19", "categories": []}
{"id": "2505.24065", "title": "Cross-Modal Characterization of Thin Film MoS$_2$ Using Generative Models", "abstract": "The growth and characterization of materials using empirical optimization typically requires a significant amount of expert time, experience, and resources. Several complementary characterization methods are routinely performed to determine the quality and properties of a grown sample. Machine learning (ML) can support the conventional approaches by using historical data to guide and provide speed and efficiency to the growth and characterization of materials. Specifically, ML can provide quantitative information from characterization data that is typically obtained from a different modality. In this study, we have investigated the feasibility of projecting the quantitative metric from microscopy measurements, such as atomic force microscopy (AFM), using data obtained from spectroscopy measurements, like Raman spectroscopy. Generative models were also trained to generate the full and specific features of the Raman and photoluminescence spectra from each other and the AFM images of the thin film MoS$_2$. The results are promising and have provided a foundational guide for the use of ML for the cross-modal characterization of materials for their accelerated, efficient, and cost-effective discovery.", "date": "2025-05-29", "categories": []}
{"id": "2508.13559", "title": "Physics-Informed Neural Networks for Programmable Origami Metamaterials with Controlled Deployment", "abstract": "Origami-inspired structures provide unprecedented opportunities for creating lightweight, deployable systems with programmable mechanical responses. However, their design remains challenging due to complex nonlinear mechanics, multistability, and the need for precise control of deployment forces. Here, we present a physics-informed neural network (PINN) framework for both forward prediction and inverse design of conical Kresling origami (CKO) without requiring pre-collected training data. By embedding mechanical equilibrium equations directly into the learning process, the model predicts complete energy landscapes with high accuracy while minimizing non-physical artifacts. The inverse design routine specifies both target stable-state heights and separating energy barriers, enabling freeform programming of the entire energy curve. This capability is extended to hierarchical CKO assemblies, where sequential layer-by-layer deployment is achieved through programmed barrier magnitudes. Finite element simulations and experiments on physical prototypes validate the designed deployment sequences and barrier ratios, confirming the robustness of the approach. This work establishes a versatile, data-free route for programming complex mechanical energy landscapes in origami-inspired metamaterials, offering broad potential for deployable aerospace systems, morphing structures, and soft robotic actuators.", "date": "2025-08-19", "categories": []}
{"id": "2508.11685", "title": "Enhancing Corrosion Resistance of Aluminum Alloys Through AI and ML Modeling", "abstract": "Corrosion poses a significant challenge to the performance of aluminum alloys, particularly in marine environments. This study investigates the application of machine learning (ML) algorithms to predict and optimize corrosion resistance, utilizing a comprehensive open-source dataset compiled from various sources. The dataset encompasses corrosion rate data and environmental conditions, preprocessed to standardize units and formats. We explored two different approaches, a direct approach, where the material's composition and environmental conditions were used as inputs to predict corrosion rates; and an inverse approach, where corrosion rate served as the input to identify suitable material compositions as output. We employed and compared three distinct ML methodologies for forward predictions: Random Forest regression, optimized via grid search; a feed-forward neural network, utilizing ReLU activation and Adam optimization; and Gaussian Process Regression (GPR), implemented with GPyTorch and employing various kernel functions. The Random Forest and neural network models provided predictive capabilities based on elemental compositions and environmental conditions. Notably, Gaussian Process Regression demonstrated superior performance, particularly with hybrid kernel functions. Log-transformed GPR further refined predictions. This study highlights the efficacy of ML, particularly GPR, in predicting corrosion rates and material properties.", "date": "2025-08-10", "categories": []}
{"id": "2508.10718", "title": "Symmetry-Constrained Multi-Scale Physics-Informed Neural Networks for Graphene Electronic Band Structure Prediction", "abstract": "Accurate prediction of electronic band structures in two-dimensional materials remains a fundamental challenge, with existing methods struggling to balance computational efficiency and physical accuracy. We present the Symmetry-Constrained Multi-Scale Physics-Informed Neural Network (SCMS-PINN) v35, which directly learns graphene band structures while rigorously enforcing crystallographic symmetries through a multi-head architecture. Our approach introduces three specialized ResNet-6 pathways -- K-head for Dirac physics, M-head for saddle points, and General head for smooth interpolation -- operating on 31 physics-informed features extracted from k-points. Progressive Dirac constraint scheduling systematically increases the weight parameter from 5.0 to 25.0, enabling hierarchical learning from global topology to local critical physics. Training on 10,000 k-points over 300 epochs achieves 99.99\\% reduction in training loss (34.597 to 0.003) with validation loss of 0.0085. The model predicts Dirac point gaps within 30.3 $\\mu$eV of theoretical zero and achieves average errors of 53.9 meV (valence) and 40.5 meV (conduction) across the Brillouin zone. All twelve C$_{6v}$ operations are enforced through systematic averaging, guaranteeing exact symmetry preservation. This framework establishes a foundation for extending physics-informed learning to broader two-dimensional materials for accelerated discovery.", "date": "2025-08-14", "categories": []}
{"id": "2508.06591", "title": "Generative Artificial Intelligence Extracts Structure-Function Relationships from Plants for New Materials", "abstract": "Large language models (LLMs) have reshaped the research landscape by enabling new approaches to knowledge retrieval and creative ideation. Yet their application in discipline-specific experimental science, particularly in highly multi-disciplinary domains like materials science, remains limited. We present a first-of-its-kind framework that integrates generative AI with literature from hitherto-unconnected fields such as plant science, biomimetics, and materials engineering to extract insights and design experiments for materials. We focus on humidity-responsive systems such as pollen-based materials and Rhapis excelsa (broadleaf lady palm) leaves, which exhibit self-actuation and adaptive performance. Using a suite of AI tools, including a fine-tuned model (BioinspiredLLM), Retrieval-Augmented Generation (RAG), agentic systems, and a Hierarchical Sampling strategy, we extract structure-property relationships and translate them into new classes of bioinspired materials. Structured inference protocols generate and evaluate hundreds of hypotheses from a single query, surfacing novel and experimentally tractable ideas. We validate our approach through real-world implementation: LLM-generated procedures, materials designs, and mechanical predictions were tested in the laboratory, culminating in the fabrication of a novel pollen-based adhesive with tunable morphology and measured shear strength, establishing a foundation for future plant-derived adhesive design. This work demonstrates how AI-assisted ideation can drive real-world materials design and enable effective human-AI collaboration.", "date": "2025-08-08", "categories": []}
{"id": "2508.06642", "title": "Benchmarking Self-Driving Labs", "abstract": "A key goal of modern materials science is accelerating the pace of materials discovery. Self-driving labs, or systems that select experiments using machine learning and then execute them using automation, are designed to fulfil this promise by performing experiments faster, more intelligently, more reliably, and with richer metadata than conventional means. This review summarizes progress in understanding the degree to which SDLs accelerate learning by quantifying how much they reduce the number of experiments required for a given goal. The review begins by summarizing the theory underlying two key metrics, namely acceleration factor AF and enhancement factor EF, which quantify how much faster and better an algorithm is relative to a reference strategy. Next, we provide a comprehensive review of the literature, which reveals a wide range of AFs with a median of 6, and that tends to increase with the dimensionality of the space, reflecting an interesting blessing of dimensionality. In contrast, reported EF values vary by over two orders of magnitude, although they consistently peak at 10-20 experiments per dimension. To understand these results, we perform a series of simulated Bayesian optimization campaigns that reveal how EF depends upon the statistical properties of the parameter space while AF depends on its complexity. Collectively, these results reinforce the motivation for using SDLs by revealing their value across a wide range of material parameter spaces and provide a common language for quantifying and understanding this acceleration.", "date": "2025-08-08", "categories": []}
{"id": "2508.06691", "title": "Role of Large Language Models and Retrieval-Augmented Generation for Accelerating Crystalline Material Discovery: A Systematic Review", "abstract": "Large language models (LLMs) have emerged as powerful tools for knowledge-intensive tasks across domains. In materials science, to find novel materials for various energy efficient devices for various real-world applications, requires several time and cost expensive simulations and experiments. In order to tune down the uncharted material search space, minimizing the experimental cost, LLMs can play a bigger role to first provide an accelerated search of promising known material candidates. Furthermore, the integration of LLMs with domain-specific information via retrieval-augmented generation (RAG) is poised to revolutionize how researchers predict materials structures, analyze defects, discover novel compounds, and extract knowledge from literature and databases. In motivation to the potentials of LLMs and RAG in accelerating material discovery, this paper presents a broad and systematic review to examine the recent advancements in applying LLMs and RAG to key materials science problems. We survey state-of-the-art developments in crystal structure prediction, defect analysis, materials discovery, literature mining, database integration, and multi-modal retrieval, highlighting how combining LLMs with external knowledge sources enables new capabilities. We discuss the performance, limitations, and implications of these approaches, and outline future directions for leveraging LLMs to accelerate materials research and discovery for advancement in technologies in the area of electronics, optics, biomedical, and energy storage.", "date": "2025-08-08", "categories": []}
{"id": "2502.06836", "title": "CAST: Cross Attention based multimodal fusion of Structure and Text for materials property prediction", "abstract": "Recent advancements in graph neural networks (GNNs) have significantly enhanced the prediction of material properties by modeling crystal structures as graphs. However, GNNs often struggle to capture global structural characteristics, such as crystal systems, limiting their predictive performance. To overcome this issue, we propose CAST, a cross-attention-based multimodal model that integrates graph representations with textual descriptions of materials, effectively preserving critical structural and compositional information. Unlike previous approaches, such as CrysMMNet and MultiMat, which rely on aggregated material-level embeddings, CAST leverages cross-attention mechanisms to combine fine-grained graph node-level and text token-level features. Additionally, we introduce a masked node prediction pretraining strategy that further enhances the alignment between node and text embeddings. Our experimental results demonstrate that CAST outperforms existing baseline models across four key material properties-formation energy, band gap, bulk modulus, and shear modulus-with average relative MAE improvements ranging from 10.2% to 35.7%. Analysis of attention maps confirms the importance of pretraining in effectively aligning multimodal representations. This study underscores the potential of multimodal learning frameworks for developing more accurate and globally informed predictive models in materials science.", "date": "2025-02-06", "categories": []}
{"id": "2508.04841", "title": "Data Driven Insights into Composition Property Relationships in FCC High Entropy Alloys", "abstract": "Structural High Entropy Alloys (HEAs) are crucial in advancing technology across various sectors, including aerospace, automotive, and defense industries. However, the scarcity of integrated chemistry, process, structure, and property data presents significant challenges for predictive property modeling. Given the vast design space of these alloys, uncovering the underlying patterns is essential yet difficult, requiring advanced methods capable of learning from limited and heterogeneous datasets. This work presents several sensitivity analyses, highlighting key elemental contributions to mechanical behavior, including insights into the compositional factors associated with brittle and fractured responses observed during nanoindentation testing in the BIRDSHOT center NiCoFeCrVMnCuAl system dataset. Several encoder decoder based chemistry property models, carefully tuned through Bayesian multi objective hyperparameter optimization, are evaluated for mapping alloy composition to six mechanical properties. The models achieve competitive or superior performance to conventional regressors across all properties, particularly for yield strength and the UTS/YS ratio, demonstrating their effectiveness in capturing complex composition property relationships.", "date": "2025-08-06", "categories": []}
{"id": "2508.02956", "title": "Autonomous Inorganic Materials Discovery via Multi-Agent Physics-Aware Scientific Reasoning", "abstract": "Conventional machine learning approaches accelerate inorganic materials design via accurate property prediction and targeted material generation, yet they operate as single-shot models limited by the latent knowledge baked into their training data. A central challenge lies in creating an intelligent system capable of autonomously executing the full inorganic materials discovery cycle, from ideation and planning to experimentation and iterative refinement. We introduce SparksMatter, a multi-agent AI model for automated inorganic materials design that addresses user queries by generating ideas, designing and executing experimental workflows, continuously evaluating and refining results, and ultimately proposing candidate materials that meet the target objectives. SparksMatter also critiques and improves its own responses, identifies research gaps and limitations, and suggests rigorous follow-up validation steps, including DFT calculations and experimental synthesis and characterization, embedded in a well-structured final report. The model's performance is evaluated across case studies in thermoelectrics, semiconductors, and perovskite oxides materials design. The results demonstrate the capacity of SparksMatter to generate novel stable inorganic structures that target the user's needs. Benchmarking against frontier models reveals that SparksMatter consistently achieves higher scores in relevance, novelty, and scientific rigor, with a significant improvement in novelty across multiple real-world design tasks as assessed by a blinded evaluator. These results demonstrate SparksMatter's unique capacity to generate chemically valid, physically meaningful, and creative inorganic materials hypotheses beyond existing materials knowledge.", "date": "2025-08-04", "categories": []}
{"id": "2508.03278", "title": "Artificial Intelligence and Generative Models for Materials Discovery -- A Review", "abstract": "High throughput experimentation tools, machine learning (ML) methods, and open material databases are radically changing the way new materials are discovered. From the experimentally driven approach in the past, we are moving quickly towards the artificial intelligence (AI) driven approach, realizing the 'inverse design' capabilities that allow the discovery of new materials given the desired properties. This review aims to discuss different principles of AI-driven generative models that are applicable for materials discovery, including different materials representations available for this purpose. We will also highlight specific applications of generative models in designing new catalysts, semiconductors, polymers, or crystals while addressing challenges such as data scarcity, computational cost, interpretability, synthesizability, and dataset biases. Emerging approaches to overcome limitations and integrate AI with experimental workflows will be discussed, including multimodal models, physics informed architectures, and closed-loop discovery systems. This review aims to provide insights for researchers aiming to harness AI's transformative potential in accelerating materials discovery for sustainability, healthcare, and energy innovation.", "date": "2025-08-05", "categories": []}
{"id": "2508.03371", "title": "A neural network machine-learning approach for characterising hydrogen trapping parameters from TDS experiments", "abstract": "The hydrogen trapping behaviour of metallic alloys is generally characterised using Thermal Desorption Spectroscopy (TDS). However, as an indirect method, extracting key parameters (trap binding energies and densities) remains a significant challenge. To address these limitations, this work introduces a machine learning-based scheme for parameter identification from TDS spectra. A multi-Neural Network (NN) model is developed and trained exclusively on synthetic data to predict trapping parameters directly from experimental data. The model comprises two multi-layer, fully connected, feed-forward NNs trained with backpropagation. The first network (classification model) predicts the number of distinct trap types. The second network (regression model) then predicts the corresponding trap densities and binding energies. The NN architectures, hyperparameters, and data pre-processing were optimised to minimise the amount of training data. The proposed model demonstrated strong predictive capabilities when applied to three tempered martensitic steels of different compositions. The code developed is freely provided.", "date": "2025-08-05", "categories": []}
{"id": "2508.03405", "title": "Model Accuracy and Data Heterogeneity Shape Uncertainty Quantification in Machine Learning Interatomic Potentials", "abstract": "Machine learning interatomic potentials (MLIPs) enable accurate atomistic modelling, but reliable uncertainty quantification (UQ) remains elusive. In this study, we investigate two UQ strategies, ensemble learning and D-optimality, within the atomic cluster expansion framework. It is revealed that higher model accuracy strengthens the correlation between predicted uncertainties and actual errors and improves novelty detection, with D-optimality yielding more conservative estimates. Both methods deliver well calibrated uncertainties on homogeneous training sets, yet they underpredict errors and exhibit reduced novelty sensitivity on heterogeneous datasets. To address this limitation, we introduce clustering-enhanced local D-optimality, which partitions configuration space into clusters during training and applies D-optimality within each cluster. This approach substantially improves the detection of novel atomic environments in heterogeneous datasets. Our findings clarify the roles of model fidelity and data heterogeneity in UQ performance and provide a practical route to robust active learning and adaptive sampling strategies for MLIP development.", "date": "2025-08-05", "categories": []}
{"id": "2508.00833", "title": "Deep Kernel Bayesian Optimisation for Closed-Loop Electrode Microstructure Design with User-Defined Properties based on GANs", "abstract": "The generation of multiphase porous electrode microstructures with optimum morphological and transport properties is essential in the design of improved electrochemical energy storage devices, such as lithium-ion batteries. Electrode characteristics directly influence battery performance by acting as the main sites where the electrochemical reactions coupled with transport processes occur. This work presents a generation-optimisation closed-loop algorithm for the design of microstructures with tailored properties. A deep convolutional Generative Adversarial Network is used as a deep kernel and employed to generate synthetic three-phase three-dimensional images of a porous lithium-ion battery cathode material. A Gaussian Process Regression uses the latent space of the generator and serves as a surrogate model to correlate the morphological and transport properties of the synthetic microstructures. This surrogate model is integrated into a deep kernel Bayesian optimisation framework, which optimises cathode properties as a function of the latent space of the generator. A set of objective functions were defined to perform the maximisation of morphological properties (e.g., volume fraction, specific surface area) and transport properties (relative diffusivity). We demonstrate the ability to perform simultaneous maximisation of correlated properties (specific surface area and relative diffusivity), as well as constrained optimisation of these properties. This is the maximisation of morphological or transport properties constrained by constant values of the volume fraction of the phase of interest. Visualising the optimised latent space reveals its correlation with morphological properties, enabling the fast generation of visually realistic microstructures with customised properties.", "date": "2025-06-07", "categories": []}
{"id": "2508.00896", "title": "Phase-fraction guided denoising diffusion model for augmenting multiphase steel microstructure segmentation via micrograph image-mask pair synthesis", "abstract": "The effectiveness of machine learning in metallographic microstructure segmentation is often constrained by the lack of human-annotated phase masks, particularly for rare or compositionally complex morphologies within the metal alloy. We introduce PF-DiffSeg, a phase-fraction controlled, one-stage denoising diffusion framework that jointly synthesizes microstructure images and their corresponding segmentation masks in a single generative trajectory to further improve segmentation accuracy. By conditioning on global phase-fraction vectors, augmented to represent real data distribution and emphasize minority classes, our model generates compositionally valid and structurally coherent microstructure image and mask samples that improve both data diversity and training efficiency. Evaluated on the MetalDAM benchmark for additively manufactured multiphase steel, our synthetic augmentation method yields notable improvements in segmentation accuracy compared to standard augmentation strategies especially in minority classes and further outperforms a two-stage mask-guided diffusion and generative adversarial network (GAN) baselines, while also reducing inference time compared to conventional approach. The method integrates generation and conditioning into a unified framework, offering a scalable solution for data augmentation in metallographic applications.", "date": "2025-07-28", "categories": []}
{"id": "2508.02641", "title": "FastCSP: Accelerated Molecular Crystal Structure Prediction with Universal Model for Atoms", "abstract": "Crystal Structure Prediction (CSP) of molecular crystals plays a central role in applications, such as pharmaceuticals and organic electronics. CSP is challenging and computationally expensive due to the need to explore a large search space with sufficient accuracy to capture energy differences of a few kJ/mol between polymorphs. Dispersion-inclusive density functional theory (DFT) provides the required accuracy but its computational cost is impractical for a large number of putative structures. We introduce FastCSP, an open-source, high-throughput CSP workflow based on machine learning interatomic potentials (MLIPs). FastCSP combines random structure generation using Genarris 3.0 with geometry relaxation and free energy calculations powered entirely by the Universal Model for Atoms (UMA) MLIP. We benchmark FastCSP on a curated set of 28 mostly rigid molecules, demonstrating that our workflow consistently generates known experimental structures and ranks them within 5 kJ/mol per molecule of the global minimum. Our results demonstrate that universal MLIPs can be used across diverse compounds without requiring system-specific tuning. Moreover, the speed and accuracy afforded by UMA eliminate the need for classical force fields in the early stages of CSP and for final re-ranking with DFT. The open-source release of the entire FastCSP workflow significantly lowers the barrier to accessing CSP. CSP results for a single system can be obtained within hours on tens of modern GPUs, making high-throughput crystal structure prediction feasible for a broad range of scientific applications.", "date": "2025-08-04", "categories": []}
{"id": "2507.23160", "title": "Extended Factorization Machine Annealing for Rapid Discovery of Transparent Conducting Materials", "abstract": "The development of novel transparent conducting materials (TCMs) is essential for enhancing the performance and reducing the cost of next-generation devices such as solar cells and displays. In this research, we focus on the (Al$_x$Ga$_y$In$_z$)$_2$O$_3$ system and extend the FMA framework, which combines a Factorization Machine (FM) and annealing, to search for optimal compositions and crystal structures with high accuracy and low cost. The proposed method introduces (i) the binarization of continuous variables, (ii) the utilization of good solutions using a Hopfield network, (iii) the activation of global search through adaptive random flips, and (iv) fine-tuning via a bit-string local search. Validation using the (Al$_x$Ga$_y$In$_z$)$_2$O$_3$ data from the Kaggle \"Nomad2018 Predicting Transparent Conductors\" competition demonstrated that our method achieves faster and more accurate searches than Bayesian optimization and genetic algorithms. Furthermore, its application to multi-objective optimization showed its capability in designing materials by simultaneously considering both the band gap and formation energy. These results suggest that applying our method to larger, more complex search problems and diverse material designs that reflect realistic experimental conditions is expected to contribute to the further advancement of materials informatics.", "date": "2025-07-30", "categories": []}
{"id": "2410.19568", "title": "Prediction of microstructural representativity from a single image", "abstract": "In this study, we present a method for predicting the representativity of the phase fraction observed in a single image (2D or 3D) of a material. Traditional approaches often require large datasets and extensive statistical analysis to estimate the Integral Range, a key factor in determining the variance of microstructural properties. Our method leverages the Two-Point Correlation function to directly estimate the variance from a single image, thereby enabling phase fraction prediction with associated confidence levels. We validate our approach using open-source datasets, demonstrating its efficacy across diverse microstructures. This technique significantly reduces the data requirements for representativity analysis, providing a practical tool for material scientists and engineers working with limited microstructural data. To make the method easily accessible, we have created a web-application, www.imagerep.io, for quick, simple and informative use of the method.", "date": "2024-10-25", "categories": []}
{"id": "2507.19799", "title": "Enhancing Materials Discovery with Valence Constrained Design in Generative Modeling", "abstract": "Diffusion-based deep generative models have emerged as powerful tools for inverse materials design. Yet, many existing approaches overlook essential chemical constraints such as oxidation state balance, which can lead to chemically invalid structures. Here we introduce CrysVCD (Crystal generator with Valence-Constrained Design), a modular framework that integrates chemical rules directly into the generative process. CrysVCD first employs a transformer-based elemental language model to generate valence-balanced compositions, followed by a diffusion model to generate crystal structures. The valence constraint enables orders-of-magnitude more efficient chemical valence checking, compared to pure data-driven approaches with post-screening. When fine-tuned on stability metrics, CrysVCD achieves 85% thermodynamic stability and 68% phonon stability. Moreover, CrysVCD supports conditional generation of functional materials, enabling discovery of candidates such as high thermal conductivity semiconductors and high-$\\kappa$ dielectric compounds. Designed as a general-purpose plugin, CrysVCD can be integrated into diverse generative pipeline to promote chemical validity, offering a reliable, scientifically grounded path for materials discovery.", "date": "2025-07-26", "categories": []}
{"id": "2507.20326", "title": "MIPS: a Multimodal Infinite Polymer Sequence Pre-training Framework for Polymer Property Prediction", "abstract": "Polymers, composed of repeating structural units called monomers, are fundamental materials in daily life and industry. Accurate property prediction for polymers is essential for their design, development, and application. However, existing modeling approaches, which typically represent polymers by the constituent monomers, struggle to capture the whole properties of polymer, since the properties change during the polymerization process. In this study, we propose a Multimodal Infinite Polymer Sequence (MIPS) pre-training framework, which represents polymers as infinite sequences of monomers and integrates both topological and spatial information for comprehensive modeling. From the topological perspective, we generalize message passing mechanism (MPM) and graph attention mechanism (GAM) to infinite polymer sequences. For MPM, we demonstrate that applying MPM to infinite polymer sequences is equivalent to applying MPM on the induced star-linking graph of monomers. For GAM, we propose to further replace global graph attention with localized graph attention (LGA). Moreover, we show the robustness of the \"star linking\" strategy through Repeat and Shift Invariance Test (RSIT). Despite its robustness, \"star linking\" strategy exhibits limitations when monomer side chains contain ring structures, a common characteristic of polymers, as it fails the Weisfeiler-Lehman~(WL) test. To overcome this issue, we propose backbone embedding to enhance the capability of MPM and LGA on infinite polymer sequences. From the spatial perspective, we extract 3D descriptors of repeating monomers to capture spatial information. Finally, we design a cross-modal fusion mechanism to unify the topological and spatial information. Experimental validation across eight diverse polymer property prediction tasks reveals that MIPS achieves state-of-the-art performance.", "date": "2025-07-27", "categories": []}
{"id": "2208.04146", "title": "Linking Properties to Microstructure in Liquid Metal Embedded Elastomers via Machine Learning", "abstract": "Liquid metals (LM) are embedded in an elastomer matrix to obtain soft composites with unique thermal, dielectric, and mechanical properties. They have applications in soft robotics, biomedical engineering, and wearable electronics. By linking the structure to the properties of these materials, it is possible to perform material design rationally. Liquid-metal embedded elastomers (LMEEs) have been designed for targeted electro-thermo-mechanical properties by semi-supervised learning of structure-property (SP) links in a variational autoencoder network (VAE). The design parameters are the microstructural descriptors that are physically meaningful and have affine relationships with the synthetization of the studied particulate composite. The machine learning (ML) model is trained on a generated dataset of microstructural descriptors with their multifunctional property quantities as their labels. Sobol sequence is used for in-silico Design of Experiment (DoE) by sampling the design space to generate a comprehensive dataset of 3D microstructure realizations via a packing algorithm. The mechanical responses of the generated microstructures are simulated using a previously developed Finite Element (FE) model, considering the surface tension induced by LM inclusions, while the linear thermal and dielectric constants are homogenized with the help of our in-house Fast Fourier Transform (FFT) package. Following the training by minimization of an appropriate loss function, the VAE encoder acts as the surrogate of numerical solvers of the multifunctional homogenizations, and its decoder is used for the material design. Our results indicate the satisfactory performance of the surrogate model and the inverse calculator with respect to high-fidelity numerical simulations validated with LMEE experimental results.", "date": "2022-07-24", "categories": []}
{"id": "2402.04379", "title": "Fine-Tuned Language Models Generate Stable Inorganic Materials as Text", "abstract": "We propose fine-tuning large language models for generation of stable materials. While unorthodox, fine-tuning large language models on text-encoded atomistic data is simple to implement yet reliable, with around 90% of sampled structures obeying physical constraints on atom positions and charges. Using energy above hull calculations from both learned ML potentials and gold-standard DFT calculations, we show that our strongest model (fine-tuned LLaMA-2 70B) can generate materials predicted to be metastable at about twice the rate (49% vs 28%) of CDVAE, a competing diffusion model. Because of text prompting's inherent flexibility, our models can simultaneously be used for unconditional generation of stable material, infilling of partial structures and text-conditional generation. Finally, we show that language models' ability to capture key symmetries of crystal structures improves with model scale, suggesting that the biases of pretrained LLMs are surprisingly well-suited for atomistic data.", "date": "2024-02-06", "categories": []}
{"id": "2412.20796", "title": "FastCHGNet: Training one Universal Interatomic Potential to 1.5 Hours with 32 GPUs", "abstract": "Graph neural network universal interatomic potentials (GNN-UIPs) have demonstrated remarkable generalization and transfer capabilities in material discovery and property prediction. These models can accelerate molecular dynamics (MD) simulation by several orders of magnitude while maintaining \\textit{ab initio} accuracy, making them a promising new paradigm in material simulations. One notable example is Crystal Hamiltonian Graph Neural Network (CHGNet), pretrained on the energies, forces, stresses, and magnetic moments from the MPtrj dataset, representing a state-of-the-art GNN-UIP model for charge-informed MD simulations. However, training the CHGNet model is time-consuming(8.3 days on one A100 GPU) for three reasons: (i) requiring multi-layer propagation to reach more distant atom information, (ii) requiring second-order derivatives calculation to finish weights updating and (iii) the implementation of reference CHGNet does not fully leverage the computational capabilities. This paper introduces FastCHGNet, an optimized CHGNet, with three contributions: Firstly, we design innovative Force/Stress Readout modules to decompose Force/Stress prediction. Secondly, we adopt massive optimizations such as kernel fusion, redundancy bypass, etc, to exploit GPU computation power sufficiently. Finally, we extend CHGNet to support multiple GPUs and propose a load-balancing technique to enhance GPU utilization. Numerical results show that FastCHGNet reduces memory footprint by a factor of 3.59. The final training time of FastCHGNet can be decreased to \\textbf{1.53 hours} on 32 GPUs without sacrificing model accuracy.", "date": "2024-12-30", "categories": []}
{"id": "2507.17907", "title": "Deep learning-aided inverse design of porous metamaterials", "abstract": "The ultimate aim of the study is to explore the inverse design of porous metamaterials using a deep learning-based generative framework. Specifically, we develop a property-variational autoencoder (pVAE), a variational autoencoder (VAE) augmented with a regressor, to generate structured metamaterials with tailored hydraulic properties, such as porosity and permeability. While this work uses the lattice Boltzmann method (LBM) to generate intrinsic permeability tensor data for limited porous microstructures, a convolutional neural network (CNN) is trained using a bottom-up approach to predict effective hydraulic properties. This significantly reduces the computational cost compared to direct LBM simulations. The pVAE framework is trained on two datasets: a synthetic dataset of artificial porous microstructures and CT-scan images of volume elements from real open-cell foams. The encoder-decoder architecture of the VAE captures key microstructural features, mapping them into a compact and interpretable latent space for efficient structure-property exploration. The study provides a detailed analysis and interpretation of the latent space, demonstrating its role in structure-property mapping, interpolation, and inverse design. This approach facilitates the generation of new metamaterials with desired properties. The datasets and codes used in this study will be made open-access to support further research.", "date": "2025-07-23", "categories": []}
{"id": "2507.16218", "title": "Toward Routine CSP of Pharmaceuticals: A Fully Automated Protocol Using Neural Network Potentials", "abstract": "Crystal structure prediction (CSP) is a useful tool in pharmaceutical development for identifying and assessing risks associated with polymorphism, yet widespread adoption has been hindered by high computational costs and the need for both manual specification and expert knowledge to achieve useful results. Here, we introduce a fully automated, high-throughput CSP protocol designed to overcome these barriers. The protocol's efficiency is driven by Lavo-NN, a novel neural network potential (NNP) architected and trained specifically for pharmaceutical crystal structure generation and ranking. This NNP-driven crystal generation phase is integrated into a scalable cloud-based workflow. We validate this CSP protocol on an extensive retrospective benchmark of 49 unique molecules, almost all of which are drug-like, successfully generating structures that match all 110 $Z' = 1$ experimental polymorphs. The average CSP in this benchmark is performed with approximately 8.4k CPU hours, which is a significant reduction compared to other protocols. The practical utility of the protocol is further demonstrated through case studies that resolve ambiguities in experimental data and a semi-blinded challenge that successfully identifies and ranks polymorphs of three modern drugs from powder X-ray diffraction patterns alone. By significantly reducing the required time and cost, the protocol enables CSP to be routinely deployed earlier in the drug discovery pipeline, such as during lead optimization. Rapid turnaround times and high throughput also enable CSP that can be run in parallel with experimental screening, providing chemists with real-time insights to guide their work in the lab.", "date": "2025-07-22", "categories": []}
{"id": "2506.00302", "title": "Beyond Atomic Geometry Representations in Materials Science: A Human-in-the-Loop Multimodal Framework", "abstract": "Most materials science datasets are limited to atomic geometries (e.g., XYZ files), restricting their utility for multimodal learning and comprehensive data-centric analysis. These constraints have historically impeded the adoption of advanced machine learning techniques in the field. This work introduces MultiCrystalSpectrumSet (MCS-Set), a curated framework that expands materials datasets by integrating atomic structures with 2D projections and structured textual annotations, including lattice parameters and coordination metrics. MCS-Set enables two key tasks: (1) multimodal property and summary prediction, and (2) constrained crystal generation with partial cluster supervision. Leveraging a human-in-the-loop pipeline, MCS-Set combines domain expertise with standardized descriptors for high-quality annotation. Evaluations using state-of-the-art language and vision-language models reveal substantial modality-specific performance gaps and highlight the importance of annotation quality for generalization. MCS-Set offers a foundation for benchmarking multimodal models, advancing annotation practices, and promoting accessible, versatile materials science datasets. The dataset and implementations are available at https://github.com/KurbanIntelligenceLab/MultiCrystalSpectrumSet.", "date": "2025-05-30", "categories": []}
{"id": "2507.14267", "title": "DREAMS: Density Functional Theory Based Research Engine for Agentic Materials Simulation", "abstract": "Materials discovery relies on high-throughput, high-fidelity simulation techniques such as Density Functional Theory (DFT), which require years of training, extensive parameter fine-tuning and systematic error handling. To address these challenges, we introduce the DFT-based Research Engine for Agentic Materials Screening (DREAMS), a hierarchical, multi-agent framework for DFT simulation that combines a central Large Language Model (LLM) planner agent with domain-specific LLM agents for atomistic structure generation, systematic DFT convergence testing, High-Performance Computing (HPC) scheduling, and error handling. In addition, a shared canvas helps the LLM agents to structure their discussions, preserve context and prevent hallucination. We validate DREAMS capabilities on the Sol27LC lattice-constant benchmark, achieving average errors below 1\\% compared to the results of human DFT experts. Furthermore, we apply DREAMS to the long-standing CO/Pt(111) adsorption puzzle, demonstrating its long-term and complex problem-solving capabilities. The framework again reproduces expert-level literature adsorption-energy differences. Finally, DREAMS is employed to quantify functional-driven uncertainties with Bayesian ensemble sampling, confirming the Face Centered Cubic (FCC)-site preference at the Generalized Gradient Approximation (GGA) DFT level. In conclusion, DREAMS approaches L3-level automation - autonomous exploration of a defined design space - and significantly reduces the reliance on human expertise and intervention, offering a scalable path toward democratized, high-throughput, high-fidelity computational materials discovery.", "date": "2025-07-18", "categories": []}
{"id": "2507.15303", "title": "Universal crystal material property prediction via multi-view geometric fusion in graph transformers", "abstract": "Accurately and comprehensively representing crystal structures is critical for advancing machine learning in large-scale crystal materials simulations, however, effectively capturing and leveraging the intricate geometric and topological characteristics of crystal structures remains a core, long-standing challenge for most existing methods in crystal property prediction. Here, we propose MGT, a multi-view graph transformer framework that synergistically fuses SE3 invariant and SO3 equivariant graph representations, which respectively captures rotation-translation invariance and rotation equivariance in crystal geometries. To strategically incorporate these complementary geometric representations, we employ a lightweight mixture of experts router in MGT to adaptively adjust the weight assigned to SE3 and SO3 embeddings based on the specific target task. Compared with previous state-of-the-art models, MGT reduces the mean absolute error by up to 21% on crystal property prediction tasks through multi-task self-supervised pretraining. Ablation experiments and interpretable investigations confirm the effectiveness of each technique implemented in our framework. Additionally, in transfer learning scenarios including crystal catalyst adsorption energy and hybrid perovskite bandgap prediction, MGT achieves performance improvements of up to 58% over existing baselines, demonstrating domain-agnostic scalability across diverse application domains. As evidenced by the above series of studies, we believe that MGT can serve as useful model for crystal material property prediction, providing a valuable tool for the discovery of novel materials.", "date": "2025-07-21", "categories": []}
{"id": "2501.02707", "title": "Refining Coarse-Grained Molecular Topologies: A Bayesian Optimization Approach", "abstract": "Molecular Dynamics (MD) simulations are essential for accurately predicting the physical and chemical properties of large molecular systems across various pressure and temperature ensembles. However, the high computational costs associated with All-Atom (AA) MD simulations have led to the development of Coarse-Grained Molecular Dynamics (CGMD), providing a lower-dimensional compression of the AA structure into representative CG beads, offering reduced computational expense at the cost of predictive accuracy. Existing CGMD methods, such as CG-Martini (calibrated against experimental data), aim to generate an embedding of a topology that sufficiently generalizes across a range of structures. Detrimentally, in attempting to specify parameterization with applicability across molecular classes, it is unable to specialize to domain-specific applications, where sufficient accuracy and computational speed are critical. This work presents a novel approach to optimize derived results from CGMD simulations by refining the general-purpose Martini3 topologies specifically the bonded interaction parameters within a given coarse-grained mapping - for domain-specific applications using Bayesian Optimization methodologies. We have developed and validated a CG potential applicable to any degree of polymerization, representing a significant advancement in the field. Our optimized CG potential, based on the Martini3 framework, aims to achieve accuracy comparable to AAMD while maintaining the computational efficiency of CGMD. This approach bridges the gap between efficiency and accuracy in multiscale molecular simulations, potentially enabling more rapid and cost-effective molecular discovery across various scientific and technological domains.", "date": "2025-01-06", "categories": []}
{"id": "2507.13246", "title": "The carbon cost of materials discovery: Can machine learning really accelerate the discovery of new photovoltaics?", "abstract": "Computational screening has become a powerful complement to experimental efforts in the discovery of high-performance photovoltaic (PV) materials. Most workflows rely on density functional theory (DFT) to estimate electronic and optical properties relevant to solar energy conversion. Although more efficient than laboratory-based methods, DFT calculations still entail substantial computational and environmental costs. Machine learning (ML) models have recently gained attention as surrogates for DFT, offering drastic reductions in resource use with competitive predictive performance. In this study, we reproduce a canonical DFT-based workflow to estimate the maximum efficiency limit and progressively replace its components with ML surrogates. By quantifying the CO$_2$ emissions associated with each computational strategy, we evaluate the trade-offs between predictive efficacy and environmental cost. Our results reveal multiple hybrid ML/DFT strategies that optimize different points along the accuracy--emissions front. We find that direct prediction of scalar quantities, such as maximum efficiency, is significantly more tractable than using predicted absorption spectra as an intermediate step. Interestingly, ML models trained on DFT data can outperform DFT workflows using alternative exchange--correlation functionals in screening applications, highlighting the consistency and utility of data-driven approaches. We also assess strategies to improve ML-driven screening through expanded datasets and improved model architectures tailored to PV-relevant features. This work provides a quantitative framework for building low-emission, high-throughput discovery pipelines.", "date": "2025-07-17", "categories": []}
{"id": "2507.12404", "title": "Neural Network-Guided Symbolic Regression for Interpretable Descriptor Discovery in Perovskite Catalysts", "abstract": "Understanding and predicting the activity of oxide perovskite catalysts for the oxygen evolution reaction (OER) requires descriptors that are both accurate and physically interpretable. While symbolic regression (SR) offers a path to discover such formulas, its performance degrades with high-dimensional inputs and small datasets. We present a two-phase framework that combines neural networks (NN), feature importance analysis, and symbolic regression (SR) to discover interpretable descriptors for OER activity in oxide perovskites. In Phase I, using a small dataset and seven structural features, we reproduce and improve the known {\\mu}/t descriptor by engineering composite features and applying symbolic regression, achieving training and validation MAEs of 22.8 and 20.8 meV, respectively. In Phase II, we expand to 164 features, reduce dimensionality, and identify LUMO energy as a key electronic descriptor. A final formula using {\\mu}/t, {\\mu}/RA, and LUMO energy achieves improved accuracy (training and validation MAEs of 22.1 and 20.6 meV) with strong physical interpretability. Our results demonstrate that NN-guided symbolic regression enables accurate, interpretable, and physically meaningful descriptor discovery in data-scarce regimes, indicating interpretability need not sacrifice accuracy for materials informatics.", "date": "2025-07-16", "categories": []}
{"id": "2506.06935", "title": "An Agentic Framework for Autonomous Metamaterial Modeling and Inverse Design", "abstract": "Recent significant advances in integrating multiple Large Language Model (LLM) systems have enabled Agentic Frameworks capable of performing complex tasks autonomously, including novel scientific research. We develop and demonstrate such a framework specifically for the inverse design of photonic metamaterials. When queried with a desired optical spectrum, the Agent autonomously proposes and develops a forward deep learning model, accesses external tools via APIs for tasks like simulation and optimization, utilizes memory, and generates a final design via a deep inverse method. The framework's effectiveness is demonstrated in its ability to automate, reason, plan, and adapt. Notably, the Agentic Framework possesses internal reflection and decision flexibility, permitting highly varied and potentially novel outputs.", "date": "2025-06-07", "categories": []}
{"id": "2507.05559", "title": "MP-ALOE: An r2SCAN dataset for universal machine learning interatomic potentials", "abstract": "We present MP-ALOE, a dataset of nearly 1 million DFT calculations using the accurate r2SCAN meta-generalized gradient approximation. Covering 89 elements, MP-ALOE was created using active learning and primarily consists of off-equilibrium structures. We benchmark a machine learning interatomic potential trained on MP-ALOE, and evaluate its performance on a series of benchmarks, including predicting the thermochemical properties of equilibrium structures; predicting forces of far-from-equilibrium structures; maintaining physical soundness under static extreme deformations; and molecular dynamic stability under extreme temperatures and pressures. MP-ALOE shows strong performance on all of these benchmarks, and is made public for the broader community to utilize.", "date": "2025-07-08", "categories": []}
{"id": "2410.10885", "title": "Adaptive AI-Driven Material Synthesis: Towards Autonomous 2D Materials Growth", "abstract": "Two-dimensional (2D) materials are poised to revolutionize current solid-state technology with their extraordinary properties. Yet, the primary challenge remains their scalable production. While there have been significant advancements, much of the scientific progress has depended on the exfoliation of materials, a method that poses severe challenges for large-scale applications. With the advent of artificial intelligence (AI) in materials science, innovative synthesis methodologies are now on the horizon. This study explores the forefront of autonomous materials synthesis using an artificial neural network (ANN) trained by evolutionary methods, focusing on the efficient production of graphene. Our approach demonstrates that a neural network can iteratively and autonomously learn a time-dependent protocol for the efficient growth of graphene, without requiring pretraining on what constitutes an effective recipe. Evaluation criteria are based on the proximity of the Raman signature to that of monolayer graphene: higher scores are granted to outcomes whose spectrum more closely resembles that of an ideal continuous monolayer structure. This feedback mechanism allows for iterative refinement of the ANN's time-dependent synthesis protocols, progressively improving sample quality. Through the advancement and application of AI methodologies, this work makes a substantial contribution to the field of materials engineering, fostering a new era of innovation and efficiency in the synthesis process.", "date": "2024-10-10", "categories": []}
{"id": "2411.15351", "title": "Accelerating CALPHAD-based Phase Diagram Predictions in Complex Alloys Using Universal Machine Learning Potentials: Opportunities and Challenges", "abstract": "Accurate phase diagram prediction is crucial for understanding alloy thermodynamics and advancing materials design. While traditional CALPHAD methods are robust, they are resource-intensive and limited by experimentally assessed data. This work explores the use of machine learning interatomic potentials (MLIPs) such as M3GNet, CHGNet, MACE, SevenNet, and ORB to significantly accelerate phase diagram calculations by using the Alloy Theoretic Automated Toolkit (ATAT) to map calculations of the energies and free energies of atomistic systems to CALPHAD-compatible thermodynamic descriptions. Using case studies including Cr-Mo, Cu-Au, and Pt-W, we demonstrate that MLIPs, particularly ORB, achieve computational speedups exceeding three orders of magnitude compared to DFT while maintaining phase stability predictions within acceptable accuracy. Extending this approach to liquid phases and ternary systems like Cr-Mo-V highlights its versatility for high-entropy alloys and complex chemical spaces. This work demonstrates that MLIPs, integrated with tools like ATAT within a CALPHAD framework, provide an efficient and accurate framework for high-throughput thermodynamic modeling, enabling rapid exploration of novel alloy systems. While many challenges remain to be addressed, the accuracy of some of these MLIPs (ORB in particular) are on the verge of paving the way toward high-throughput generation of CALPHAD thermodynamic descriptions of multi-component, multi-phase alloy systems.", "date": "2024-11-22", "categories": []}
{"id": "2507.03602", "title": "Kinetic Langevin Diffusion for Crystalline Materials Generation", "abstract": "Generative modeling of crystalline materials using diffusion models presents a series of challenges: the data distribution is characterized by inherent symmetries and involves multiple modalities, with some defined on specific manifolds. Notably, the treatment of fractional coordinates representing atomic positions in the unit cell requires careful consideration, as they lie on a hypertorus. In this work, we introduce Kinetic Langevin Diffusion for Materials (KLDM), a novel diffusion model for crystalline materials generation, where the key innovation resides in the modeling of the coordinates. Instead of resorting to Riemannian diffusion on the hypertorus directly, we generalize Trivialized Diffusion Model (TDM) to account for the symmetries inherent to crystals. By coupling coordinates with auxiliary Euclidean variables representing velocities, the diffusion process is now offset to a flat space. This allows us to effectively perform diffusion on the hypertorus while providing a training objective that accounts for the periodic translation symmetry of the true data distribution. We evaluate KLDM on both Crystal Structure Prediction (CSP) and De-novo Generation (DNG) tasks, demonstrating its competitive performance with current state-of-the-art models.", "date": "2025-07-04", "categories": []}
{"id": "2507.04053", "title": "TopoMAS: Large Language Model Driven Topological Materials Multiagent System", "abstract": "Topological materials occupy a frontier in condensed-matter physics thanks to their remarkable electronic and quantum properties, yet their cross-scale design remains bottlenecked by inefficient discovery workflows. Here, we introduce TopoMAS (Topological materials Multi-Agent System), an interactive human-AI framework that seamlessly orchestrates the entire materials-discovery pipeline: from user-defined queries and multi-source data retrieval, through theoretical inference and crystal-structure generation, to first-principles validation. Crucially, TopoMAS closes the loop by autonomously integrating computational outcomes into a dynamic knowledge graph, enabling continuous knowledge refinement. In collaboration with human experts, it has already guided the identification of novel topological phases SrSbO3, confirmed by first-principles calculations. Comprehensive benchmarks demonstrate robust adaptability across base Large Language Model, with the lightweight Qwen2.5-72B model achieving 94.55% accuracy while consuming only 74.3-78.4% of tokens required by Qwen3-235B and 83.0% of DeepSeek-V3's usage--delivering responses twice as fast as Qwen3-235B. This efficiency establishes TopoMAS as an accelerator for computation-driven discovery pipelines. By harmonizing rational agent orchestration with a self-evolving knowledge graph, our framework not only delivers immediate advances in topological materials but also establishes a transferable, extensible paradigm for materials-science domain.", "date": "2025-07-05", "categories": []}
{"id": "2507.04493", "title": "Machine Learning-Based Prediction of Metal-Organic Framework Materials: A Comparative Analysis of Multiple Models", "abstract": "Metal-organic frameworks (MOFs) have emerged as promising materials for various applications due to their unique structural properties and versatile functionalities. This study presents a comprehensive investigation of machine learning approaches for predicting MOF material properties. We employed five different machine learning models: Random Forest, XGBoost, LightGBM, Support Vector Machine, and Neural Network, to analyze and predict MOF characteristics using a dataset from the Kaggle platform. The models were evaluated using multiple performance metrics, including RMSE, R^2, MAE, and cross-validation scores. Results demonstrated that the Random Forest model achieved superior performance with an R^2 value of 0.891 and RMSE of 0.152, significantly outperforming other models. LightGBM showed remarkable computational efficiency, completing training in 25.7 seconds while maintaining high accuracy. Our comparative analysis revealed that ensemble learning methods generally exhibited better performance than traditional single models in MOF property prediction. This research provides valuable insights into the application of machine learning in materials science and establishes a robust framework for future MOF material design and property prediction.", "date": "2025-07-06", "categories": []}
{"id": "2502.18314", "title": "Learning atomic forces from uncertainty-calibrated adversarial attacks", "abstract": "Adversarial approaches, which intentionally challenge machine learning models by generating difficult examples, are increasingly being adopted to improve machine learning interatomic potentials (MLIPs). While already providing great practical value, little is known about the actual prediction errors of MLIPs on adversarial structures and whether these errors can be controlled. We propose the Calibrated Adversarial Geometry Optimization (CAGO) algorithm to discover adversarial structures with user-assigned errors. Through uncertainty calibration, the estimated uncertainty of MLIPs is unified with real errors. By performing geometry optimization for calibrated uncertainty, we reach adversarial structures with the user-assigned target MLIP prediction error. Integrating with active learning pipelines, we benchmark CAGO, demonstrating stable MLIPs that systematically converge structural, dynamical, and thermodynamical properties for liquid water and water adsorption in a metal-organic framework within only hundreds of training structures, where previously many thousands were typically required.", "date": "2025-02-25", "categories": []}
{"id": "2507.01054", "title": "XxaCT-NN: Structure Agnostic Multimodal Learning for Materials Science", "abstract": "Recent advances in materials discovery have been driven by structure-based models, particularly those using crystal graphs. While effective for computational datasets, these models are impractical for real-world applications where atomic structures are often unknown or difficult to obtain. We propose a scalable multimodal framework that learns directly from elemental composition and X-ray diffraction (XRD) -- two of the more available modalities in experimental workflows without requiring crystal structure input. Our architecture integrates modality-specific encoders with a cross-attention fusion module and is trained on the 5-million-sample Alexandria dataset. We present masked XRD modeling (MXM), and apply MXM and contrastive alignment as self-supervised pretraining strategies. Pretraining yields faster convergence (up to 4.2x speedup) and improves both accuracy and representation quality. We further demonstrate that multimodal performance scales more favorably with dataset size than unimodal baselines, with gains compounding at larger data regimes. Our results establish a path toward structure-free, experimentally grounded foundation models for materials science.", "date": "2025-06-27", "categories": []}
{"id": "2507.01913", "title": "Advancing Magnetic Materials Discovery -- A structure-based machine learning approach for magnetic ordering and magnetic moment prediction", "abstract": "Accurately predicting magnetic behavior across diverse materials systems remains a longstanding challenge due to the complex interplay of structural and electronic factors and is pivotal for the accelerated discovery and design of next-generation magnetic materials. In this work, a refined descriptor is proposed that significantly improves the prediction of two critical magnetic properties -- magnetic ordering (Ferromagnetic vs. Ferrimagnetic) and magnetic moment per atom -- using only the structural information of materials. Unlike previous models limited to Mn-based or lanthanide-transition metal compounds, the present approach generalizes across a diverse dataset of 5741 stable, binary and ternary, ferromagnetic and ferrimagnetic compounds sourced from the Materials Project. Leveraging an enriched elemental vector representation and advanced feature engineering, including nonlinear terms and reduced matrix sparsity, the LightGBM-based model achieves an accuracy of 82.4% for magnetic ordering classification and balanced recall across FM and FiM classes, addressing a key limitation in prior studies. The model predicts magnetic moment per atom with a correlation coefficient of 0.93, surpassing the Hund's matrix and orbital field matrix descriptors. Additionally, it accurately estimates formation energy per atom, enabling assessment of both magnetic behavior and material stability. This generalized and computationally efficient framework offers a robust tool for high-throughput screening of magnetic materials with tailored properties.", "date": "2025-07-02", "categories": []}
{"id": "2501.02144", "title": "Establishing baselines for generative discovery of inorganic crystals", "abstract": "Generative artificial intelligence offers a promising avenue for materials discovery, yet its advantages over traditional methods remain unclear. In this work, we introduce and benchmark two baseline approaches - random enumeration of charge-balanced prototypes and data-driven ion exchange of known compounds - against four generative techniques based on diffusion models, variational autoencoders, and large language models. Our results show that established methods such as ion exchange are better at generating novel materials that are stable, although many of these closely resemble known compounds. In contrast, generative models excel at proposing novel structural frameworks and, when sufficient training data is available, can more effectively target properties such as electronic band gap and bulk modulus. To enhance the performance of both the baseline and generative approaches, we implement a post-generation screening step in which all proposed structures are passed through stability and property filters from pre-trained machine learning models including universal interatomic potentials. This low-cost filtering step leads to substantial improvement in the success rates of all methods, remains computationally efficient, and ultimately provides a practical pathway toward more effective generative strategies for materials discovery. By establishing baselines for comparison, this work highlights opportunities for continued advancement of generative models, especially for the targeted generation of novel materials that are thermodynamically stable.", "date": "2025-01-04", "categories": []}
{"id": "2502.13899", "title": "AI-Driven Discovery of High Performance Polymer Electrodes for Next-Generation Batteries", "abstract": "The use of transition group metals in electric batteries requires extensive usage of critical elements like lithium, cobalt and nickel, which poses significant environmental challenges. Replacing these metals with redox-active organic materials offers a promising alternative, thereby reducing the carbon footprint of batteries by one order of magnitude. However, this approach faces critical obstacles, including the limited availability of suitable redox-active organic materials and issues such as lower electronic conductivity, voltage, specific capacity, and long-term stability. To overcome the limitations for lower voltage and specific capacity, a machine learning (ML) driven battery informatics framework is developed and implemented. This framework utilizes an extensive battery dataset and advanced ML techniques to accelerate and enhance the identification, optimization, and design of redox-active organic materials. In this contribution, a data-fusion ML coupled meta learning model capable of predicting the battery properties, voltage and specific capacity, for various organic negative electrodes and charge carriers (positive electrode materials) combinations is presented. The ML models accelerate experimentation, facilitate the inverse design of battery materials, and identify suitable candidates from three extensive material libraries to advance sustainable energy-storage technologies.", "date": "2025-02-19", "categories": []}
{"id": "2505.20094", "title": "SwarmThinkers: Learning Physically Consistent Atomic KMC Transitions at Scale", "abstract": "Can a scientific simulation system be physically consistent, interpretable by design, and scalable across regimes--all at once? Despite decades of progress, this trifecta remains elusive. Classical methods like Kinetic Monte Carlo ensure thermodynamic accuracy but scale poorly; learning-based methods offer efficiency but often sacrifice physical consistency and interpretability. We present SwarmThinkers, a reinforcement learning framework that recasts atomic-scale simulation as a physically grounded swarm intelligence system. Each diffusing particle is modeled as a local decision-making agent that selects transitions via a shared policy network trained under thermodynamic constraints. A reweighting mechanism fuses learned preferences with transition rates, preserving statistical fidelity while enabling interpretable, step-wise decision making. Training follows a centralized-training, decentralized-execution paradigm, allowing the policy to generalize across system sizes, concentrations, and temperatures without retraining. On a benchmark simulating radiation-induced Fe-Cu alloy precipitation, SwarmThinkers is the first system to achieve full-scale, physically consistent simulation on a single A100 GPU, previously attainable only via OpenKMC on a supercomputer. It delivers up to 4963x (3185x on average) faster computation with 485x lower memory usage. By treating particles as decision-makers, not passive samplers, SwarmThinkers marks a paradigm shift in scientific simulation--one that unifies physical consistency, interpretability, and scalability through agent-driven intelligence.", "date": "2025-05-26", "categories": []}
{"id": "2507.00024", "title": "AIMatDesign: Knowledge-Augmented Reinforcement Learning for Inverse Materials Design under Data Scarcity", "abstract": "With the growing demand for novel materials, machine learning-driven inverse design methods face significant challenges in reconciling the high-dimensional materials composition space with limited experimental data. Existing approaches suffer from two major limitations: (I) machine learning models often lack reliability in high-dimensional spaces, leading to prediction biases during the design process; (II) these models fail to effectively incorporate domain expert knowledge, limiting their capacity to support knowledge-guided inverse design. To address these challenges, we introduce AIMatDesign, a reinforcement learning framework that addresses these limitations by augmenting experimental data using difference-based algorithms to build a trusted experience pool, accelerating model convergence. To enhance model reliability, an automated refinement strategy guided by large language models (LLMs) dynamically corrects prediction inconsistencies, reinforcing alignment between reward signals and state value functions. Additionally, a knowledge-based reward function leverages expert domain rules to improve stability and efficiency during training. Our experiments demonstrate that AIMatDesign significantly surpasses traditional machine learning and reinforcement learning methods in discovery efficiency, convergence speed, and success rates. Among the numerous candidates proposed by AIMatDesign, experimental synthesis of representative Zr-based alloys yielded a top-performing BMG with 1.7GPa yield strength and 10.2\\% elongation, closely matching predictions. Moreover, the framework accurately captured the trend of yield strength variation with composition, demonstrating its reliability and potential for closed-loop materials discovery.", "date": "2025-06-17", "categories": []}
{"id": "2412.17283", "title": "Emerging Microelectronic Materials by Design: Navigating Combinatorial Design Space with Scarce and Dispersed Data", "abstract": "The increasing demands of sustainable energy, electronics, and biomedical applications call for next-generation functional materials with unprecedented properties. Of particular interest are emerging materials that display exceptional physical properties, making them promising candidates in energy-efficient microelectronic devices. As the conventional Edisonian approach becomes significantly outpaced by growing societal needs, emerging computational modeling and machine learning (ML) methods are employed for the rational design of materials. However, the complex physical mechanisms, cost of first-principles calculations, and the dispersity and scarcity of data pose challenges to both physics-based and data-driven materials modeling. Moreover, the combinatorial composition-structure design space is high-dimensional and often disjoint, making design optimization nontrivial. In this Account, we review a team effort toward establishing a framework that integrates data-driven and physics-based methods to address these challenges and accelerate materials design. We begin by presenting our integrated materials design framework and its three components in a general context. We then provide an example of applying this materials design framework to metal-insulator transition (MIT) materials, a specific type of emerging materials with practical importance in next-generation memory technologies. We identify multiple new materials which may display this property and propose pathways for their synthesis. Finally, we identify some outstanding challenges in data-driven materials design, such as materials data quality issues and property-performance mismatch. We seek to raise awareness of these overlooked issues hindering materials design, thus stimulating efforts toward developing methods to mitigate the gaps.", "date": "2024-12-23", "categories": []}
{"id": "2501.18637", "title": "Machine learning of microstructure--property relationships in materials leveraging microstructure representation from foundational vision transformers", "abstract": "Machine learning of microstructure--property relationships from data is an emerging approach in computational materials science. Most existing machine learning efforts focus on the development of task-specific models for each microstructure--property relationship. We propose utilizing pre-trained foundational vision transformers for the extraction of task-agnostic microstructure features and subsequent light-weight machine learning of a microstructure-dependent property. We demonstrate our approach with pre-trained state-of-the-art vision transformers (CLIP, DINOv2, SAM) in two case studies on machine-learning: (i) elastic modulus of two-phase microstructures based on simulations data; and (ii) Vicker's hardness of Ni-base and Co-base superalloys based on experimental data published in literature. Our results show the potential of foundational vision transformers for robust microstructure representation and efficient machine learning of microstructure--property relationships without the need for expensive task-specific training or fine-tuning of bespoke deep learning models.", "date": "2025-01-28", "categories": []}
{"id": "2506.20743", "title": "A Survey of AI for Materials Science: Foundation Models, LLM Agents, Datasets, and Tools", "abstract": "Foundation models (FMs) are catalyzing a transformative shift in materials science (MatSci) by enabling scalable, general-purpose, and multimodal AI systems for scientific discovery. Unlike traditional machine learning models, which are typically narrow in scope and require task-specific engineering, FMs offer cross-domain generalization and exhibit emergent capabilities. Their versatility is especially well-suited to materials science, where research challenges span diverse data types and scales. This survey provides a comprehensive overview of foundation models, agentic systems, datasets, and computational tools supporting this growing field. We introduce a task-driven taxonomy encompassing six broad application areas: data extraction, interpretation and Q\\&A; atomistic simulation; property prediction; materials structure, design and discovery; process planning, discovery, and optimization; and multiscale modeling. We discuss recent advances in both unimodal and multimodal FMs, as well as emerging large language model (LLM) agents. Furthermore, we review standardized datasets, open-source tools, and autonomous experimental platforms that collectively fuel the development and integration of FMs into research workflows. We assess the early successes of foundation models and identify persistent limitations, including challenges in generalizability, interpretability, data imbalance, safety concerns, and limited multimodal fusion. Finally, we articulate future research directions centered on scalable pretraining, continual learning, data governance, and trustworthiness.", "date": "2025-06-25", "categories": []}
{"id": "2506.20839", "title": "Uncertainty-Aware Machine-Learning Framework for Predicting Dislocation Plasticity and Stress-Strain Response in FCC Alloys", "abstract": "Machine learning has significantly advanced the understanding and application of structural materials, with an increasing emphasis on integrating existing data and quantifying uncertainties in predictive modeling. This study presents a comprehensive methodology utilizing a mixed density network (MDN) model, trained on extensive experimental data from literature. This approach uniquely predicts the distribution of dislocation density, inferred as a latent variable, and the resulting stress distribution at the grain level. The incorporation of statistical parameters of those predicted distributions into a dislocation-mediated plasticity model allows for accurate stress-strain predictions with explicit uncertainty quantification. This strategy not only improves the accuracy and reliability of mechanical property predictions but also plays a vital role in optimizing alloy design, thereby facilitating the development of new materials in a rapidly evolving industry.", "date": "2025-06-25", "categories": []}
{"id": "2503.17427", "title": "3D variational autoencoder for fingerprinting microstructure volume elements", "abstract": "Microstructure quantification is an important step towards establishing structure-property relationships in materials. Machine learning-based image processing methods have been shown to outperform conventional image processing techniques and are increasingly applied to microstructure quantification tasks. In this work, we present a 3D variational autoencoder (VAE) for encoding microstructure volume elements (VEs) comprising voxelated crystallographic orientation data. Crystal symmetries in the orientation space are accounted for by mapping to the crystallographic fundamental zone as a preprocessing step, which allows for a continuous loss function to be used and improves the training convergence rate. The VAE is then used to encode a training set of VEs with an equiaxed polycrystalline microstructure with random texture. Accurate reconstructions are achieved with a relative average misorientation error of 3x10^-2 on the test dataset, for a continuous latent space with dimension 256. We show that the model generalises well to microstructures with textures, grain sizes and aspect ratios outside the training distribution. Structure-property relationships are explored through using the training set of VEs as initial configurations in various crystal plasticity (CP) simulations. Microstructural fingerprints extracted from the VAE, which parameterise the VEs in a low-dimensional latent space, are stored alongside the volume-averaged stress response, at each strain increment, to uniaxial tensile deformation from CP simulations. This is then used to train a fully connected neural network mapping the input fingerprint to the resulting stress response, which acts as a surrogate model for the CP simulation. The fingerprint-based surrogate model is shown to accurately predict the microstructural dependence in the CP stress response, with a relative mean-squared error of 2.75 MPa on unseen test data.", "date": "2025-03-21", "categories": []}
{"id": "2506.19628", "title": "Operator Forces For Coarse-Grained Molecular Dynamics", "abstract": "Coarse-grained (CG) molecular dynamics simulations extend the length and time scale of atomistic simulations by replacing groups of correlated atoms with CG beads. Machine-learned coarse-graining (MLCG) has recently emerged as a promising approach to construct highly accurate force fields for CG molecular dynamics. However, the calibration of MLCG force fields typically hinges on force matching, which demands extensive reference atomistic trajectories with corresponding force labels. In practice, atomistic forces are often not recorded, making traditional force matching infeasible on pre-existing datasets. Recently, noise-based kernels have been introduced to adapt force matching to the low-data regime, including situations in which reference atomistic forces are not present. While this approach produces force fields which recapitulate slow collective motion, it introduces significant local distortions due to the corrupting effects of the noise-based kernel. In this work, we introduce more general kernels based on normalizing flows that substantially reduce these local distortions while preserving global conformational accuracy. We demonstrate our method on small proteins, showing that flow-based kernels can generate high-quality CG forces solely from configurational samples.", "date": "2025-06-24", "categories": []}
{"id": "2506.17680", "title": "Enhancing Stress-Strain Predictions with Seq2Seq and Cross-Attention based on Small Punch Test", "abstract": "This paper introduces a novel deep-learning approach to predict true stress-strain curves of high-strength steels from small punch test (SPT) load-displacement data. The proposed approach uses Gramian Angular Field (GAF) to transform load-displacement sequences into images, capturing spatial-temporal features and employs a Sequence-to-Sequence (Seq2Seq) model with an LSTM-based encoder-decoder architecture, enhanced by multi-head cross-attention to improved accuracy. Experimental results demonstrate that the proposed approach achieves superior prediction accuracy, with minimum and maximum mean absolute errors of 0.15 MPa and 5.58 MPa, respectively. The proposed method offers a promising alternative to traditional experimental techniques in materials science, enhancing the accuracy and efficiency of true stress-strain relationship predictions.", "date": "2025-06-21", "categories": []}
{"id": "2506.18497", "title": "Leveraging neural network interatomic potentials for a foundation model of chemistry", "abstract": "Large-scale foundation models, including neural network interatomic potentials (NIPs) in computational materials science, have demonstrated significant potential. However, despite their success in accelerating atomistic simulations, NIPs face challenges in directly predicting electronic properties and often require coupling to higher-scale models or extensive simulations for macroscopic properties. Machine learning (ML) offers alternatives for structure-to-property mapping but faces trade-offs: feature-based methods often lack generalizability, while deep neural networks require significant data and computational power. To address these trade-offs, we introduce HackNIP, a two-stage pipeline that leverages pretrained NIPs. This method first extracts fixed-length feature vectors (embeddings) from NIP foundation models and then uses these embeddings to train shallow ML models for downstream structure-to-property predictions. This study investigates whether such a hybridization approach, by ``hacking\" the NIP, can outperform end-to-end deep neural networks, determines the dataset size at which this transfer learning approach surpasses direct fine-tuning of the NIP, and identifies which NIP embedding depths yield the most informative features. HackNIP is benchmarked on Matbench, evaluated for data efficiency, and tested on diverse tasks including \\textit{ab initio}, experimental, and molecular properties. We also analyze how embedding depth impacts performance. This work demonstrates a hybridization strategy to overcome ML trade-offs in materials science, aiming to democratize high-performance predictive modeling.", "date": "2025-06-23", "categories": []}
{"id": "2501.12919", "title": "Bridging Text and Crystal Structures: Literature-driven Contrastive Learning for Materials Science", "abstract": "Understanding structure-property relationships is an essential yet challenging aspect of materials discovery and development. To facilitate this process, recent studies in materials informatics have sought latent embedding spaces of crystal structures to capture their similarities based on properties and functionalities. However, abstract feature-based embedding spaces are human-unfriendly and prevent intuitive and efficient exploration of the vast materials space. Here we introduce Contrastive Language--Structure Pre-training (CLaSP), a learning paradigm for constructing crossmodal embedding spaces between crystal structures and texts. CLaSP aims to achieve material embeddings that 1) capture property- and functionality-related similarities between crystal structures and 2) allow intuitive retrieval of materials via user-provided description texts as queries. To compensate for the lack of sufficient datasets linking crystal structures with textual descriptions, CLaSP leverages a dataset of over 400,000 published crystal structures and corresponding publication records, including paper titles and abstracts, for training. We demonstrate the effectiveness of CLaSP through text-based crystal structure screening and embedding space visualization.", "date": "2025-01-22", "categories": []}
{"id": "2506.15722", "title": "UniMate: A Unified Model for Mechanical Metamaterial Generation, Property Prediction, and Condition Confirmation", "abstract": "Metamaterials are artificial materials that are designed to meet unseen properties in nature, such as ultra-stiffness and negative materials indices. In mechanical metamaterial design, three key modalities are typically involved, i.e., 3D topology, density condition, and mechanical property. Real-world complex application scenarios place the demanding requirements on machine learning models to consider all three modalities together. However, a comprehensive literature review indicates that most existing works only consider two modalities, e.g., predicting mechanical properties given the 3D topology or generating 3D topology given the required properties. Therefore, there is still a significant gap for the state-of-the-art machine learning models capturing the whole. Hence, we propose a unified model named UNIMATE, which consists of a modality alignment module and a synergetic diffusion generation module. Experiments indicate that UNIMATE outperforms the other baseline models in topology generation task, property prediction task, and condition confirmation task by up to 80.2%, 5.1%, and 50.2%, respectively. We opensource our proposed UNIMATE model and corresponding results at https://github.com/wzhan24/UniMate.", "date": "2025-06-05", "categories": []}
{"id": "2506.16918", "title": "A Neural Operator based Hybrid Microscale Model for Multiscale Simulation of Rate-Dependent Materials", "abstract": "The behavior of materials is influenced by a wide range of phenomena occurring across various time and length scales. To better understand the impact of microstructure on macroscopic response, multiscale modeling strategies are essential. Numerical methods, such as the $\\text{FE}^2$ approach, account for micro-macro interactions to predict the global response in a concurrent manner. However, these methods are computationally intensive due to the repeated evaluations of the microscale. This challenge has led to the integration of deep learning techniques into computational homogenization frameworks to accelerate multiscale simulations. In this work, we employ neural operators to predict the microscale physics, resulting in a hybrid model that combines data-driven and physics-based approaches. This allows for physics-guided learning and provides flexibility for different materials and spatial discretizations. We apply this method to time-dependent solid mechanics problems involving viscoelastic material behavior, where the state is represented by internal variables only at the microscale. The constitutive relations of the microscale are incorporated into the model architecture and the internal variables are computed based on established physical principles. The results for homogenized stresses ($<6\\%$ error) show that the approach is computationally efficient ($\\sim 100 \\times$ faster).", "date": "2025-06-20", "categories": []}
{"id": "2506.14828", "title": "Accurate and Uncertainty-Aware Multi-Task Prediction of HEA Properties Using Prior-Guided Deep Gaussian Processes", "abstract": "Surrogate modeling techniques have become indispensable in accelerating the discovery and optimization of high-entropy alloys(HEAs), especially when integrating computational predictions with sparse experimental observations. This study systematically evaluates the fitting performance of four prominent surrogate models conventional Gaussian Processes(cGP), Deep Gaussian Processes(DGP), encoder-decoder neural networks for multi-output regression and XGBoost applied to a hybrid dataset of experimental and computational properties in the AlCoCrCuFeMnNiV HEA system. We specifically assess their capabilities in predicting correlated material properties, including yield strength, hardness, modulus, ultimate tensile strength, elongation, and average hardness under dynamic and quasi-static conditions, alongside auxiliary computational properties. The comparison highlights the strengths of hierarchical and deep modeling approaches in handling heteroscedastic, heterotopic, and incomplete data commonly encountered in materials informatics. Our findings illustrate that DGP infused with machine learning-based prior outperform other surrogates by effectively capturing inter-property correlations and input-dependent uncertainty. This enhanced predictive accuracy positions advanced surrogate models as powerful tools for robust and data-efficient materials design.", "date": "2025-06-13", "categories": []}
{"id": "2506.12516", "title": "Information fusion strategy integrating pre-trained language model and contrastive learning for materials knowledge mining", "abstract": "Machine learning has revolutionized materials design, yet predicting complex properties like alloy ductility remains challenging due to the influence of processing conditions and microstructural features that resist quantification through traditional reductionist approaches. Here, we present an innovative information fusion architecture that integrates domain-specific texts from materials science literature with quantitative physical descriptors to overcome these limitations. Our framework employs MatSciBERT for advanced textual comprehension and incorporates contrastive learning to automatically extract implicit knowledge regarding processing parameters and microstructural characteristics. Through rigorous ablation studies and comparative experiments, the model demonstrates superior performance, achieving coefficient of determination (R2) values of 0.849 and 0.680 on titanium alloy validation set and refractory multi-principal-element alloy test set. This systematic approach provides a holistic framework for property prediction in complex material systems where quantitative descriptors are incomplete and establishes a foundation for knowledge-guided materials design and informatics-driven materials discovery.", "date": "2025-06-14", "categories": []}
{"id": "2506.12557", "title": "Language Models Enable Data-Augmented Synthesis Planning for Inorganic Materials", "abstract": "Inorganic synthesis planning currently relies primarily on heuristic approaches or machine-learning models trained on limited datasets, which constrains its generality. We demonstrate that language models, without task-specific fine-tuning, can recall synthesis conditions. Off-the-shelf models, such as GPT-4.1, Gemini 2.0 Flash and Llama 4 Maverick, achieve a Top-1 precursor-prediction accuracy of up to 53.8 % and a Top-5 performance of 66.1 % on a held-out set of 1,000 reactions. They also predict calcination and sintering temperatures with mean absolute errors below 126 {\\deg}C, matching specialized regression methods. Ensembling these language models further enhances predictive accuracy and reduces inference cost per prediction by up to 70 %. We subsequently employ language models to generate 28,548 synthetic reaction recipes, which we combine with literature-mined examples to pretrain a transformer-based model, SyntMTE. After fine-tuning on the combined dataset, SyntMTE reduces mean-absolute error in sintering temperature prediction to 73 {\\deg}C and in calcination temperature to 98 {\\deg}C. This strategy improves models by up to 8.7 % compared with baselines trained exclusively on experimental data. Finally, in a case study on Li7La3Zr2O12 solid-state electrolytes, we demonstrate that SyntMTE reproduces the experimentally observed dopant-dependent sintering trends. Our hybrid workflow enables scalable, data-efficient inorganic synthesis planning.", "date": "2025-06-14", "categories": []}
{"id": "2503.14293", "title": "Ensemble Knowledge Distillation for Machine Learning Interatomic Potentials", "abstract": "The quality of machine learning interatomic potentials (MLIPs) strongly depends on the quantity of training data as well as the quantum chemistry (QC) level of theory used. Datasets generated with high-fidelity QC methods are typically restricted to small molecules and may be missing energy gradients, which make it difficult to train accurate MLIPs. We present an ensemble knowledge distillation (EKD) method to improve MLIP accuracy when trained to energy-only datasets. First, multiple teacher models are trained to QC energies and then generate atomic forces for all configurations in the dataset. Next, the student MLIP is trained to both QC energies and to ensemble-averaged forces generated by the teacher models. We apply this workflow on the ANI-1ccx dataset where the configuration energies computed at the coupled cluster level of theory. The resulting student MLIPs achieve new state-of-the-art accuracy on the COMP6 benchmark and show improved stability for molecular dynamics simulations.", "date": "2025-03-18", "categories": []}
{"id": "2506.11055", "title": "PolyMicros: Bootstrapping a Foundation Model for Polycrystalline Material Structure", "abstract": "Recent advances in Foundation Models for Materials Science are poised to revolutionize the discovery, manufacture, and design of novel materials with tailored properties and responses. Although great strides have been made, successes have been restricted to materials classes where multi-million sample data repositories can be readily curated (e.g., atomistic structures). Unfortunately, for many structural and functional materials (e.g., mesoscale structured metal alloys), such datasets are too costly or prohibitive to construct; instead, datasets are limited to very few examples. To address this challenge, we introduce a novel machine learning approach for learning from hyper-sparse, complex spatial data in scientific domains. Our core contribution is a physics-driven data augmentation scheme that leverages an ensemble of local generative models, trained on as few as five experimental observations, and coordinates them through a novel diversity curation strategy to generate a large-scale, physically diverse dataset. We utilize this framework to construct PolyMicros, the first Foundation Model for polycrystalline materials (a structural material class important across a broad range of industrial and scientific applications). We demonstrate the utility of PolyMicros by zero-shot solving several long standing challenges related to accelerating 3D experimental microscopy. Finally, we make both our models and datasets openly available to the community.", "date": "2025-05-22", "categories": []}
{"id": "2506.11332", "title": "Polymorphism Crystal Structure Prediction with Adaptive Space Group Diversity Control", "abstract": "Crystalline materials can form different structural arrangements (i.e. polymorphs) with the same chemical composition, exhibiting distinct physical properties depending on how they were synthesized or the conditions under which they operate. For example, carbon can exist as graphite (soft, conductive) or diamond (hard, insulating). Computational methods that can predict these polymorphs are vital in materials science, which help understand stability relationships, guide synthesis efforts, and discover new materials with desired properties without extensive trial-and-error experimentation. However, effective crystal structure prediction (CSP) algorithms for inorganic polymorph structures remain limited. We propose ParetoCSP2, a multi-objective genetic algorithm for polymorphism CSP that incorporates an adaptive space group diversity control technique, preventing over-representation of any single space group in the population guided by a neural network interatomic potential. Using an improved population initialization method and performing iterative structure relaxation, ParetoCSP2 not only alleviates premature convergence but also achieves improved convergence speed. Our results show that ParetoCSP2 achieves excellent performance in polymorphism prediction, including a nearly perfect space group and structural similarity accuracy for formulas with two polymorphs but with the same number of unit cell atoms. Evaluated on a benchmark dataset, it outperforms baseline algorithms by factors of 2.46-8.62 for these accuracies and improves by 44.8\\%-87.04\\% across key performance metrics for regular CSP. Our source code is freely available at https://github.com/usccolumbia/ParetoCSP2.", "date": "2025-06-12", "categories": []}
{"id": "2506.10944", "title": "Coupled reaction and diffusion governing interface evolution in solid-state batteries", "abstract": "Understanding and controlling the atomistic-level reactions governing the formation of the solid-electrolyte interphase (SEI) is crucial for the viability of next-generation solid state batteries. However, challenges persist due to difficulties in experimentally characterizing buried interfaces and limits in simulation speed and accuracy. We conduct large-scale explicit reactive simulations with quantum accuracy for a symmetric battery cell, {\\symcell}, enabled by active learning and deep equivariant neural network interatomic potentials. To automatically characterize the coupled reactions and interdiffusion at the interface, we formulate and use unsupervised classification techniques based on clustering in the space of local atomic environments. Our analysis reveals the formation of a previously unreported crystalline disordered phase, Li$_2$S$_{0.72}$P$_{0.14}$Cl$_{0.14}$, in the SEI, that evaded previous predictions based purely on thermodynamics, underscoring the importance of explicit modeling of full reaction and transport kinetics. Our simulations agree with and explain experimental observations of the SEI formations and elucidate the Li creep mechanisms, critical to dendrite initiation, characterized by significant Li motion along the interface. Our approach is to crease a digital twin from first principles, without adjustable parameters fitted to experiment. As such, it offers capabilities to gain insights into atomistic dynamics governing complex heterogeneous processes in solid-state synthesis and electrochemistry.", "date": "2025-06-12", "categories": []}
{"id": "2505.21646", "title": "Iterative Corpus Refinement for Materials Property Prediction Based on Scientific Texts", "abstract": "The discovery and optimization of materials for specific applications is hampered by the practically infinite number of possible elemental combinations and associated properties, also known as the `combinatorial explosion'. By nature of the problem, data are scarce and all possible data sources should be used. In addition to simulations and experimental results, the latent knowledge in scientific texts is not yet used to its full potential. We present an iterative framework that refines a given scientific corpus by strategic selection of the most diverse documents, training Word2Vec models, and monitoring the convergence of composition-property correlations in embedding space. Our approach is applied to predict high-performing materials for oxygen reduction (ORR), hydrogen evolution (HER), and oxygen evolution (OER) reactions for a large number of possible candidate compositions. Our method successfully predicts the highest performing compositions among a large pool of candidates, validated by experimental measurements of the electrocatalytic performance in the lab. This work demonstrates and validates the potential of iterative corpus refinement to accelerate materials discovery and optimization, offering a scalable and efficient tool for screening large compositional spaces where reliable data are scarce or non-existent.", "date": "2025-05-27", "categories": []}
{"id": "2506.01994", "title": "Re-experiment Smart: a Novel Method to Enhance Data-driven Prediction of Mechanical Properties of Epoxy Polymers", "abstract": "Accurate prediction of polymer material properties through data-driven approaches greatly accelerates novel material development by reducing redundant experiments and trial-and-error processes. However, inevitable outliers in empirical measurements can severely skew machine learning results, leading to erroneous prediction models and suboptimal material designs. To address this limitation, we propose a novel approach to enhance dataset quality efficiently by integrating multi-algorithm outlier detection with selective re-experimentation of unreliable outlier cases. To validate the empirical effectiveness of the approach, we systematically construct a new dataset containing 701 measurements of three key mechanical properties: glass transition temperature ($T_g$), tan $\\delta$ peak, and crosslinking density ($v_{c}$). To demonstrate its general applicability, we report the performance improvements across multiple machine learning models, including Elastic Net, SVR, Random Forest, and TPOT, to predict the three key properties. Our method reliably reduces prediction error (RMSE) and significantly improves accuracy with minimal additional experimental work, requiring only about 5% of the dataset to be re-measured. These findings highlight the importance of data quality enhancement in achieving reliable machine learning applications in polymer science and present a scalable strategy for improving predictive reliability in materials science.", "date": "2025-05-19", "categories": []}
{"id": "2407.19326", "title": "Accounting for plasticity: An extension of inelastic Constitutive Artificial Neural Networks", "abstract": "In this work, we extend the existing framework of inelastic constitutive artificial neural networks (iCANNs) by incorporating plasticity to increase their applicability to model more complex material behavior. The proposed approach ensures objectivity, material symmetry, and thermodynamic consistency, providing a robust basis for automatic model discovery of constitutive equations at finite strains. These are predicted by discovering formulations for the Helmholtz free energy and plastic potentials for the yield function and evolution equations in terms of feed-forward networks. Our framework captures both linear and nonlinear kinematic hardening behavior. Investigation of our model's prediction showed that the extended iCANNs successfully predict both linear and nonlinear kinematic hardening behavior based on experimental and artificially generated datasets, showcasing promising capabilities of this framework. Nonetheless, challenges remain in discovering more complex yield criteria with tension-compression asymmetry and addressing deviations in experimental data at larger strains. Despite these limitations, the proposed framework provides a promising basis for incorporating plasticity into iCANNs, offering a platform for advancing in the field of automated model discovery.", "date": "2024-07-27", "categories": []}
{"id": "2505.06711", "title": "Efficient Parallelization of Message Passing Neural Network Potentials for Large-scale Molecular Dynamics", "abstract": "Machine learning potentials have achieved great success in accelerating atomistic simulations. Many of them relying on atom-centered local descriptors are natural for parallelization. More recent message passing neural network (MPNN) models have demonstrated their superior accuracy and become increasingly popular. However, efficiently parallelizing MPNN models across multiple nodes remains challenging, limiting their practical applications in large-scale simulations. Here, we propose an efficient parallel algorithm for MPNN models, in which additional data communication is minimized among local atoms only in each MP layer without redundant computation, thus scaling linearly with the layer number. Integrated with our recursively embedded atom neural network model, this algorithm demonstrates excellent strong scaling and weak scaling behaviors in several benchmark systems. This approach enables massive molecular dynamics simulations on MPNN models as fast as on strictly local models for over 100 million atoms, vastly extending the applicability of the MPNN potential to an unprecedented scale. This general parallelization framework can empower various MPNN models to efficiently simulate very large and complex systems.", "date": "2025-05-10", "categories": []}
{"id": "2506.05616", "title": "Toward Greater Autonomy in Materials Discovery Agents: Unifying Planning, Physics, and Scientists", "abstract": "We aim at designing language agents with greater autonomy for crystal materials discovery. While most of existing studies restrict the agents to perform specific tasks within predefined workflows, we aim to automate workflow planning given high-level goals and scientist intuition. To this end, we propose Materials Agent unifying Planning, Physics, and Scientists, known as MAPPS. MAPPS consists of a Workflow Planner, a Tool Code Generator, and a Scientific Mediator. The Workflow Planner uses large language models (LLMs) to generate structured and multi-step workflows. The Tool Code Generator synthesizes executable Python code for various tasks, including invoking a force field foundation model that encodes physics. The Scientific Mediator coordinates communications, facilitates scientist feedback, and ensures robustness through error reflection and recovery. By unifying planning, physics, and scientists, MAPPS enables flexible and reliable materials discovery with greater autonomy, achieving a five-fold improvement in stability, uniqueness, and novelty rates compared with prior generative models when evaluated on the MP-20 data. We provide extensive experiments across diverse tasks to show that MAPPS is a promising framework for autonomous materials discovery.", "date": "2025-06-05", "categories": []}
{"id": "2506.07083", "title": "Inverse Design of Metamaterials with Manufacturing-Guiding Spectrum-to-Structure Conditional Diffusion Model", "abstract": "Metamaterials are artificially engineered structures that manipulate electromagnetic waves, having optical properties absent in natural materials. Recently, machine learning for the inverse design of metamaterials has drawn attention. However, the highly nonlinear relationship between the metamaterial structures and optical behaviour, coupled with fabrication difficulties, poses challenges for using machine learning to design and manufacture complex metamaterials. Herein, we propose a general framework that implements customised spectrum-to-shape and size parameters to address one-to-many metamaterial inverse design problems using conditional diffusion models. Our method exhibits superior spectral prediction accuracy, generates a diverse range of patterns compared to other typical generative models, and offers valuable prior knowledge for manufacturing through the subsequent analysis of the diverse generated results, thereby facilitating the experimental fabrication of metamaterial designs. We demonstrate the efficacy of the proposed method by successfully designing and fabricating a free-form metamaterial with a tailored selective emission spectrum for thermal camouflage applications.", "date": "2025-06-08", "categories": []}
{"id": "2503.05617", "title": "Can KAN CANs? Input-convex Kolmogorov-Arnold Networks (KANs) as hyperelastic constitutive artificial neural networks (CANs)", "abstract": "Traditional constitutive models rely on hand-crafted parametric forms with limited expressivity and generalizability, while neural network-based models can capture complex material behavior but often lack interpretability. To balance these trade-offs, we present monotonic Input-Convex Kolmogorov-Arnold Networks (ICKANs) for learning polyconvex hyperelastic constitutive laws. ICKANs leverage the Kolmogorov-Arnold representation, decomposing the model into compositions of trainable univariate spline-based activation functions for rich expressivity. We introduce trainable monotonic input-convex splines within the KAN architecture, ensuring physically admissible polyconvex models for isotropic compressible hyperelasticity. The resulting models are both compact and interpretable, enabling explicit extraction of analytical constitutive relationships through a monotonic input-convex symbolic regression technique. Through unsupervised training on full-field strain data and limited global force measurements, ICKANs accurately capture nonlinear stress-strain behavior across diverse strain states. Finite element simulations of unseen geometries with trained ICKAN hyperelastic constitutive models confirm the framework's robustness and generalization capability.", "date": "2025-03-07", "categories": []}
{"id": "2506.03910", "title": "Enhancing Experimental Efficiency in Materials Design: A Comparative Study of Taguchi and Machine Learning Methods", "abstract": "Materials design problems often require optimizing multiple variables, rendering full factorial exploration impractical. Design of experiment (DOE) methods, such as Taguchi technique, are commonly used to efficiently sample the design space but they inherently lack the ability to capture non-linear dependency of process variables. In this work, we demonstrate how machine learning (ML) methods can be used to overcome these limitations. We compare the performance of Taguchi method against an active learning based Gaussian process regression (GPR) model in a wire arc additive manufacturing (WAAM) process to accurately predict aspects of bead geometry, including penetration depth, bead width, and height. While Taguchi method utilized a three-factor, five-level L25 orthogonal array to suggest weld parameters, the GPR model used an uncertainty-based exploration acquisition function coupled with latin hypercube sampling for initial training data. Accuracy and efficiency of both models was evaluated on 15 test cases, with GPR outperforming Taguchi in both metrics. This work applies to broader materials processing domain requiring efficient exploration of complex parameters.", "date": "2025-06-04", "categories": []}
{"id": "2506.04055", "title": "chemtrain-deploy: A parallel and scalable framework for machine learning potentials in million-atom MD simulations", "abstract": "Machine learning potentials (MLPs) have advanced rapidly and show great promise to transform molecular dynamics (MD) simulations. However, most existing software tools are tied to specific MLP architectures, lack integration with standard MD packages, or are not parallelizable across GPUs. To address these challenges, we present chemtrain-deploy, a framework that enables model-agnostic deployment of MLPs in LAMMPS. chemtrain-deploy supports any JAX-defined semi-local potential, allowing users to exploit the functionality of LAMMPS and perform large-scale MLP-based MD simulations on multiple GPUs. It achieves state-of-the-art efficiency and scales to systems containing millions of atoms. We validate its performance and scalability using graph neural network architectures, including MACE, Allegro, and PaiNN, applied to a variety of systems, such as liquid-vapor interfaces, crystalline materials, and solvated peptides. Our results highlight the practical utility of chemtrain-deploy for real-world, high-performance simulations and provide guidance for MLP architecture selection and future design.", "date": "2025-06-04", "categories": []}
{"id": "2506.04195", "title": "MACS: Multi-Agent Reinforcement Learning for Optimization of Crystal Structures", "abstract": "Geometry optimization of atomic structures is a common and crucial task in computational chemistry and materials design. Following the learning to optimize paradigm, we propose a new multi-agent reinforcement learning method called Multi-Agent Crystal Structure optimization (MACS) to address periodic crystal structure optimization. MACS treats geometry optimization as a partially observable Markov game in which atoms are agents that adjust their positions to collectively discover a stable configuration. We train MACS across various compositions of reported crystalline materials to obtain a policy that successfully optimizes structures from the training compositions as well as structures of larger sizes and unseen compositions, confirming its excellent scalability and zero-shot transferability. We benchmark our approach against a broad range of state-of-the-art optimization methods and demonstrate that MACS optimizes periodic crystal structures significantly faster, with fewer energy calculations, and the lowest failure rate.", "date": "2025-06-04", "categories": []}
{"id": "2506.00198", "title": "MOFGPT: Generative Design of Metal-Organic Frameworks using Language Models", "abstract": "The discovery of Metal-Organic Frameworks (MOFs) with application-specific properties remains a central challenge in materials chemistry, owing to the immense size and complexity of their structural design space. Conventional computational screening techniques such as molecular simulations and density functional theory (DFT), while accurate, are computationally prohibitive at scale. Machine learning offers an exciting alternative by leveraging data-driven approaches to accelerate materials discovery. The complexity of MOFs, with their extended periodic structures and diverse topologies, creates both opportunities and challenges for generative modeling approaches. To address these challenges, we present a reinforcement learning-enhanced, transformer-based framework for the de novo design of MOFs. Central to our approach is MOFid, a chemically-informed string representation encoding both connectivity and topology, enabling scalable generative modeling. Our pipeline comprises three components: (1) a generative GPT model trained on MOFid sequences, (2) MOFormer, a transformer-based property predictor, and (3) a reinforcement learning (RL) module that optimizes generated candidates via property-guided reward functions. By integrating property feedback into sequence generation, our method drives the model toward synthesizable, topologically valid MOFs with desired functional attributes. This work demonstrates the potential of large language models, when coupled with reinforcement learning, to accelerate inverse design in reticular chemistry and unlock new frontiers in computational MOF discovery.", "date": "2025-05-30", "categories": []}
{"id": "2506.00725", "title": "A Foundation Model for Non-Destructive Defect Identification from Vibrational Spectra", "abstract": "Defects are ubiquitous in solids and strongly influence materials' mechanical and functional properties. However, non-destructive characterization and quantification of defects, especially when multiple types coexist, remain a long-standing challenge. Here we introduce DefectNet, a foundation machine learning model that predicts the chemical identity and concentration of substitutional point defects with multiple coexisting elements directly from vibrational spectra, specifically phonon density-of-states (PDoS). Trained on over 16,000 simulated spectra from 2,000 semiconductors, DefectNet employs a tailored attention mechanism to identify up to six distinct defect elements at concentrations ranging from 0.2% to 25%. The model generalizes well to unseen crystals across 56 elements and can be fine-tuned on experimental data. Validation using inelastic scattering measurements of SiGe alloys and MgB$_2$ superconductor demonstrates its accuracy and transferability. Our work establishes vibrational spectroscopy as a viable, non-destructive probe for point defect quantification in bulk materials, and highlights the promise of foundation models in data-driven defect engineering.", "date": "2025-05-31", "categories": []}
{"id": "2506.01868", "title": "NepTrain and NepTrainKit: Automated Active Learning and Visualization Toolkit for Neuroevolution Potentials", "abstract": "As a machine-learned potential, the neuroevolution potential (NEP) method features exceptional computational efficiency and has been successfully applied in materials science. Constructing high-quality training datasets is crucial for developing accurate NEP models. However, the preparation and screening of NEP training datasets remain a bottleneck for broader applications due to their time-consuming, labor-intensive, and resource-intensive nature. In this work, we have developed NepTrain and NepTrainKit, which are dedicated to initializing and managing training datasets to generate high-quality training sets while automating NEP model training. NepTrain is an open-source Python package that features a bond length filtering method to effectively identify and remove non-physical structures from molecular dynamics trajectories, thereby ensuring high-quality training datasets. NepTrainKit is a graphical user interface (GUI) software designed specifically for NEP training datasets, providing functionalities for data editing, visualization, and interactive exploration. It integrates key features such as outlier identification, farthest-point sampling, non-physical structure detection, and configuration type selection. The combination of these tools enables users to process datasets more efficiently and conveniently. Using $\\rm CsPbI_3$ as a case study, we demonstrate the complete workflow for training NEP models with NepTrain and further validate the models through materials property predictions. We believe this toolkit will greatly benefit researchers working with machine learning interatomic potentials.", "date": "2025-06-02", "categories": []}
{"id": "2503.05771", "title": "A Materials Foundation Model via Hybrid Invariant-Equivariant Architectures", "abstract": "Machine learning interatomic potentials (MLIPs) can predict energy, force, and stress of materials and enable a wide range of downstream discovery tasks. A key design choice in MLIPs involves the trade-off between invariant and equivariant architectures. Invariant models offer computational efficiency but may not perform as well, especially when predicting high-order outputs. In contrast, equivariant models can capture high-order symmetries, but are computationally expensive. In this work, we propose HIENet, a hybrid invariant-equivariant materials interatomic potential model that integrates both invariant and equivariant message passing layers, while provably satisfying key physical constraints. HIENet achieves state-of-the-art performance with considerable computational speedups over prior models. Experimental results on both common benchmarks and downstream materials discovery tasks demonstrate the efficiency and effectiveness of HIENet.", "date": "2025-02-25", "categories": []}
{"id": "2403.13627", "title": "A Straightforward Gradient-Based Approach for High-Tc Superconductor Design: Leveraging Domain Knowledge via Adaptive Constraints", "abstract": "Materials design aims to discover novel compounds with desired properties. However, prevailing strategies face critical trade-offs. Conventional element-substitution approaches readily and adaptively incorporate various domain knowledge but remain confined to a narrow search space. In contrast, deep generative models efficiently explore vast compositional landscapes, yet they struggle to flexibly integrate domain knowledge. To address these trade-offs, we propose a gradient-based material design framework that combines these strengths, offering both efficiency and adaptability. In our method, chemical compositions are optimised to achieve target properties by using property prediction models and their gradients. In order to seamlessly enforce diverse constraints, including those reflecting domain insights such as oxidation states, discretised compositional ratios, types of elements, and their abundance, we apply masks and employ a special loss function, namely the integer loss. Furthermore, we initialise the optimisation using promising candidates from existing dataset, effectively guiding the search away from unfavourable regions and thus helping to avoid poor solutions. Our approach demonstrates a more efficient exploration of superconductor candidates, uncovering candidate materials with higher critical temperature than conventional element-substitution and generative models. Importantly, it could propose new compositions beyond those found in existing databases, including new hydride superconductors absent from the training dataset but which share compositional similarities with materials found in literature. This synergy of domain knowledge and machine-learning-based scalability provides a robust foundation for rapid, adaptive, and comprehensive materials design for superconductors and beyond.", "date": "2024-03-20", "categories": []}
{"id": "2505.22208", "title": "LaMM: Semi-Supervised Pre-Training of Large-Scale Materials Models", "abstract": "Neural network potentials (NNPs) are crucial for accelerating computational materials science by surrogating density functional theory (DFT) calculations. Improving their accuracy is possible through pre-training and fine-tuning, where an NNP model is first pre-trained on a large-scale dataset and then fine-tuned on a smaller target dataset. However, this approach is computationally expensive, mainly due to the cost of DFT-based dataset labeling and load imbalances during large-scale pre-training. To address this, we propose LaMM, a semi-supervised pre-training method incorporating improved denoising self-supervised learning and a load-balancing algorithm for efficient multi-node training. We demonstrate that our approach effectively leverages a large-scale dataset of $\\sim$300 million semi-labeled samples to train a single NNP model, resulting in improved fine-tuning performance in terms of both speed and accuracy.", "date": "2025-05-28", "categories": []}
{"id": "2505.20299", "title": "MetamatBench: Integrating Heterogeneous Data, Computational Tools, and Visual Interface for Metamaterial Discovery", "abstract": "Metamaterials, engineered materials with architected structures across multiple length scales, offer unprecedented and tunable mechanical properties that surpass those of conventional materials. However, leveraging advanced machine learning (ML) for metamaterial discovery is hindered by three fundamental challenges: (C1) Data Heterogeneity Challenge arises from heterogeneous data sources, heterogeneous composition scales, and heterogeneous structure categories; (C2) Model Complexity Challenge stems from the intricate geometric constraints of ML models, which complicate their adaptation to metamaterial structures; and (C3) Human-AI Collaboration Challenge comes from the \"dual black-box'' nature of sophisticated ML models and the need for intuitive user interfaces. To tackle these challenges, we introduce a unified framework, named MetamatBench, that operates on three levels. (1) At the data level, we integrate and standardize 5 heterogeneous, multi-modal metamaterial datasets. (2) The ML level provides a comprehensive toolkit that adapts 17 state-of-the-art ML methods for metamaterial discovery. It also includes a comprehensive evaluation suite with 12 novel performance metrics with finite element-based assessments to ensure accurate and reliable model validation. (3) The user level features a visual-interactive interface that bridges the gap between complex ML techniques and non-ML researchers, advancing property prediction and inverse design of metamaterials for research and applications. MetamatBench offers a unified platform deployed at http://zhoulab-1.cs.vt.edu:5550 that enables machine learning researchers and practitioners to develop and evaluate new methodologies in metamaterial discovery. For accessibility and reproducibility, we open-source our benchmark and the codebase at https://github.com/cjpcool/Metamaterial-Benchmark.", "date": "2025-05-08", "categories": []}
{"id": "2502.03638", "title": "SymmCD: Symmetry-Preserving Crystal Generation with Diffusion Models", "abstract": "Generating novel crystalline materials has the potential to lead to advancements in fields such as electronics, energy storage, and catalysis. The defining characteristic of crystals is their symmetry, which plays a central role in determining their physical properties. However, existing crystal generation methods either fail to generate materials that display the symmetries of real-world crystals, or simply replicate the symmetry information from examples in a database. To address this limitation, we propose SymmCD, a novel diffusion-based generative model that explicitly incorporates crystallographic symmetry into the generative process. We decompose crystals into two components and learn their joint distribution through diffusion: 1) the asymmetric unit, the smallest subset of the crystal which can generate the whole crystal through symmetry transformations, and; 2) the symmetry transformations needed to be applied to each atom in the asymmetric unit. We also use a novel and interpretable representation for these transformations, enabling generalization across different crystallographic symmetry groups. We showcase the competitive performance of SymmCD on a subset of the Materials Project, obtaining diverse and valid crystals with realistic symmetries and predicted properties.", "date": "2025-02-05", "categories": []}
{"id": "2505.18162", "title": "Accelerating Battery Material Optimization through iterative Machine Learning", "abstract": "The performance of battery materials is determined by their composition and the processing conditions employed during commercial-scale fabrication, where raw materials undergo complex processing steps with various additives to yield final products. As the complexity of these parameters expands with the development of industry, conventional one-factor-at-a-time (OFAT) experiment becomes old fashioned. While domain expertise aids in parameter optimization, this traditional approach becomes increasingly vulnerable to cognitive limitations and anthropogenic biases as the complexity of factors grows. Herein, we introduce an iterative machine learning (ML) framework that integrates active learning to guide targeted experimentation and facilitate incremental model refinement. This method systematically leverages comprehensive experimental observations, including both successful and unsuccessful results, effectively mitigating human-induced biases and alleviating data scarcity. Consequently, it significantly accelerates exploration within the high-dimensional design space. Our results demonstrate that active-learning-driven experimentation markedly reduces the total number of experimental cycles necessary, underscoring the transformative potential of ML-based strategies in expediting battery material optimization.", "date": "2025-05-12", "categories": []}
{"id": "2505.18693", "title": "Simultaneous Optimization of Efficiency and Degradation in Tunable HTL-Free Perovskite Solar Cells with MWCNT-Integrated Back Contact Using a Machine Learning-Derived Polynomial Regressor", "abstract": "Perovskite solar cells (PSCs) without a hole transport layer (HTL) offer a cost-effective and stable alternative to conventional architectures, utilizing only an absorber layer and an electron transport layer (ETL). This study presents a machine learning (ML)-driven framework to optimize the efficiency and stability of HTL-free PSCs by integrating experimental validation with numerical simulations. Excellent agreement is achieved between a fabricated device and its simulated counterpart at a molar fraction \\( x = 68.7\\% \\) in \\(\\mathrm{MAPb}_{1-x}\\mathrm{Sb}_{2x/3}\\mathrm{I}_3\\), where MA is methylammonium. A dataset of 1650 samples is generated by varying molar fraction, absorber defect density, thickness, and ETL doping, with corresponding efficiency and 50-hour degradation as targets. A fourth-degree polynomial regressor (PR-4) shows the best performance, achieving RMSEs of 0.0179 and 0.0117, and \\( R^2 \\) scores of 1 and 0.999 for efficiency and degradation, respectively. The derived model generalizes beyond the training range and is used in an L-BFGS-B optimization algorithm with a weighted objective function to maximize efficiency and minimize degradation. This improves device efficiency from 13.7\\% to 16.84\\% and reduces degradation from 6.61\\% to 2.39\\% over 1000 hours. Finally, the dataset is labeled into superior and inferior classes, and a multilayer perceptron (MLP) classifier achieves 100\\% accuracy, successfully identifying optimal configurations.", "date": "2025-05-24", "categories": []}
{"id": "2503.03965", "title": "All-atom Diffusion Transformers: Unified generative modelling of molecules and materials", "abstract": "Diffusion models are the standard toolkit for generative modelling of 3D atomic systems. However, for different types of atomic systems -- such as molecules and materials -- the generative processes are usually highly specific to the target system despite the underlying physics being the same. We introduce the All-atom Diffusion Transformer (ADiT), a unified latent diffusion framework for jointly generating both periodic materials and non-periodic molecular systems using the same model: (1) An autoencoder maps a unified, all-atom representations of molecules and materials to a shared latent embedding space; and (2) A diffusion model is trained to generate new latent embeddings that the autoencoder can decode to sample new molecules or materials. Experiments on MP20, QM9 and GEOM-DRUGS datasets demonstrate that jointly trained ADiT generates realistic and valid molecules as well as materials, obtaining state-of-the-art results on par with molecule and crystal-specific models. ADiT uses standard Transformers with minimal inductive biases for both the autoencoder and diffusion model, resulting in significant speedups during training and inference compared to equivariant diffusion models. Scaling ADiT up to half a billion parameters predictably improves performance, representing a step towards broadly generalizable foundation models for generative chemistry. Open source code: https://github.com/facebookresearch/all-atom-diffusion-transformer", "date": "2025-03-05", "categories": []}
{"id": "2504.08859", "title": "PolyConf: Unlocking Polymer Conformation Generation through Hierarchical Generative Models", "abstract": "Polymer conformation generation is a critical task that enables atomic-level studies of diverse polymer materials. While significant advances have been made in designing conformation generation methods for small molecules and proteins, these methods struggle to generate polymer conformations due to their unique structural characteristics. Meanwhile, the scarcity of polymer conformation datasets further limits the progress, making this important area largely unexplored. In this work, we propose PolyConf, a pioneering tailored polymer conformation generation method that leverages hierarchical generative models to unlock new possibilities. Specifically, we decompose the polymer conformation into a series of local conformations (i.e., the conformations of its repeating units), generating these local conformations through an autoregressive model, and then generating their orientation transformations via a diffusion model to assemble them into the complete polymer conformation. Moreover, we develop the first benchmark with a high-quality polymer conformation dataset derived from molecular dynamics simulations to boost related research in this area. The comprehensive evaluation demonstrates that PolyConf consistently outperforms existing conformation generation methods, thus facilitating advancements in polymer modeling and simulation.", "date": "2025-04-11", "categories": []}
{"id": "2505.00625", "title": "SA-GAT-SR: Self-Adaptable Graph Attention Networks with Symbolic Regression for high-fidelity material property prediction", "abstract": "Recent advances in machine learning have demonstrated an enormous utility of deep learning approaches, particularly Graph Neural Networks (GNNs) for materials science. These methods have emerged as powerful tools for high-throughput prediction of material properties, offering a compelling enhancement and alternative to traditional first-principles calculations. While the community has predominantly focused on developing increasingly complex and universal models to enhance predictive accuracy, such approaches often lack physical interpretability and insights into materials behavior. Here, we introduce a novel computational paradigm, Self-Adaptable Graph Attention Networks integrated with Symbolic Regression (SA-GAT-SR), that synergistically combines the predictive capability of GNNs with the interpretative power of symbolic regression. Our framework employs a self-adaptable encoding algorithm that automatically identifies and adjust attention weights so as to screen critical features from an expansive 180-dimensional feature space while maintaining O(n) computational scaling. The integrated SR module subsequently distills these features into compact analytical expressions that explicitly reveal quantum-mechanically meaningful relationships, achieving 23 times acceleration compared to conventional SR implementations that heavily rely on first principle calculations-derived features as input. This work suggests a new framework in computational materials science, bridging the gap between predictive accuracy and physical interpretability, offering valuable physical insights into material behavior.", "date": "2025-05-01", "categories": []}
{"id": "2505.16379", "title": "Materials Generation in the Era of Artificial Intelligence: A Comprehensive Survey", "abstract": "Materials are the foundation of modern society, underpinning advancements in energy, electronics, healthcare, transportation, and infrastructure. The ability to discover and design new materials with tailored properties is critical to solving some of the most pressing global challenges. In recent years, the growing availability of high-quality materials data combined with rapid advances in Artificial Intelligence (AI) has opened new opportunities for accelerating materials discovery. Data-driven generative models provide a powerful tool for materials design by directly create novel materials that satisfy predefined property requirements. Despite the proliferation of related work, there remains a notable lack of up-to-date and systematic surveys in this area. To fill this gap, this paper provides a comprehensive overview of recent progress in AI-driven materials generation. We first organize various types of materials and illustrate multiple representations of crystalline materials. We then provide a detailed summary and taxonomy of current AI-driven materials generation approaches. Furthermore, we discuss the common evaluation metrics and summarize open-source codes and benchmark datasets. Finally, we conclude with potential future directions and challenges in this fast-growing field. The related sources can be found at https://github.com/ZhixunLEE/Awesome-AI-for-Materials-Generation.", "date": "2025-05-22", "categories": []}
{"id": "2404.00050", "title": "Grappa -- A Machine Learned Molecular Mechanics Force Field", "abstract": "Simulating large molecular systems over long timescales requires force fields that are both accurate and efficient. In recent years, E(3) equivariant neural networks have lifted the tension between computational efficiency and accuracy of force fields, but they are still several orders of magnitude more expensive than established molecular mechanics (MM) force fields. Here, we propose Grappa, a machine learning framework to predict MM parameters from the molecular graph, employing a graph attentional neural network and a transformer with symmetry-preserving positional encoding. The resulting Grappa force field outperformstabulated and machine-learned MM force fields in terms of accuracy at the same computational efficiency and can be used in existing Molecular Dynamics (MD) engines like GROMACS and OpenMM. It predicts energies and forces of small molecules, peptides, RNA and - showcasing its extensibility to uncharted regions of chemical space - radicals at state-of-the-art MM accuracy. We demonstrate Grappa's transferability to macromolecules in MD simulations from a small fast folding protein up to a whole virus particle. Our force field sets the stage for biomolecular simulations closer to chemical accuracy, but with the same computational cost as established protein force fields.", "date": "2024-03-25", "categories": []}
{"id": "2405.18944", "title": "Predicting Many Crystal Properties via an Adaptive Transformer-based Framework", "abstract": "Machine learning has revolutionized many fields, including materials science. However, predicting properties of crystalline materials using machine learning faces challenges in input encoding, output versatility, and interpretability. We introduce CrystalBERT, an adaptable transformer-based framework integrating space group, elemental, and unit cell information. This novel structure can seamlessly combine diverse features and accurately predict various physical properties, including topological properties, superconducting transition temperatures, dielectric constants, and more. CrystalBERT provides insightful interpretations of features influencing target properties. Our results indicate that space group and elemental information are crucial for predicting topological and superconducting properties, underscoring their intricate nature. By incorporating these features, we achieve 91\\% accuracy in topological classification, surpassing prior studies and identifying previously misclassified materials. This research demonstrates that integrating diverse material information enhances the prediction of complex material properties, paving the way for more accurate and interpretable machine learning models in materials science.", "date": "2024-05-29", "categories": []}
{"id": "2412.11970", "title": "DARWIN 1.5: Large Language Models as Materials Science Adapted Learners", "abstract": "Materials discovery and design aim to find compositions and structures with desirable properties over highly complex and diverse physical spaces. Traditional solutions, such as high-throughput simulations or machine learning, often rely on complex descriptors, which hinder generalizability and transferability across different material systems. Moreover, These descriptors may inadequately represent macro-scale material properties, which are influenced by structural imperfections and compositional variations in real-world samples, thus limiting their practical applicability. To address these challenges, we propose DARWIN 1.5, the largest open-source large language model tailored for materials science. By leveraging natural language as input, DARWIN eliminates the need for task-specific descriptors and enables a flexible, unified approach to material property prediction and discovery. Our approach integrates 6M material domain papers and 21 experimental datasets from 49,256 materials across modalities while enabling cross-task knowledge transfer. The enhanced model achieves up to 59.1% improvement in prediction accuracy over the base LLaMA-7B architecture and outperforms SOTA machine learning approaches across 8 materials design tasks. These results establish LLMs as a promising foundation for developing versatile and scalable models in materials science.", "date": "2024-12-16", "categories": []}
{"id": "2411.06804", "title": "Predicting ionic conductivity in solids from the machine-learned potential energy landscape", "abstract": "Discovering new superionic materials is essential for advancing solid-state batteries, which offer improved energy density and safety compared to the traditional lithium-ion batteries with liquid electrolytes. Conventional computational methods for identifying such materials are resource-intensive and not easily scalable. Recently, universal interatomic potential models have been developed using equivariant graph neural networks. These models are trained on extensive datasets of first-principles force and energy calculations. One can achieve significant computational advantages by leveraging them as the foundation for traditional methods of assessing the ionic conductivity, such as molecular dynamics or nudged elastic band techniques. However, the generalization error from model inference on diverse atomic structures arising in such calculations can compromise the reliability of the results. In this work, we propose an approach for the quick and reliable screening of ionic conductors through the analysis of a universal interatomic potential. Our method incorporates a set of heuristic structure descriptors that effectively employ the rich knowledge of the underlying model while requiring minimal generalization capabilities. Using our descriptors, we rank lithium-containing materials in the Materials Project database according to their expected ionic conductivity. Eight out of the ten highest-ranked materials are confirmed to be superionic at room temperature in first-principles calculations. Notably, our method achieves a speed-up factor of approximately 50 compared to molecular dynamics driven by a machine-learning potential, and is at least 3,000 times faster compared to first-principles molecular dynamics.", "date": "2024-11-11", "categories": []}
{"id": "2505.13571", "title": "Autonomous nanoparticle synthesis by design", "abstract": "Controlled synthesis of materials with specified atomic structures underpins technological advances yet remains reliant on iterative, trial-and-error approaches. Nanoparticles (NPs), whose atomic arrangement dictates their emergent properties, are particularly challenging to synthesise due to numerous tunable parameters. Here, we introduce an autonomous approach explicitly targeting synthesis of atomic-scale structures. Our method autonomously designs synthesis protocols by matching real time experimental total scattering (TS) and pair distribution function (PDF) data to simulated target patterns, without requiring prior synthesis knowledge. We demonstrate this capability at a synchrotron, successfully synthesising two structurally distinct gold NPs: 5 nm decahedral and 10 nm face-centred cubic structures. Ultimately, specifying a simulated target scattering pattern, thus representing a bespoke atomic structure, and obtaining both the synthesised material and its reproducible synthesis protocol on demand may revolutionise materials design. Thus, ScatterLab provides a generalisable blueprint for autonomous, atomic structure-targeted synthesis across diverse systems and applications.", "date": "2025-05-19", "categories": []}
{"id": "2505.14245", "title": "Path-integral molecular dynamics with actively-trained and universal machine learning force fields", "abstract": "Accounting for nuclear quantum effects (NQEs) can significantly alter material properties at finite temperatures. Atomic modeling using the path-integral molecular dynamics (PIMD) method can fully account for such effects, but requires computationally efficient and accurate models of interatomic interactions. Empirical potentials are fast but may lack sufficient accuracy, whereas quantum-mechanical calculations are highly accurate but computationally expensive. Machine-learned interatomic potentials offer a solution to this challenge, providing near-quantum-mechanical accuracy while maintaining high computational efficiency compared to density functional theory (DFT) calculations. In this context, an interface was developed to integrate moment tensor potentials (MTPs) from the MLIP-2 software package into PIMD calculations using the i-PI software package. This interface was then applied to active learning of potentials and to investigate the influence of NQEs on material properties, namely the temperature dependence of lattice parameters and thermal expansion coefficients, as well as radial distribution functions, for lithium hydride (LiH) and silicon (Si) systems. The results were compared with experimental data, quasi-harmonic approximation calculations, and predictions from the universal machine learning force field MatterSim. These comparisons demonstrated the high accuracy and effectiveness of the MTP-PIMD approach.", "date": "2025-05-20", "categories": []}
{"id": "2505.12650", "title": "AutoMat: Enabling Automated Crystal Structure Reconstruction from Microscopy via Agentic Tool Use", "abstract": "Machine learning-based interatomic potentials and force fields depend critically on accurate atomic structures, yet such data are scarce due to the limited availability of experimentally resolved crystals. Although atomic-resolution electron microscopy offers a potential source of structural data, converting these images into simulation-ready formats remains labor-intensive and error-prone, creating a bottleneck for model training and validation. We introduce AutoMat, an end-to-end, agent-assisted pipeline that automatically transforms scanning transmission electron microscopy (STEM) images into atomic crystal structures and predicts their physical properties. AutoMat combines pattern-adaptive denoising, physics-guided template retrieval, symmetry-aware atomic reconstruction, fast relaxation and property prediction via MatterSim, and coordinated orchestration across all stages. We propose the first dedicated STEM2Mat-Bench for this task and evaluate performance using lattice RMSD, formation energy MAE, and structure-matching success rate. By orchestrating external tool calls, AutoMat enables a text-only LLM to outperform vision-language models in this domain, achieving closed-loop reasoning throughout the pipeline. In large-scale experiments over 450 structure samples, AutoMat substantially outperforms existing multimodal large language models and tools. These results validate both AutoMat and STEM2Mat-Bench, marking a key step toward bridging microscopy and atomistic simulation in materials science.The code and dataset are publicly available at https://github.com/yyt-2378/AutoMat and https://huggingface.co/datasets/yaotianvector/STEM2Mat.", "date": "2025-05-19", "categories": []}
{"id": "2403.13952", "title": "Considerations in the use of ML interaction potentials for free energy calculations", "abstract": "Machine learning force fields (MLFFs) promise to accurately describe the potential energy surface of molecules at the ab initio level of theory with improved computational efficiency. Within MLFFs, equivariant graph neural networks (EQNNs) have shown great promise in accuracy and performance and are the focus of this work. The capability of EQNNs to recover free energy surfaces (FES) remains to be thoroughly investigated. In this work, we investigate the impact of collective variables (CVs) distribution within the training data on the accuracy of EQNNs predicting the FES of butane and alanine dipeptide (ADP). A generalizable workflow is presented in which training configurations are generated with classical molecular dynamics simulations, and energies and forces are obtained with ab initio calculations. We evaluate how bond and angle constraints in the training data influence the accuracy of EQNN force fields in reproducing the FES of the molecules at both classical and ab initio levels of theory. Results indicate that the model's accuracy is unaffected by the distribution of sampled CVs during training, given that the training data includes configurations from characteristic regions of the system's FES. However, when the training data is obtained from classical simulations, the EQNN struggles to extrapolate the free energy for configurations with high free energy. In contrast, models trained with the same configurations on ab initio data show improved extrapolation accuracy. The findings underscore the difficulties in creating a comprehensive training dataset for EQNNs to predict FESs and highlight the importance of prior knowledge of the system's FES.", "date": "2024-03-20", "categories": []}
{"id": "2501.12222", "title": "High-temperature superconductivity in Li$_2$AuH$_6$ mediated by strong electron-phonon coupling under ambient pressure", "abstract": "We used our developed AI search engine~(InvDesFlow) to perform extensive investigations regarding ambient stable superconducting hydrides. A cubic structure Li$_2$AuH$_6$ with Au-H octahedral motifs is identified to be a candidate. After performing thermodynamical analysis, we provide a feasible route to experimentally synthesize this material via the known LiAu and LiH compounds under ambient pressure. The further first-principles calculations suggest that Li$_2$AuH$_6$ shows a high superconducting transition temperature ($T_c$) $\\sim$ 140 K under ambient pressure. The H-1$s$ electrons strongly couple with phonon modes of vibrations of Au-H octahedrons as well as vibrations of Li atoms, where the latter is not taken seriously in other previously similar cases. Hence, different from previous claims of searching metallic covalent bonds to find high-$T_c$ superconductors, we emphasize here the importance of those phonon modes with strong electron-phonon coupling (EPC). And we suggest that one can intercalate atoms into binary or ternary hydrides to introduce more potential phonon modes with strong EPC, which is an effective approach to find high-$T_c$ superconductors within multicomponent compounds.", "date": "2025-01-21", "categories": []}
{"id": "2505.09161", "title": "Bridging Theory and Experiment in Materials Discovery: Machine-Learning-Assisted Prediction of Synthesizable Structures", "abstract": "Even though thermodynamic energy-based crystal structure prediction (CSP) has revolutionized materials discovery, the energy-driven CSP approaches often struggle to identify experimentally realizable metastable materials synthesized through kinetically controlled pathways, creating a critical gap between theoretical predictions and experimental synthesis. Here, we propose a synthesizability-driven CSP framework that integrates symmetry-guided structure derivation with a Wyckoff encode-based machine-learning model, allowing for the efficient localization of subspaces likely to yield highly synthesizable structures. Within the identified promising subspaces, a structure-based synthesizability evaluation model, fine-tuned using recently synthesized structures to enhance predictive accuracy, is employed in conjunction with ab initio calculations to systematically identify synthesizable candidates. The framework successfully reproduces 13 experimentally known XSe (X = Sc, Ti, Mn, Fe, Ni, Cu, Zn) structures, demonstrating its effectiveness in predicting synthesizable structures. Notably, 92,310 structures are filtered from the 554,054 candidates predicted by GNoME, exhibiting great potential for promising synthesizability. Additionally, eight thermodynamically favorable Hf-X-O (X = Ti, V, and Mn) structures have been identified, among which three HfV$_2$O$_7$ candidates exhibit high synthesizability, presenting viable candidates for experimental realization and potentially associated with experimentally observed temperature-induced phase transitions. This work establishes a data-driven paradigm for machine-learning-assisted inorganic materials synthesis, highlighting its potential to bridge the gap between computational predictions and experimental realization while unlocking new opportunities for the targeted discovery of novel functional materials.", "date": "2025-05-14", "categories": []}
{"id": "2505.09174", "title": "Quotient Complex Transformer (QCformer) for Perovskite Data Analysis", "abstract": "The discovery of novel functional materials is crucial in addressing the challenges of sustainable energy generation and climate change. Hybrid organic-inorganic perovskites (HOIPs) have gained attention for their exceptional optoelectronic properties in photovoltaics. Recently, geometric deep learning, particularly graph neural networks (GNNs), has shown strong potential in predicting material properties and guiding material design. However, traditional GNNs often struggle to capture the periodic structures and higher-order interactions prevalent in such systems. To address these limitations, we propose a novel representation based on quotient complexes (QCs) and introduce the Quotient Complex Transformer (QCformer) for material property prediction. A material structure is modeled as a quotient complex, which encodes both pairwise and many-body interactions via simplices of varying dimensions and captures material periodicity through a quotient operation. Our model leverages higher-order features defined on simplices and processes them using a simplex-based Transformer module. We pretrain QCformer on benchmark datasets such as the Materials Project and JARVIS, and fine-tune it on HOIP datasets. The results show that QCformer outperforms state-of-the-art models in HOIP property prediction, demonstrating its effectiveness. The quotient complex representation and QCformer model together contribute a powerful new tool for predictive modeling of perovskite materials.", "date": "2025-05-14", "categories": []}
{"id": "2311.04418", "title": "AI-accelerated Discovery of Altermagnetic Materials", "abstract": "Altermagnetism, a new magnetic phase, has been theoretically proposed and experimentally verified to be distinct from ferromagnetism and antiferromagnetism. Although altermagnets have been found to possess many exotic physical properties, the limited availability of known altermagnetic materials hinders the study of such properties. Hence, discovering more types of altermagnetic materials with different properties is crucial for a comprehensive understanding of altermagnetism and thus facilitating new applications in the next generation information technologies, e.g., storage devices and high-sensitivity sensors. Since each altermagnetic material has a unique crystal structure, we propose an automated discovery approach empowered by an AI search engine that employs a pre-trained graph neural network to learn the intrinsic features of the material crystal structure, followed by fine-tuning a classifier with limited positive samples to predict the altermagnetism probability of a given material candidate. Finally, we successfully discovered 50 new altermagnetic materials that cover metals, semiconductors, and insulators confirmed by the first-principles electronic structure calculations. The wide range of electronic structural characteristics reveals that various novel physical properties manifest in these newly discovered altermagnetic materials, e.g., anomalous Hall effect, anomalous Kerr effect, and topological property. Noteworthy, we discovered 4 $i$-wave altermagnetic materials for the first time. Overall, the AI search engine performs much better than human experts and suggests a set of new altermagnetic materials with unique properties, outlining its potential for accelerated discovery of the materials with targeted properties.", "date": "2023-11-08", "categories": []}
{"id": "2409.08065", "title": "InvDesFlow: An AI-driven materials inverse design workflow to explore possible high-temperature superconductors", "abstract": "The discovery of new superconducting materials, particularly those exhibiting high critical temperature ($T_c$), has been a vibrant area of study within the field of condensed matter physics. Conventional approaches primarily rely on physical intuition to search for potential superconductors within the existing databases. However, the known materials only scratch the surface of the extensive array of possibilities within the realm of materials. Here, we develop InvDesFlow, an AI search engine that integrates deep model pre-training and fine-tuning techniques, diffusion models, and physics-based approaches (e.g., first-principles electronic structure calculation) for the discovery of high-$T_c$ superconductors. Utilizing InvDesFlow, we have obtained 74 dynamically stable materials with critical temperatures predicted by the AI model to be $T_c \\geq$ 15 K based on a very small set of samples. Notably, these materials are not contained in any existing dataset. Furthermore, we analyze trends in our dataset and individual materials including B$_4$CN$_3$ (at 5 GPa) and B$_5$CN$_2$ (at ambient pressure) whose $T_c$s are 24.08 K and 15.93 K, respectively. We demonstrate that AI technique can discover a set of new high-$T_c$ superconductors, outline its potential for accelerating discovery of the materials with targeted properties.", "date": "2024-09-12", "categories": []}
{"id": "2505.07906", "title": "Image-Guided Microstructure Optimization using Diffusion Models: Validated with Li-Mn-rich Cathode Precursors", "abstract": "Microstructure often dictates materials performance, yet it is rarely treated as an explicit design variable because microstructure is hard to quantify, predict, and optimize. Here, we introduce an image centric, closed-loop framework that makes microstructural morphology into a controllable objective and demonstrate its use case with Li- and Mn-rich layered oxide cathode precursors. This work presents an integrated, AI driven framework for the predictive design and optimization of lithium-ion battery cathode precursor synthesis. This framework integrates a diffusion-based image generation model, a quantitative image analysis pipeline, and a particle swarm optimization (PSO) algorithm. By extracting key morphological descriptors such as texture, sphericity, and median particle size (D50) from SEM images, the platform accurately predicts SEM like morphologies resulting from specific coprecipitation conditions, including reaction time-, solution concentration-, and pH-dependent structural changes. Optimization then pinpoints synthesis parameters that yield user defined target morphologies, as experimentally validated by the close agreement between predicted and synthesized structures. This framework offers a practical strategy for data driven materials design, enabling both forward prediction and inverse design of synthesis conditions and paving the way toward autonomous, image guided microstructure engineering.", "date": "2025-05-12", "categories": []}
{"id": "2505.07801", "title": "Automatically Differentiable Model Updating (ADiMU): conventional, hybrid, and neural network material model discovery including history-dependency", "abstract": "We introduce the first Automatically Differentiable Model Updating (ADiMU) framework that finds any history-dependent material model from full-field displacement and global force data (global, indirect discovery) or from strain-stress data (local, direct discovery). We show that ADiMU can update conventional (physics-based), neural network (data-driven), and hybrid material models. Moreover, this framework requires no fine-tuning of hyperparameters or additional quantities beyond those inherent to the user-selected material model architecture and optimizer. The robustness and versatility of ADiMU is extensively exemplified by updating different models spanning tens to millions of parameters, in both local and global discovery settings. Relying on fully differentiable code, the algorithmic implementation leverages vectorizing maps that enable history-dependent automatic differentiation via efficient batched execution of shared computation graphs. This contribution also aims to facilitate the integration, evaluation and application of future material model architectures by openly supporting the research community. Therefore, ADiMU is released as an open-source computational tool, integrated into a carefully designed and documented software named HookeAI.", "date": "2025-05-12", "categories": []}
{"id": "2505.04634", "title": "MatMMFuse: Multi-Modal Fusion model for Material Property Prediction", "abstract": "  The recent progress of using graph based encoding of crystal structures for\nhigh throughput material property prediction has been quite successful.\nHowever, using a single modality model prevents us from exploiting the\nadvantages of an enhanced features space by combining different\nrepresentations. Specifically, pre-trained Large language models(LLMs) can\nencode a large amount of knowledge which is beneficial for training of models.\nMoreover, the graph encoder is able to learn the local features while the text\nencoder is able to learn global information such as space group and crystal\nsymmetry. In this work, we propose Material Multi-Modal Fusion(MatMMFuse), a\nfusion based model which uses a multi-head attention mechanism for the\ncombination of structure aware embedding from the Crystal Graph Convolution\nNetwork (CGCNN) and text embeddings from the SciBERT model. We train our model\nin an end-to-end framework using data from the Materials Project Dataset. We\nshow that our proposed model shows an improvement compared to the vanilla CGCNN\nand SciBERT model for all four key properties: formation energy, band gap,\nenergy above hull and fermi energy. Specifically, we observe an improvement of\n40% compared to the vanilla CGCNN model and 68% compared to the SciBERT model\nfor predicting the formation energy per atom. Importantly, we demonstrate the\nzero shot performance of the trained model on small curated datasets of\nPerovskites, Chalcogenides and the Jarvis Dataset. The results show that the\nproposed model exhibits better zero shot performance than the individual plain\nvanilla CGCNN and SciBERT model. This enables researchers to deploy the model\nfor specialized industrial applications where collection of training data is\nprohibitively expensive.\n", "date": "2025-04-30", "categories": []}
{"id": "2505.05354", "title": "High-fidelity Grain Growth Modeling: Leveraging Deep Learning for Fast\n  Computations", "abstract": "  Grain growth simulation is crucial for predicting metallic material\nmicrostructure evolution during annealing and resulting final mechanical\nproperties, but traditional partial differential equation-based methods are\ncomputationally expensive, creating bottlenecks in materials design and\nmanufacturing. In this work, we introduce a machine learning framework that\ncombines a Convolutional Long Short-Term Memory networks with an Autoencoder to\nefficiently predict grain growth evolution. Our approach captures both spatial\nand temporal aspects of grain evolution while encoding high-dimensional grain\nstructure data into a compact latent space for pattern learning, enhanced by a\nnovel composite loss function combining Mean Squared Error, Structural\nSimilarity Index Measurement, and Boundary Preservation to maintain structural\nintegrity of grain boundary topology of the prediction. Results demonstrated\nthat our machine learning approach accelerates grain growth prediction by up to\n\\SI{89}{\\times} faster, reducing computation time from \\SI{10}{\\minute} to\napproximately \\SI{10}{\\second} while maintaining high-fidelity predictions. The\nbest model (S-30-30) achieving a structural similarity score of\n\\SI{86.71}{\\percent} and mean grain size error of just \\SI{0.07}{\\percent}. All\nmodels accurately captured grain boundary topology, morphology, and size\ndistributions. This approach enables rapid microstructural prediction for\napplications where conventional simulations are prohibitively time-consuming,\npotentially accelerating innovation in materials science and manufacturing.\n", "date": "2025-05-08", "categories": []}
{"id": "2505.03779", "title": "Neural Co-Optimization of Structural Topology, Manufacturable Layers,\n  and Path Orientations for Fiber-Reinforced Composites", "abstract": "  We propose a neural network-based computational framework for the\nsimultaneous optimization of structural topology, curved layers, and path\norientations to achieve strong anisotropic strength in fiber-reinforced\nthermoplastic composites while ensuring manufacturability. Our framework\nemploys three implicit neural fields to represent geometric shape, layer\nsequence, and fiber orientation. This enables the direct formulation of both\ndesign and manufacturability objectives - such as anisotropic strength,\nstructural volume, machine motion control, layer curvature, and layer thickness\n- into an integrated and differentiable optimization process. By incorporating\nthese objectives as loss functions, the framework ensures that the resultant\ncomposites exhibit optimized mechanical strength while remaining its\nmanufacturability for filament-based multi-axis 3D printing across diverse\nhardware platforms. Physical experiments demonstrate that the composites\ngenerated by our co-optimization method can achieve an improvement of up to\n33.1% in failure loads compared to composites with sequentially optimized\nstructures and manufacturing sequences.\n", "date": "2025-04-30", "categories": []}
{"id": "2502.00798", "title": "Deep Neural Network for Phonon-Assisted Optical Spectra in\n  Semiconductors", "abstract": "  Ab initio based accurate simulation of phonon-assisted optical spectra of\nsemiconductors at finite temperatures remains a formidable challenge, as it\nrequires large supercells for phonon sampling and computationally expensive\nhigh-accuracy exchange-correlation (XC) functionals. In this work, we present\nan efficient approach that combines deep learning tight-binding and potential\nmodels to address this challenge with ab initio fidelity. By leveraging\nmolecular dynamics for atomic configuration sampling and deep learning-enabled\nrapid Hamiltonian evaluation, our approach enables large-scale simulations of\ntemperature-dependent optical properties using advanced XC functionals (HSE,\nSCAN). Demonstrated on silicon and gallium arsenide across temperature 100-400\nK, the method accurately captures phonon-induced bandgap renormalization and\nindirect/direct absorption processes which are in excellent agreement with\nexperimental findings over five orders of magnitude. This work establishes a\npathway for high-throughput investigation of electron-phonon coupled phenomena\nin complex materials, overcoming traditional computational limitations arising\nfrom large supercell used with computationally expensive XC-functionals.\n", "date": "2025-02-02", "categories": []}
{"id": "2502.05463", "title": "Learning Memory and Material Dependent Constitutive Laws", "abstract": "  We propose and study a neural operator framework for learning memory- and\nmaterial microstructure-dependent constitutive laws for heterogeneous\nmaterials. We work in the two-scale setting where homogenization theory\nprovides a systematic approach to deriving macroscale constitutive laws,\nobviating the need to resolve complex microstructure repeatedly. However, the\nunit cell problems defining these constitutive models are typically not\namenable to explicit evaluation. It is therefore of interest to learn\nconstitutive models from data generated by the unit cell problem. Our proposed\nframework models homogenized constitutive laws with both memory- and\nmicrostructure-dependence through the use of Markovian recurrent and Fourier\nneural operators. The homogenization problem for Kelvin-Voigt viscoelastic\nmaterials is studied to provide firm theoretical foundations for our model. The\ntheoretical properties of the cell problem in this Kelvin-Voigt setting\nmotivate the proposed learning framework; and are also used to prove a\nuniversal approximation theorem for the learned macroscale constitutive model.\nNumerical experiments show that the proposed learning framework accurately\nlearns memory- and microstructure-dependent viscoelastic and\nelasto-viscoplastic constitutive models, beyond the setting of the theory.\nFurthermore, we show that the learned constitutive models can be successfully\ndeployed in macroscale simulation of material deformation for different\nmicrostructures without retraining.\n", "date": "2025-02-08", "categories": []}
{"id": "2408.05215", "title": "Physics-Informed Weakly Supervised Learning for Interatomic Potentials", "abstract": "  Machine learning plays an increasingly important role in computational\nchemistry and materials science, complementing computationally intensive ab\ninitio and first-principles methods. Despite their utility, machine-learning\nmodels often lack generalization capability and robustness during atomistic\nsimulations, yielding unphysical energy and force predictions that hinder their\nreal-world applications. We address this challenge by introducing a\nphysics-informed, weakly supervised approach for training machine-learned\ninteratomic potentials (MLIPs). We introduce two novel loss functions,\nextrapolating the potential energy via a Taylor expansion and using the concept\nof conservative forces. Our approach improves the accuracy of MLIPs applied to\ntraining tasks with sparse training data sets and reduces the need for\npre-training computationally demanding models with large data sets.\nParticularly, we perform extensive experiments demonstrating reduced energy and\nforce errors -- often lower by a factor of two -- for various baseline models\nand benchmark data sets. Moreover, we demonstrate improved robustness during MD\nsimulations of the MLIP models trained with the proposed weakly supervised\nloss. Finally, our approach improves the fine-tuning of foundation models on\nsparse, highly accurate ab initio data. An implementation of our method and\nscripts for executing experiments are available at\nhttps://github.com/nec-research/PICPS-ML4Sci.\n", "date": "2024-07-23", "categories": []}
{"id": "2505.01438", "title": "Global Stress Generation and Spatiotemporal Super-Resolution\n  Physics-Informed Operator under Dynamic Loading for Two-Phase Random\n  Materials", "abstract": "  Material stress analysis is a critical aspect of material design and\nperformance optimization. Under dynamic loading, the global stress evolution in\nmaterials exhibits complex spatiotemporal characteristics, especially in\ntwo-phase random materials (TRMs). Such kind of material failure is often\nassociated with stress concentration, and the phase boundaries are key\nlocations where stress concentration occurs. In practical engineering\napplications, the spatiotemporal resolution of acquired microstructural data\nand its dynamic stress evolution is often limited. This poses challenges for\ndeep learning methods in generating high-resolution spatiotemporal stress\nfields, particularly for accurately capturing stress concentration regions. In\nthis study, we propose a framework for global stress generation and\nspatiotemporal super-resolution in TRMs under dynamic loading. First, we\nintroduce a diffusion model-based approach, named as Spatiotemporal Stress\nDiffusion (STS-diffusion), for generating global spatiotemporal stress data.\nThis framework incorporates Space-Time U-Net (STU-net), and we systematically\ninvestigate the impact of different attention positions on model accuracy.\nNext, we develop a physics-informed network for spatiotemporal\nsuper-resolution, termed as Spatiotemporal Super-Resolution Physics-Informed\nOperator (ST-SRPINN). The proposed ST-SRPINN is an unsupervised learning\nmethod. The influence of data-driven and physics-informed loss function weights\non model accuracy is explored in detail. Benefiting from physics-based\nconstraints, ST-SRPINN requires only low-resolution stress field data during\ntraining and can upscale the spatiotemporal resolution of stress fields to\narbitrary magnifications.\n", "date": "2025-04-26", "categories": []}
{"id": "2504.21331", "title": "Towards Space Group Determination from EBSD Patterns: The Role of Deep\n  Learning and High-throughput Dynamical Simulations", "abstract": "  The design of novel materials hinges on the understanding of\nstructure-property relationships. However, in recent times, our capability to\nsynthesize a large number of materials has outpaced our speed at characterizing\nthem. While the overall chemical constituents can be readily known during\nsynthesis, the structural evolution and characterization of newly synthesized\nsamples remains a bottleneck for the ultimate goal of high throughput\nnanomaterials discovery. Thus, scalable methods for crystal symmetry\ndetermination that can analyze a large volume of material samples within a\nshort time-frame are especially needed. Kikuchi diffraction in the SEM is a\npromising technique for this due to its sensitivity to dynamical scattering,\nwhich may provide information beyond just the seven crystal systems and\nfourteen Bravais lattices. After diffraction patterns are collected from\nmaterial samples, deep learning methods may be able to classify the space group\nsymmetries using the patterns as input, which paired with the elemental\ncomposition, would help enable the determination of the crystal structure. To\ninvestigate the feasibility of this solution, neural networks were trained to\npredict the space group type of background corrected EBSD patterns. Our\nnetworks were first trained and tested on an artificial dataset of EBSD\npatterns of 5,148 different cubic phases, created through physics-based\ndynamical simulations. Next, Maximum Classifier Discrepancy, an unsupervised\ndeep learning-based domain adaptation method, was utilized to train neural\nnetworks to make predictions for experimental EBSD patterns. We introduce a\nrelabeling scheme, which enables our models to achieve accuracy scores higher\nthan 90% on simulated and experimental data, suggesting that neural networks\nare capable of making predictions of crystal symmetry from an EBSD pattern.\n", "date": "2025-04-30", "categories": []}
{"id": "2505.01060", "title": "Monotone Peridynamic Neural Operator for Nonlinear Material Modeling\n  with Conditionally Unique Solutions", "abstract": "  Data-driven methods have emerged as powerful tools for modeling the responses\nof complex nonlinear materials directly from experimental measurements. Among\nthese methods, the data-driven constitutive models present advantages in\nphysical interpretability and generalizability across different boundary\nconditions/domain settings. However, the well-posedness of these learned models\nis generally not guaranteed a priori, which makes the models prone to\nnon-physical solutions in downstream simulation tasks. In this study, we\nintroduce monotone peridynamic neural operator (MPNO), a novel data-driven\nnonlocal constitutive model learning approach based on neural operators. Our\napproach learns a nonlocal kernel together with a nonlinear constitutive\nrelation, while ensuring solution uniqueness through a monotone gradient\nnetwork. This architectural constraint on gradient induces convexity of the\nlearnt energy density function, thereby guaranteeing solution uniqueness of\nMPNO in small deformation regimes. To validate our approach, we evaluate MPNO's\nperformance on both synthetic and real-world datasets. On synthetic datasets\nwith manufactured kernel and constitutive relation, we show that the learnt\nmodel converges to the ground-truth as the measurement grid size decreases both\ntheoretically and numerically. Additionally, our MPNO exhibits superior\ngeneralization capabilities than the conventional neural networks: it yields\nsmaller displacement solution errors in down-stream tasks with new and unseen\nloadings. Finally, we showcase the practical utility of our approach through\napplications in learning a homogenized model from molecular dynamics data,\nhighlighting its expressivity and robustness in real-world scenarios.\n", "date": "2025-05-02", "categories": []}
{"id": "2505.01283", "title": "Reduced-order structure-property linkages for stochastic metamaterials", "abstract": "  The capabilities of additive manufacturing have facilitated the design and\nproduction of mechanical metamaterials with diverse unit cell geometries.\nEstablishing linkages between the vast design space of unit cells and their\neffective mechanical properties is critical for the efficient design and\nperformance evaluation of such metamaterials. However, physics-based\nsimulations of metamaterial unit cells across the entire design space are\ncomputationally expensive, necessitating a materials informatics framework to\nefficiently capture complex structure-property relationships. In this work,\nprincipal component analysis of 2-point correlation functions is performed to\nextract the salient features from a large dataset of randomly generated 2D\nmetamaterials. Physics-based simulations are performed using a fast Fourier\ntransform (FFT)-based homogenization approach to efficiently compute the\nhomogenized effective elastic stiffness across the extensive unit cell designs.\nSubsequently, Gaussian process regression is used to generate reduced-order\nsurrogates, mapping unit cell designs to their homogenized effective elastic\nconstant. It is demonstrated that the adopted workflow enables a high-value\nlow-dimensional representation of the voluminous stochastic metamaterial\ndataset, facilitating the construction of robust structure-property maps.\nFinally, an uncertainty-based active learning framework is utilized to train a\nsurrogate model with a significantly smaller number of data points compared to\nthe original full dataset. It is shown that a dataset as small as $0.61\\%$ of\nthe entire dataset is sufficient to generate accurate and robust\nstructure-property maps.\n", "date": "2025-05-02", "categories": []}
{"id": "2404.12367", "title": "Model-free quantification of completeness, uncertainties, and outliers\n  in atomistic machine learning using information theory", "abstract": "  An accurate description of information is relevant for a range of problems in\natomistic machine learning (ML), such as crafting training sets, performing\nuncertainty quantification (UQ), or extracting physical insights from large\ndatasets. However, atomistic ML often relies on unsupervised learning or model\npredictions to analyze information contents from simulation or training data.\nHere, we introduce a theoretical framework that provides a rigorous, model-free\ntool to quantify information contents in atomistic simulations. We demonstrate\nthat the information entropy of a distribution of atom-centered environments\nexplains known heuristics in ML potential developments, from training set sizes\nto dataset optimality. Using this tool, we propose a model-free UQ method that\nreliably predicts epistemic uncertainty and detects out-of-distribution\nsamples, including rare events in systems such as nucleation. This method\nprovides a general tool for data-driven atomistic modeling and combines efforts\nin ML, simulations, and physical explainability.\n", "date": "2024-04-18", "categories": []}
{"id": "2504.10560", "title": "Molecular Learning Dynamics", "abstract": "  We apply the physics-learning duality to molecular systems by complementing\nthe physical description of interacting particles with a dual learning\ndescription, where each particle is modeled as an agent minimizing a loss\nfunction. In the traditional physics framework, the equations of motion are\nderived from the Lagrangian function, while in the learning framework, the same\nequations emerge from learning dynamics driven by the agent loss function. The\nloss function depends on scalar quantities that describe invariant properties\nof all other agents or particles. To demonstrate this approach, we first infer\nthe loss functions of oxygen and hydrogen directly from a dataset generated by\nthe CP2K physics-based simulation of water molecules. We then employ the loss\nfunctions to develop a learning-based simulation of water molecules, which\nachieves comparable accuracy while being significantly more computationally\nefficient than standard physics-based simulations.\n", "date": "2025-04-14", "categories": []}
{"id": "2504.20112", "title": "Supervised Pretraining for Material Property Prediction", "abstract": "  Accurate prediction of material properties facilitates the discovery of novel\nmaterials with tailored functionalities. Deep learning models have recently\nshown superior accuracy and flexibility in capturing structure-property\nrelationships. However, these models often rely on supervised learning, which\nrequires large, well-annotated datasets an expensive and time-consuming\nprocess. Self-supervised learning (SSL) offers a promising alternative by\npretraining on large, unlabeled datasets to develop foundation models that can\nbe fine-tuned for material property prediction. In this work, we propose\nsupervised pretraining, where available class information serves as surrogate\nlabels to guide learning, even when downstream tasks involve unrelated material\nproperties. We evaluate this strategy on two state-of-the-art SSL models and\nintroduce a novel framework for supervised pretraining. To further enhance\nrepresentation learning, we propose a graph-based augmentation technique that\ninjects noise to improve robustness without structurally deforming material\ngraphs. The resulting foundation models are fine-tuned for six challenging\nmaterial property predictions, achieving significant performance gains over\nbaselines, ranging from 2% to 6.67% improvement in mean absolute error (MAE)\nand establishing a new benchmark in material property prediction. This study\nrepresents the first exploration of supervised pertaining with surrogate labels\nin material property prediction, advancing methodology and application in the\nfield.\n", "date": "2025-04-27", "categories": []}
{"id": "2502.13107", "title": "MatterChat: A Multi-Modal LLM for Material Science", "abstract": "  Understanding and predicting the properties of inorganic materials is crucial\nfor accelerating advancements in materials science and driving applications in\nenergy, electronics, and beyond. Integrating material structure data with\nlanguage-based information through multi-modal large language models (LLMs)\noffers great potential to support these efforts by enhancing human-AI\ninteraction. However, a key challenge lies in integrating atomic structures at\nfull resolution into LLMs. In this work, we introduce MatterChat, a versatile\nstructure-aware multi-modal LLM that unifies material structural data and\ntextual inputs into a single cohesive model. MatterChat employs a bridging\nmodule to effectively align a pretrained machine learning interatomic potential\nwith a pretrained LLM, reducing training costs and enhancing flexibility. Our\nresults demonstrate that MatterChat significantly improves performance in\nmaterial property prediction and human-AI interaction, surpassing\ngeneral-purpose LLMs such as GPT-4. We also demonstrate its usefulness in\napplications such as more advanced scientific reasoning and step-by-step\nmaterial synthesis.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.15764", "title": "High-Throughput Computational Screening and Interpretable Machine\n  Learning of Metal-organic Frameworks for Iodine Capture", "abstract": "  The removal of leaked radioactive iodine isotopes in humid environments holds\nsignificant importance in nuclear waste management and nuclear accident\nmitigation. In this study, high-throughput computational screening and machine\nlearning were combined to reveal the iodine capture performance of 1816\nmetal-organic framework (MOF) materials under humid air conditions. Firstly,\nthe relationship between the structural characteristics of MOFs and their\nadsorption properties was explored, with the aim of identifying the optimal\nstructural parameters for iodine capture. Subsequently, two machine learning\nregression algorithms - Random Forest and CatBoost, were employed to predict\nthe iodine adsorption capabilities of MOFs. In addition to 6 structural\nfeatures, 25 molecular features and 8 chemical features were incorporated to\nenhance the prediction accuracy of the machine learning algorithms. Feature\nimportance was assessed to determine the relative influence of various features\non iodine adsorption performance, in which the Henry's coefficient and heat of\nadsorption to iodine were found the two most crucial chemical factors.\nFurthermore, four types of molecular fingerprints were introduced for providing\ncomprehensive and detailed structural information of MOF materials. The top 20\nmost significant MACCS molecular fingerprints were picked out, revealing that\nthe presence of six-membered ring structures and nitrogen atoms in the MOFs\nwere the key structural factors that enhanced iodine adsorption, followed by\nthe existence of oxygen atoms. This work combined high-throughput computation,\nmachine learning, and molecular fingerprints to comprehensively elucidate the\nmultifaceted factors influencing the iodine adsorption performance of MOFs,\noffering profound insightful guidelines for screening and structural design of\nadvanced MOF materials.\n", "date": "2025-02-14", "categories": []}
{"id": "2504.19787", "title": "Interpretable machine learning-guided design of Fe-based soft magnetic\n  alloys", "abstract": "  We present a machine-learning guided approach to predict saturation\nmagnetization (MS) and coercivity (HC) in Fe-rich soft magnetic alloys,\nparticularly Fe-Si-B systems. ML models trained on experimental data reveals\nthat increasing Si and B content reduces MS from 1.81T (DFT~2.04 T) to ~1.54 T\n(DFT~1.56T) in Fe-Si-B, which is attributed to decreased magnetic density and\nstructural modifications. Experimental validation of ML predicted magnetic\nsaturation on Fe-1Si-1B (2.09T), Fe-5Si-5B (2.01T) and Fe-10Si-10B (1.54T)\nalloy compositions further support our findings. These trends are consistent\nwith density functional theory (DFT) predictions, which link increased\nelectronic disorder and band broadening to lower MS values. Experimental\nvalidation on selected alloys confirms the predictive accuracy of the ML model,\nwith good agreement across compositions. Beyond predictive accuracy, detailed\nuncertainty quantification and model interpretability including through feature\nimportance and partial dependence analysis reveals that MS is governed by a\nnonlinear interplay between Fe content, early transition metal ratios, and\nannealing temperature, while HC is more sensitive to processing conditions such\nas ribbon thickness and thermal treatment windows. The ML framework was further\napplied to Fe-Si-B/Cr/Cu/Zr/Nb alloys in a pseudo-quaternary compositional\nspace, which shows comparable magnetic properties to NANOMET\n(Fe84.8Si0.5B9.4Cu0.8 P3.5C1), FINEMET (Fe73.5Si13.5B9 Cu1Nb3), NANOPERM\n(Fe88Zr7B4Cu1), and HITPERM (Fe44Co44Zr7B4Cu1. Our fundings demonstrate the\npotential of ML framework for accelerated search of high-performance, Co- and\nNi-free, soft magnetic materials.\n", "date": "2025-04-28", "categories": []}
{"id": "2504.19987", "title": "Graph Neural Network Prediction of Nonlinear Optical Properties", "abstract": "  Nonlinear optical (NLO) materials for generating lasers via second harmonic\ngeneration (SHG) are highly sought in today's technology. However, discovering\nnovel materials with considerable SHG is challenging due to the time-consuming\nand costly nature of both experimental methods and first-principles\ncalculations. In this study, we present a deep learning approach using the\nAtomistic Line Graph Neural Network (ALIGNN) to predict NLO properties.\nSourcing data from the Novel Opto-Electronic Materials Discovery (NOEMD)\ndatabase and using the Kurtz-Perry (KP) coefficient as the key target, we\ndeveloped a robust model capable of accurately estimating nonlinear optical\nresponses. Our results demonstrate that the model achieves 82.5% accuracy at a\ntolerated absolute error up to 1 pm/V and relative error not exceeding 0.5.\nThis work highlights the potential of deep learning in accelerating the\ndiscovery and design of advanced optical materials with desired properties.\n", "date": "2025-04-28", "categories": []}
{"id": "2504.04002", "title": "Machine Learning Reveals Composition Dependent Thermal Stability in\n  Halide Perovskites", "abstract": "  Halide perovskites exhibit unpredictable properties in response to\nenvironmental stressors, due to several composition-dependent degradation\nmechanisms. In this work, we apply data visualization and machine learning (ML)\ntechniques to reveal unexpected correlations between composition, temperature,\nand material properties while using high throughput, in situ environmental\nphotoluminescence (PL) experiments. Correlation heatmaps show the strong\ninfluence of Cs content on film degradation, and dimensionality reduction\nvisualization methods uncover clear composition-based data clusters. An extreme\ngradient boosting algorithm (XGBoost) effectively forecasts PL features for ten\nperovskite films with both composition-agnostic (>85% accuracy) and\ncomposition-dependent (>75% accuracy) model approaches, while elucidating the\nrelative feature importance of composition (up to 99%). This model validates a\npreviously unseen anti-correlation between Cs content and material thermal\nstability. Our ML-based framework can be expanded to any perovskite family,\nsignificantly reducing the analysis time currently employed to identify stable\noptions for photovoltaics.\n", "date": "2025-04-05", "categories": []}
{"id": "2502.12147", "title": "Learning Smooth and Expressive Interatomic Potentials for Physical\n  Property Prediction", "abstract": "  Machine learning interatomic potentials (MLIPs) have become increasingly\neffective at approximating quantum mechanical calculations at a fraction of the\ncomputational cost. However, lower errors on held out test sets do not always\ntranslate to improved results on downstream physical property prediction tasks.\nIn this paper, we propose testing MLIPs on their practical ability to conserve\nenergy during molecular dynamic simulations. If passed, improved correlations\nare found between test errors and their performance on physical property\nprediction tasks. We identify choices which may lead to models failing this\ntest, and use these observations to improve upon highly-expressive models. The\nresulting model, eSEN, provides state-of-the-art results on a range of physical\nproperty prediction tasks, including materials stability prediction, thermal\nconductivity prediction, and phonon calculations.\n", "date": "2025-02-17", "categories": []}
{"id": "2504.14143", "title": "Predicting Stress and Damage in Carbon Fiber-Reinforced Composites\n  Deformation Process using Composite U-Net Surrogate Model", "abstract": "  Carbon fiber-reinforced composites (CFRC) are pivotal in advanced engineering\napplications due to their exceptional mechanical properties. A deep\nunderstanding of CFRC behavior under mechanical loading is essential for\noptimizing performance in demanding applications such as aerospace structures.\nWhile traditional Finite Element Method (FEM) simulations, including advanced\ntechniques like Interface-enriched Generalized FEM (IGFEM), offer valuable\ninsights, they can struggle with computational efficiency. Existing data-driven\nsurrogate models partially address these challenges by predicting propagated\ndamage or stress-strain behavior but fail to comprehensively capture the\nevolution of stress and damage throughout the entire deformation history,\nincluding crack initiation and propagation. This study proposes a novel\nauto-regressive composite U-Net deep learning model to simultaneously predict\nstress and damage fields during CFRC deformation. By leveraging the U-Net\narchitecture's ability to capture spatial features and integrate macro- and\nmicro-scale phenomena, the proposed model overcomes key limitations of prior\napproaches. The model achieves high accuracy in predicting evolution of stress\nand damage distribution within the microstructure of a CFRC under\nunidirectional strain, offering a speed-up of over 60 times compared to IGFEM.\n", "date": "2025-04-19", "categories": []}
{"id": "2504.14378", "title": "Machine learning enhanced atom probe tomography analysis: a snapshot\n  review", "abstract": "  Atom probe tomography (APT) is a burgeoning characterization technique that\nprovides compositional mapping of materials in three-dimensions at near-atomic\nscale. Since its significant expansion in the past 30 years, we estimate that\none million APT datasets have been collected, each containing millions to\nbillions of individual ions. Their analysis and the extraction of\nmicrostructural information has largely relied upon individual users whose\nvaried level of expertise causes clear and documented bias. Current practices\nhinder efficient data processing, and make challenging standardization and the\ndeployment of data analysis workflows that would be compliant with FAIR data\nprinciples. Over the past decade, building upon the long-standing expertise of\nthe APT community in the development of advanced data processing or data mining\ntechniques, there has been a surge of novel machine learning (ML) approaches\naiming for user-independence, and that are efficient, reproducible, and robust\nfrom a statistics perspective. Here, we provide a snapshot review of this\nrapidly evolving field. We begin with a brief introduction to APT and the\nnature of the APT data. This is followed by an overview of relevant ML\nalgorithms and a comprehensive review of their applications to APT. We also\ndiscuss how ML can enable discoveries beyond human capability, offering new\ninsights into the mechanisms within materials. Finally, we provide guidance for\nfuture directions in this domain.\n", "date": "2025-04-19", "categories": []}
{"id": "2504.12627", "title": "Uncertainty Quantification in Graph Neural Networks with Shallow\n  Ensembles", "abstract": "  Machine-learned potentials (MLPs) have revolutionized materials discovery by\nproviding accurate and efficient predictions of molecular and material\nproperties. Graph Neural Networks (GNNs) have emerged as a state-of-the-art\napproach due to their ability to capture complex atomic interactions. However,\nGNNs often produce unreliable predictions when encountering out-of-domain data\nand it is difficult to identify when that happens. To address this challenge,\nwe explore Uncertainty Quantification (UQ) techniques, focusing on Direct\nPropagation of Shallow Ensembles (DPOSE) as a computationally efficient\nalternative to deep ensembles. By integrating DPOSE into the SchNet model, we\nassess its ability to provide reliable uncertainty estimates across diverse\nDensity Functional Theory datasets, including QM9, OC20, and Gold Molecular\nDynamics. Our findings often demonstrate that DPOSE successfully distinguishes\nbetween in-domain and out-of-domain samples, exhibiting higher uncertainty for\nunobserved molecule and material classes. This work highlights the potential of\nlightweight UQ methods in improving the robustness of GNN-based materials\nmodeling and lays the foundation for future integration with active learning\nstrategies.\n", "date": "2025-04-17", "categories": []}
{"id": "2409.19552", "title": "OmniXAS: A Universal Deep-Learning Framework for Materials X-ray\n  Absorption Spectra", "abstract": "  X-ray absorption spectroscopy (XAS) is a powerful characterization technique\nfor probing the local chemical environment of absorbing atoms. However,\nanalyzing XAS data presents significant challenges, often requiring extensive,\ncomputationally intensive simulations, as well as significant domain expertise.\nThese limitations hinder the development of fast, robust XAS analysis pipelines\nthat are essential in high-throughput studies and for autonomous\nexperimentation. We address these challenges with OmniXAS, a framework that\ncontains a suite of transfer learning approaches for XAS prediction, each\ncontributing to improved accuracy and efficiency, as demonstrated on K-edge\nspectra database covering eight 3d transition metals (Ti-Cu). The OmniXAS\nframework is built upon three distinct strategies. First, we use M3GNet to\nderive latent representations of the local chemical environment of absorption\nsites as input for XAS prediction, achieving up to order-of-magnitude\nimprovements over conventional featurization techniques. Second, we employ a\nhierarchical transfer learning strategy, training a universal multi-task model\nacross elements before fine-tuning for element-specific predictions. Models\nbased on this cascaded approach after element-wise fine-tuning outperform\nelement-specific models by up to 69%. Third, we implement cross-fidelity\ntransfer learning, adapting a universal model to predict spectra generated by\nsimulation of a different fidelity with a higher computational cost. This\napproach improves prediction accuracy by up to 11% over models trained on the\ntarget fidelity alone. Our approach boosts the throughput of XAS modeling by\norders of magnitude versus first-principles simulations and is extendable to\nXAS prediction for a broader range of elements. This transfer learning\nframework is generalizable to enhance deep-learning models that target other\nproperties in materials research.\n", "date": "2024-09-29", "categories": []}
{"id": "2504.10655", "title": "MatterTune: An Integrated, User-Friendly Platform for Fine-Tuning\n  Atomistic Foundation Models to Accelerate Materials Simulation and Discovery", "abstract": "  Geometric machine learning models such as graph neural networks have achieved\nremarkable success in recent years in chemical and materials science research\nfor applications such as high-throughput virtual screening and atomistic\nsimulations. The success of these models can be attributed to their ability to\neffectively learn latent representations of atomic structures directly from the\ntraining data. Conversely, this also results in high data requirements for\nthese models, hindering their application to problems which are data sparse\nwhich are common in this domain. To address this limitation, there is a growing\ndevelopment in the area of pre-trained machine learning models which have\nlearned general, fundamental, geometric relationships in atomistic data, and\nwhich can then be fine-tuned to much smaller application-specific datasets. In\nparticular, models which are pre-trained on diverse, large-scale atomistic\ndatasets have shown impressive generalizability and flexibility to downstream\napplications, and are increasingly referred to as atomistic foundation models.\nTo leverage the untapped potential of these foundation models, we introduce\nMatterTune, a modular and extensible framework that provides advanced\nfine-tuning capabilities and seamless integration of atomistic foundation\nmodels into downstream materials informatics and simulation workflows, thereby\nlowering the barriers to adoption and facilitating diverse applications in\nmaterials science. In its current state, MatterTune supports a number of\nstate-of-the-art foundation models such as ORB, MatterSim, JMP, and\nEquformerV2, and hosts a wide range of features including a modular and\nflexible design, distributed and customizable fine-tuning, broad support for\ndownstream informatics tasks, and more.\n", "date": "2025-04-14", "categories": []}
{"id": "2504.08810", "title": "PriM: Principle-Inspired Material Discovery through Multi-Agent\n  Collaboration", "abstract": "  Complex chemical space and limited knowledge scope with biases holds immense\nchallenge for human scientists, yet in automated materials discovery. Existing\nintelligent methods relies more on numerical computation, leading to\ninefficient exploration and results with hard-interpretability. To bridge this\ngap, we introduce a principles-guided material discovery system powered by\nlanguage inferential multi-agent system (MAS), namely PriM. Our framework\nintegrates automated hypothesis generation with experimental validation in a\nroundtable system of MAS, enabling systematic exploration while maintaining\nscientific rigor. Based on our framework, the case study of nano helix\ndemonstrates higher materials exploration rate and property value while\nproviding transparent reasoning pathways. This approach develops an\nautomated-and-transparent paradigm for material discovery, with broad\nimplications for rational design of functional materials. Code is publicly\navailable at our \\href{https://github.com/amair-lab/PriM}{GitHub}.\n", "date": "2025-04-09", "categories": []}
{"id": "2504.09152", "title": "MatWheel: Addressing Data Scarcity in Materials Science Through\n  Synthetic Data", "abstract": "  Data scarcity and the high cost of annotation have long been persistent\nchallenges in the field of materials science. Inspired by its potential in\nother fields like computer vision, we propose the MatWheel framework, which\ntrain the material property prediction model using the synthetic data generated\nby the conditional generative model. We explore two scenarios: fully-supervised\nand semi-supervised learning. Using CGCNN for property prediction and Con-CDVAE\nas the conditional generative model, experiments on two data-scarce material\nproperty datasets from Matminer database are conducted. Results show that\nsynthetic data has potential in extreme data-scarce scenarios, achieving\nperformance close to or exceeding that of real samples in all two tasks. We\nalso find that pseudo-labels have little impact on generated data quality.\nFuture work will integrate advanced models and optimize generation conditions\nto boost the effectiveness of the materials data flywheel.\n", "date": "2025-04-12", "categories": []}
{"id": "2504.08258", "title": "Accelerating Multi-Objective Collaborative Optimization of Doped\n  Thermoelectric Materials via Artificial Intelligence", "abstract": "  The thermoelectric performance of materials exhibits complex nonlinear\ndependencies on both elemental types and their proportions, rendering\ntraditional trial-and-error approaches inefficient and time-consuming for\nmaterial discovery. In this work, we present a deep learning model capable of\naccurately predicting thermoelectric properties of doped materials directly\nfrom their chemical formulas, achieving state-of-the-art performance. To\nenhance interpretability, we further incorporate sensitivity analysis\ntechniques to elucidate how physical descriptors affect the thermoelectric\nfigure of merit (zT). Moreover, we establish a coupled framework that\nintegrates a surrogate model with a multi-objective genetic algorithm to\nefficiently explore the vast compositional space for high-performance\ncandidates. Experimental validation confirms the discovery of a novel\nthermoelectric material with superior $zT$ values in the medium-temperature\nregime.\n", "date": "2025-04-11", "categories": []}
{"id": "2501.10481", "title": "Learning Latent Hardening (LLH): Enhancing Deep Learning with Domain\n  Knowledge for Material Inverse Problems", "abstract": "  Advancements in deep learning and machine learning have improved the ability\nto model complex, nonlinear relationships, such as those encountered in complex\nmaterial inverse problems. However, the effectiveness of these methods often\ndepends on large datasets, which are not always available. In this study, the\nincorporation of domain-specific knowledge of the mechanical behavior of\nmaterial microstructures is investigated to evaluate the impact on the\npredictive performance of the models in data-scarce scenarios. To overcome data\nlimitations, a two-step framework, Learning Latent Hardening (LLH), is\nproposed. In the first step of LLH, a Deep Neural Network is employed to\nreconstruct full stress-strain curves from randomly selected portions of the\nstress-strain curves to capture the latent mechanical response of a material\nbased on key microstructural features. In the second step of LLH, the results\nof the reconstructed stress-strain curves are leveraged to predict key\nmicrostructural features of porous materials. The performance of six deep\nlearning and/or machine learning models trained with and without domain\nknowledge are compared: Convolutional Neural Networks, Deep Neural Networks,\nExtreme Gradient Boosting, K-Nearest Neighbors, Long Short-Term Memory, and\nRandom Forest. The results from the models with domain-specific information\nconsistently achieved higher $R^2$ values compared to models without prior\nknowledge. Models without domain knowledge missed critical patterns linking\nstress-strain behavior to microstructural changes, whereas domain-informed\nmodels better identified essential stress-strain features predictive of\nmicrostructure. These findings highlight the importance of integrating\ndomain-specific knowledge with deep learning to achieve accurate outcomes in\nmaterials science.\n", "date": "2025-01-17", "categories": []}
{"id": "2504.06878", "title": "CRYSIM: Prediction of Symmetric Structures of Large Crystals with\n  GPU-based Ising Machines", "abstract": "  Solving black-box optimization problems with Ising machines is increasingly\ncommon in materials science. However, their application to crystal structure\nprediction (CSP) is still ineffective due to symmetry agnostic encoding of\natomic coordinates. We introduce CRYSIM, an algorithm that encodes the space\ngroup, the Wyckoff positions combination, and coordinates of independent atomic\nsites as separate variables. This encoding reduces the search space\nsubstantially by exploiting the symmetry in space groups. When CRYSIM is\ninterfaced to Fixstars Amplify, a GPU-based Ising machine, its prediction\nperformance was competitive with CALYPSO and Bayesian optimization for crystals\ncontaining more than 150 atoms in a unit cell. Although it is not realistic to\ninterface CRYSIM to current small-scale quantum devices, it has the potential\nto become the standard CSP algorithm in the coming quantum age.\n", "date": "2025-04-09", "categories": []}
{"id": "2503.07378", "title": "A Materials Map Integrating Experimental and Computational Data via\n  Graph-Based Machine Learning for Enhanced Materials Discovery", "abstract": "  Materials informatics (MI), emerging from the integration of materials\nscience and data science, is expected to significantly accelerate material\ndevelopment and discovery. The data used in MI are derived from both\ncomputational and experimental studies; however, their integration remains\nchallenging. In our previous study, we reported the integration of these\ndatasets by applying a machine learning model that is trained on the\nexperimental dataset to the compositional data stored in the computational\ndatabase. In this study, we use the obtained datasets to construct materials\nmaps, which visualize the relationships between material properties and\nstructural features, aiming to support experimental researchers. The materials\nmap is constructed using the MatDeepLearn (MDL) framework, which implements\nmaterials property prediction using graph-based representations of material\nstructure and deep learning modeling. Through statistical analysis, we find\nthat the MDL framework using the message passing neural network (MPNN)\narchitecture efficiently extracts features reflecting the structural complexity\nof materials. Moreover, we find that this advantage does not necessarily\ntranslate into improved accuracy in the prediction of material properties. We\nattribute this unexpected outcome to the high learning performance inherent in\nMPNN, which can contribute to the structuring of data points within the\nmaterials map.\n", "date": "2025-03-10", "categories": []}
{"id": "2504.05565", "title": "Cross-functional transferability in universal machine learning\n  interatomic potentials", "abstract": "  The rapid development of universal machine learning interatomic potentials\n(uMLIPs) has demonstrated the possibility for generalizable learning of the\nuniversal potential energy surface. In principle, the accuracy of uMLIPs can be\nfurther improved by bridging the model from lower-fidelity datasets to\nhigh-fidelity ones. In this work, we analyze the challenge of this transfer\nlearning problem within the CHGNet framework. We show that significant energy\nscale shifts and poor correlations between GGA and r$^2$SCAN pose challenges to\ncross-functional data transferability in uMLIPs. By benchmarking different\ntransfer learning approaches on the MP-r$^2$SCAN dataset of 0.24 million\nstructures, we demonstrate the importance of elemental energy referencing in\nthe transfer learning of uMLIPs. By comparing the scaling law with and without\nthe pre-training on a low-fidelity dataset, we show that significant data\nefficiency can still be achieved through transfer learning, even with a target\ndataset of sub-million structures. We highlight the importance of proper\ntransfer learning and multi-fidelity learning in creating next-generation\nuMLIPs on high-fidelity data.\n", "date": "2025-04-07", "categories": []}
{"id": "2504.02848", "title": "Transfer learning from first-principles calculations to experiments with\n  chemistry-informed domain transformation", "abstract": "  Simulation-to-Real (Sim2Real) transfer learning, the machine learning\ntechnique that efficiently solves a real-world task by leveraging knowledge\nfrom computational data, has received increasing attention in materials science\nas a promising solution to the scarcity of experimental data. We proposed an\nefficient transfer learning scheme from first-principles calculations to\nexperiments based on the chemistry-informed domain transformation, that\nintegrates the heterogeneous source and target domains by harnessing the\nunderlying physics and chemistry. The proposed method maps the computational\ndata from the simulation space (source domain) into the space of experimental\ndata (target domain). During this process, these qualitatively different\ndomains are efficiently integrated by a couple of prior knowledge of chemistry,\n(1) the statistical ensemble, and (2) the relationship between source and\ntarget quantities. As a proof-of-concept, we predict the catalyst activity for\nthe reverse water-gas shift reaction by using the abundant first-principles\ndata in addition to the experimental data. Through the demonstration, we\nconfirmed that the transfer learning model exhibits positive transfer in\naccuracy and data efficiency. In particular, a significantly high accuracy was\nachieved despite using a few (less than ten) target data in domain\ntransformation, whose accuracy is one order of magnitude smaller than that of a\nfull scratch model trained with over 100 target data. This result indicates\nthat the proposed method leverages the high prediction performance with few\ntarget data, which helps to save the number of trials in real laboratories.\n", "date": "2025-03-20", "categories": []}
{"id": "2504.05169", "title": "Machine learning interatomic potential can infer electrical response", "abstract": "  Modeling the response of material and chemical systems to electric fields\nremains a longstanding challenge. Machine learning interatomic potentials\n(MLIPs) offer an efficient and scalable alternative to quantum mechanical\nmethods but do not by themselves incorporate electrical response. Here, we show\nthat polarization and Born effective charge (BEC) tensors can be directly\nextracted from long-range MLIPs within the Latent Ewald Summation (LES)\nframework, solely by learning from energy and force data. Using this approach,\nwe predict the infrared spectra of bulk water under zero or finite external\nelectric fields, ionic conductivities of high-pressure superionic ice, and the\nphase transition and hysteresis in ferroelectric PbTiO$_3$ perovskite. This\nwork thus extends the capability of MLIPs to predict electrical\nresponse--without training on charges or polarization or BECs--and enables\naccurate modeling of electric-field-driven processes in diverse systems at\nscale.\n", "date": "2025-04-07", "categories": []}
{"id": "2504.03187", "title": "On the Connection Between Diffusion Models and Molecular Dynamics", "abstract": "  Neural Network Potentials (NNPs) have emerged as a powerful tool for\nmodelling atomic interactions with high accuracy and computational efficiency.\nRecently, denoising diffusion models have shown promise in NNPs by training\nnetworks to remove noise added to stable configurations, eliminating the need\nfor force data during training. In this work, we explore the connection between\nnoise and forces by providing a new, simplified mathematical derivation of\ntheir relationship. We also demonstrate how a denoising model can be\nimplemented using a conventional MD software package interfaced with a standard\nNNP architecture. We demonstrate the approach by training a diffusion-based NNP\nto simulate a coarse-grained lithium chloride solution and employ data\nduplication to enhance model performance.\n", "date": "2025-04-04", "categories": []}
{"id": "2408.16337", "title": "Do Graph Neural Networks Work for High Entropy Alloys?", "abstract": "  Graph neural networks (GNNs) have excelled in predictive modeling for both\ncrystals and molecules, owing to the expressiveness of graph representations.\nHigh-entropy alloys (HEAs), however, lack chemical long-range order, limiting\nthe applicability of current graph representations. To overcome this challenge,\nwe propose a representation of HEAs as a collection of local environment (LE)\ngraphs. Based on this representation, we introduce the LESets machine learning\nmodel, an accurate, interpretable GNN for HEA property prediction. We\ndemonstrate the accuracy of LESets in modeling the mechanical properties of\nquaternary HEAs. Through analyses and interpretation, we further extract\ninsights into the modeling and design of HEAs. In a broader sense, LESets\nextends the potential applicability of GNNs to disordered materials with\ncombinatorial complexity formed by diverse constituents and their flexible\nconfigurations.\n", "date": "2024-08-29", "categories": []}
{"id": "2503.23515", "title": "Optimal Invariant Bases for Atomistic Machine Learning", "abstract": "  The representation of atomic configurations for machine learning models has\nled to the development of numerous descriptors, often to describe the local\nenvironment of atoms. However, many of these representations are incomplete\nand/or functionally dependent. Incomplete descriptor sets are unable to\nrepresent all meaningful changes in the atomic environment. Complete\nconstructions of atomic environment descriptors, on the other hand, often\nsuffer from a high degree of functional dependence, where some descriptors can\nbe written as functions of the others. These redundant descriptors do not\nprovide additional power to discriminate between different atomic environments\nand increase the computational burden. By employing techniques from the pattern\nrecognition literature to existing atomistic representations, we remove\ndescriptors that are functions of other descriptors to produce the smallest\npossible set that satisfies completeness. We apply this in two ways: first we\nrefine an existing description, the Atomistic Cluster Expansion. We show that\nthis yields a more efficient subset of descriptors. Second, we augment an\nincomplete construction based on a scalar neural network, yielding a new\nmessage-passing network architecture that can recognize up to 5-body patterns\nin each neuron by taking advantage of an optimal set of Cartesian tensor\ninvariants. This architecture shows strong accuracy on state-of-the-art\nbenchmarks while retaining low computational cost. Our results not only yield\nimproved models, but point the way to classes of invariant bases that minimize\ncost while maximizing expressivity for a host of applications.\n", "date": "2025-03-30", "categories": []}
{"id": "2404.07181", "title": "A predictive machine learning force field framework for liquid\n  electrolyte development", "abstract": "  Despite the widespread applications of machine learning force fields (MLFF)\nin solids and small molecules, there is a notable gap in applying MLFF to\nsimulate liquid electrolyte, a critical component of the current commercial\nlithium-ion battery. In this work, we introduce BAMBOO (\\textbf{B}yteDance\n\\textbf{A}I \\textbf{M}olecular Simulation \\textbf{Boo}ster), a predictive\nframework for molecular dynamics (MD) simulations, with a demonstration of its\ncapability in the context of liquid electrolyte for lithium batteries. We\ndesign a physics-inspired graph equivariant transformer architecture as the\nbackbone of BAMBOO to learn from quantum mechanical simulations. Additionally,\nwe introduce an ensemble knowledge distillation approach and apply it to MLFFs\nto reduce the fluctuation of observations from MD simulations. Finally, we\npropose a density alignment algorithm to align BAMBOO with experimental\nmeasurements. BAMBOO demonstrates state-of-the-art accuracy in predicting key\nelectrolyte properties such as density, viscosity, and ionic conductivity\nacross various solvents and salt combinations. The current model, trained on\nmore than 15 chemical species, achieves the average density error of 0.01\ng/cm$^3$ on various compositions compared with experiment.\n", "date": "2024-04-10", "categories": []}
{"id": "2504.00223", "title": "A machine learning platform for development of low flammability polymers", "abstract": "  Flammability index (FI) and cone calorimetry outcomes, such as maximum heat\nrelease rate, time to ignition, total smoke release, and fire growth rate, are\ncritical factors in evaluating the fire safety of polymers. However, predicting\nthese properties is challenging due to the complexity of material behavior\nunder heat exposure. In this work, we investigate the use of machine learning\n(ML) techniques to predict these flammability metrics. We generated synthetic\npolymers using Synthetic Data Vault to augment the experimental dataset. Our\ncomprehensive ML investigation employed both our polymer descriptors and those\ngenerated by the RDkit library. Despite the challenges of limited experimental\ndata, our models demonstrate the potential to accurately predict FI and cone\ncalorimetry outcomes, which could be instrumental in designing safer polymers.\nAdditionally, we developed POLYCOMPRED, a module integrated into the\ncloud-based MatVerse platform, providing an accessible, web-based interface for\nflammability prediction. This work provides not only the predictive modeling of\npolymer flammability but also an interactive analysis tool for the discovery\nand design of new materials with tailored fire-resistant properties.\n", "date": "2025-03-31", "categories": []}
{"id": "2504.00712", "title": "Spectral Normalization and Voigt-Reuss net: A universal approach to\n  microstructure-property forecasting with physical guarantees", "abstract": "  Heterogeneous materials are crucial to producing lightweight components,\nfunctional components, and structures composed of them. A crucial step in the\ndesign process is the rapid evaluation of their effective mechanical, thermal,\nor, in general, constitutive properties. The established procedure is to use\nforward models that accept microstructure geometry and local constitutive\nproperties as inputs. The classical simulation-based approach, which uses,\ne.g., finite elements and FFT-based solvers, can require substantial\ncomputational resources. At the same time, simulation-based models struggle to\nprovide gradients with respect to the microstructure and the constitutive\nparameters. Such gradients are, however, of paramount importance for\nmicrostructure design and for inverting the microstructure-property mapping.\nMachine learning surrogates can excel in these situations. However, they can\nlead to unphysical predictions that violate essential bounds on the\nconstitutive response, such as the upper (Voigt-like) or the lower (Reuss-like)\nbound in linear elasticity. Therefore, we propose a novel spectral\nnormalization scheme that a priori enforces these bounds. The approach is fully\nagnostic with respect to the chosen microstructural features and the utilized\nsurrogate model. All of these will automatically and strictly predict outputs\nthat obey the upper and lower bounds by construction. The technique can be used\nfor any constitutive tensor that is symmetric and where upper and lower bounds\n(in the L\\\"owner sense) exist, i.e., for permeability, thermal conductivity,\nlinear elasticity, and many more. We demonstrate the use of spectral\nnormalization in the Voigt-Reuss net using a simple neural network. Numerical\nexamples on truly extensive datasets illustrate the improved accuracy,\nrobustness, and independence of the type of input features in comparison to\nmuch-used neural networks.\n", "date": "2025-04-01", "categories": []}
{"id": "2503.23491", "title": "POINT$^{2}$: A Polymer Informatics Training and Testing Database", "abstract": "  The advancement of polymer informatics has been significantly propelled by\nthe integration of machine learning (ML) techniques, enabling the rapid\nprediction of polymer properties and expediting the discovery of\nhigh-performance polymeric materials. However, the field lacks a standardized\nworkflow that encompasses prediction accuracy, uncertainty quantification, ML\ninterpretability, and polymer synthesizability. In this study, we introduce\nPOINT$^{2}$ (POlymer INformatics Training and Testing), a comprehensive\nbenchmark database and protocol designed to address these critical challenges.\nLeveraging the existing labeled datasets and the unlabeled PI1M dataset, a\ncollection of approximately one million virtual polymers generated via a\nrecurrent neural network trained on the realistic polymers, we develop an\nensemble of ML models, including Quantile Random Forests, Multilayer\nPerceptrons with dropout, Graph Neural Networks, and pretrained large language\nmodels. These models are coupled with diverse polymer representations such as\nMorgan, MACCS, RDKit, Topological, Atom Pair fingerprints, and graph-based\ndescriptors to achieve property predictions, uncertainty estimations, model\ninterpretability, and template-based polymerization synthesizability across a\nspectrum of properties, including gas permeability, thermal conductivity, glass\ntransition temperature, melting temperature, fractional free volume, and\ndensity. The POINT$^{2}$ database can serve as a valuable resource for the\npolymer informatics community for polymer discovery and optimization.\n", "date": "2025-03-30", "categories": []}
{"id": "2503.23766", "title": "Accelerating High-Efficiency Organic Photovoltaic Discovery via\n  Pretrained Graph Neural Networks and Generative Reinforcement Learning", "abstract": "  Organic photovoltaic (OPV) materials offer a promising avenue toward\ncost-effective solar energy utilization. However, optimizing donor-acceptor\n(D-A) combinations to achieve high power conversion efficiency (PCE) remains a\nsignificant challenge. In this work, we propose a framework that integrates\nlarge-scale pretraining of graph neural networks (GNNs) with a GPT-2\n(Generative Pretrained Transformer 2)-based reinforcement learning (RL)\nstrategy to design OPV molecules with potentially high PCE. This approach\nproduces candidate molecules with predicted efficiencies approaching 21\\%,\nalthough further experimental validation is required. Moreover, we conducted a\npreliminary fragment-level analysis to identify structural motifs recognized by\nthe RL model that may contribute to enhanced PCE, thus providing design\nguidelines for the broader research community. To facilitate continued\ndiscovery, we are building the largest open-source OPV dataset to date,\nexpected to include nearly 3,000 donor-acceptor pairs. Finally, we discuss\nplans to collaborate with experimental teams on synthesizing and characterizing\nAI-designed molecules, which will provide new data to refine and improve our\npredictive and generative models.\n", "date": "2025-03-31", "categories": []}
{"id": "2503.23794", "title": "Force-Free Molecular Dynamics Through Autoregressive Equivariant\n  Networks", "abstract": "  Molecular dynamics (MD) simulations play a crucial role in scientific\nresearch. Yet their computational cost often limits the timescales and system\nsizes that can be explored. Most data-driven efforts have been focused on\nreducing the computational cost of accurate interatomic forces required for\nsolving the equations of motion. Despite their success, however, these machine\nlearning interatomic potentials (MLIPs) are still bound to small time-steps. In\nthis work, we introduce TrajCast, a transferable and data-efficient framework\nbased on autoregressive equivariant message passing networks that directly\nupdates atomic positions and velocities lifting the constraints imposed by\ntraditional numerical integration. We benchmark our framework across various\nsystems, including a small molecule, crystalline material, and bulk liquid,\ndemonstrating excellent agreement with reference MD simulations for structural,\ndynamical, and energetic properties. Depending on the system, TrajCast allows\nfor forecast intervals up to $30\\times$ larger than traditional MD time-steps,\ngenerating over 15 ns of trajectory data per day for a solid with more than\n4,000 atoms. By enabling efficient large-scale simulations over extended\ntimescales, TrajCast can accelerate materials discovery and explore physical\nphenomena beyond the reach of traditional simulations and experiments. An\nopen-source implementation of TrajCast is accessible under\nhttps://github.com/IBM/trajcast.\n", "date": "2025-03-31", "categories": []}
{"id": "2503.21560", "title": "Statistical learning of structure-property relationships for transport\n  in porous media, using hybrid AI modeling", "abstract": "  The 3D microstructure of porous media, such as electrodes in lithium-ion\nbatteries or fiber-based materials, significantly impacts the resulting\nmacroscopic properties, including effective diffusivity or permeability.\nConsequently, quantitative structure-property relationships, which link\nstructural descriptors of 3D microstructures such as porosity or geodesic\ntortuosity to effective transport properties, are crucial for further\noptimizing the performance of porous media. To overcome the limitations of 3D\nimaging, parametric stochastic 3D microstructure modeling is a powerful tool to\ngenerate many virtual but realistic structures at the cost of computer\nsimulations. The present paper uses 90,000 virtually generated 3D\nmicrostructures of porous media derived from literature by systematically\nvarying parameters of stochastic 3D microstructure models. Previously, this\ndata set has been used to establish quantitative microstructure-property\nrelationships. The present paper extends these findings by applying a hybrid AI\nframework to this data set. More precisely, symbolic regression, powered by\ndeep neural networks, genetic algorithms, and graph attention networks, is used\nto derive precise and robust analytical equations. These equations model the\nrelationships between structural descriptors and effective transport properties\nwithout requiring manual specification of the underlying functional\nrelationship. By integrating AI with traditional computational methods, the\nhybrid AI framework not only generates predictive equations but also enhances\nconventional modeling approaches by capturing relationships influenced by\nspecific microstructural features traditionally underrepresented. Thus, this\npaper significantly advances the predictive modeling capabilities in materials\nscience, offering vital insights for designing and optimizing new materials\nwith tailored transport properties.\n", "date": "2025-03-27", "categories": []}
{"id": "2404.07943", "title": "A Pretraining-Finetuning Computational Framework for Material\n  Homogenization", "abstract": "  Homogenization is a fundamental tool for studying multiscale physical\nphenomena. Traditional numerical homogenization methods, heavily reliant on\nfinite element analysis, demand significant computational resources, especially\nfor complex geometries, materials, and high-resolution problems. To address\nthese challenges, we propose PreFine-Homo, a novel numerical homogenization\nframework comprising two phases: pretraining and fine-tuning. In the\npretraining phase, a Fourier Neural Operator (FNO) is trained on large datasets\nto learn the mapping from input geometries and material properties to\ndisplacement fields. In the fine-tuning phase, the pretrained predictions serve\nas initial solutions for iterative algorithms, drastically reducing the number\nof iterations needed for convergence. The pretraining phase of PreFine-Homo\ndelivers homogenization results up to 1000 times faster than conventional\nmethods, while the fine-tuning phase further enhances accuracy. Moreover, the\nfine-tuning phase grants PreFine-Homo unlimited generalization capabilities,\nenabling continuous learning and improvement as data availability increases. We\nvalidate PreFine-Homo by predicting the effective elastic tensor for 3D\nperiodic materials, specifically Triply Periodic Minimal Surfaces (TPMS). The\nresults demonstrate that PreFine-Homo achieves high precision, exceptional\nefficiency, robust learning capabilities, and strong extrapolation ability,\nestablishing it as a powerful tool for multiscale homogenization tasks.\n", "date": "2024-03-18", "categories": []}
{"id": "2503.17870", "title": "Accelerating and enhancing thermodynamic simulations of electrochemical\n  interfaces", "abstract": "  Electrochemical interfaces are crucial in catalysis, energy storage, and\ncorrosion, where their stability and reactivity depend on complex interactions\nbetween the electrode, adsorbates, and electrolyte. Predicting stable surface\nstructures remains challenging, as traditional surface Pourbaix diagrams tend\nto either rely on expert knowledge or costly $\\textit{ab initio}$ sampling, and\nneglect thermodynamic equilibration with the environment. Machine learning (ML)\npotentials can accelerate static modeling but often overlook dynamic surface\ntransformations. Here, we extend the Virtual Surface Site Relaxation-Monte\nCarlo (VSSR-MC) method to autonomously sample surface reconstructions modeled\nunder aqueous electrochemical conditions. Through fine-tuning foundational ML\nforce fields, we accurately and efficiently predict surface energetics,\nrecovering known Pt(111) phases and revealing new LaMnO$_\\mathrm{3}$(001)\nsurface reconstructions. By explicitly accounting for bulk-electrolyte\nequilibria, our framework enhances electrochemical stability predictions,\noffering a scalable approach to understanding and designing materials for\nelectrochemical applications.\n", "date": "2025-03-22", "categories": []}
{"id": "2501.06775", "title": "Hierarchy-Boosted Funnel Learning for Identifying Semiconductors with\n  Ultralow Lattice Thermal Conductivity", "abstract": "  Data-driven machine learning (ML) has demonstrated tremendous potential in\nmaterial property predictions. However, the scarcity of materials data with\ncostly property labels in the vast chemical space presents a significant\nchallenge for ML in efficiently predicting properties and uncovering\nstructure-property relationships. Here, we propose a novel hierarchy-boosted\nfunnel learning (HiBoFL) framework, which is successfully applied to identify\nsemiconductors with ultralow lattice thermal conductivity\n($\\kappa_\\mathrm{L}$). By training on only a few hundred materials targeted by\nunsupervised learning from a pool of hundreds of thousands, we achieve\nefficient and interpretable supervised predictions of ultralow\n$\\kappa_\\mathrm{L}$, thereby circumventing large-scale brute-force \\textit{ab\ninitio} calculations without clear objectives. As a result, we provide a list\nof candidates with ultralow $\\kappa_\\mathrm{L}$ for potential thermoelectric\napplications and discover a new factor that significantly influences structural\nanharmonicity. This HiBoFL framework offers a novel practical pathway for\naccelerating the discovery of functional materials.\n", "date": "2025-01-12", "categories": []}
{"id": "2503.16667", "title": "A preliminary data fusion study to assess the feasibility of Foundation\n  Process-Property Models in Laser Powder Bed Fusion", "abstract": "  Foundation models are at the forefront of an increasing number of critical\napplications. In regards to technologies such as additive manufacturing (AM),\nthese models have the potential to dramatically accelerate process optimization\nand, in turn, design of next generation materials. A major challenge that\nimpedes the construction of foundation process-property models is data\nscarcity. To understand the impact of this challenge, and since foundation\nmodels rely on data fusion, in this work we conduct controlled experiments\nwhere we focus on the transferability of information across different material\nsystems and properties. More specifically, we generate experimental datasets\nfrom 17-4 PH and 316L stainless steels (SSs) in Laser Powder Bed Fusion (LPBF)\nwhere we measure the effect of five process parameters on porosity and\nhardness. We then leverage Gaussian processes (GPs) for process-property\nmodeling in various configurations to test if knowledge about one material\nsystem or property can be leveraged to build more accurate machine learning\nmodels for other material systems or properties. Through extensive\ncross-validation studies and probing the GPs' interpretable hyperparameters, we\nstudy the intricate relation among data size and dimensionality, complexity of\nthe process-property relations, noise, and characteristics of machine learning\nmodels. Our findings highlight the need for structured learning approaches that\nincorporate domain knowledge in building foundation process-property models\nrather than relying on uninformed data fusion in data-limited applications.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.17171", "title": "Generative adversarial framework to calibrate excursion set models for\n  the 3D morphology of all-solid-state battery cathodes", "abstract": "  This paper presents a computational method for generating virtual 3D\nmorphologies of functional materials using low-parametric stochastic geometry\nmodels, i.e., digital twins, calibrated with 2D microscopy images. These\ndigital twins allow systematic parameter variations to simulate various\nmorphologies, that can be deployed for virtual materials testing by means of\nspatially resolved numerical simulations of macroscopic properties. Generative\nadversarial networks (GANs) have gained popularity for calibrating models to\ngenerate realistic 3D morphologies. However, GANs often comprise of numerous\nuninterpretable parameters make systematic variation of morphologies for\nvirtual materials testing challenging. In contrast, low-parametric stochastic\ngeometry models (e.g., based on Gaussian random fields) enable targeted\nvariation but may struggle to mimic complex morphologies. Combining GANs with\nadvanced stochastic geometry models (e.g., excursion sets of more general\nrandom fields) addresses these limitations, allowing model calibration solely\nfrom 2D image data. This approach is demonstrated by generating a digital twin\nof all-solid-state battery (ASSB) cathodes. Since the digital twins are\nparametric, they support systematic exploration of structural scenarios and\ntheir macroscopic properties. The proposed method facilitates simulation\nstudies for optimizing 3D morphologies, benefiting not only ASSB cathodes but\nalso other materials with similar structures.\n", "date": "2025-03-21", "categories": []}
{"id": "2501.18137", "title": "Tensor Completion for Surrogate Modeling of Material Property Prediction", "abstract": "  When designing materials to optimize certain properties, there are often many\npossible configurations of designs that need to be explored. For example, the\nmaterials' composition of elements will affect properties such as strength or\nconductivity, which are necessary to know when developing new materials.\nExploring all combinations of elements to find optimal materials becomes very\ntime consuming, especially when there are more design variables. For this\nreason, there is growing interest in using machine learning (ML) to predict a\nmaterial's properties. In this work, we model the optimization of certain\nmaterial properties as a tensor completion problem, to leverage the structure\nof our datasets and navigate the vast number of combinations of material\nconfigurations. Across a variety of material property prediction tasks, our\nexperiments show tensor completion methods achieving 10-20% decreased error\ncompared with baseline ML models such as GradientBoosting and Multilayer\nPerceptron (MLP), while maintaining similar training speed.\n", "date": "2025-01-30", "categories": []}
{"id": "2502.16457", "title": "Towards Fully-Automated Materials Discovery via Large-Scale Synthesis\n  Dataset and Expert-Level LLM-as-a-Judge", "abstract": "  Materials synthesis is vital for innovations such as energy storage,\ncatalysis, electronics, and biomedical devices. Yet, the process relies heavily\non empirical, trial-and-error methods guided by expert intuition. Our work aims\nto support the materials science community by providing a practical,\ndata-driven resource. We have curated a comprehensive dataset of 17K\nexpert-verified synthesis recipes from open-access literature, which forms the\nbasis of our newly developed benchmark, AlchemyBench. AlchemyBench offers an\nend-to-end framework that supports research in large language models applied to\nsynthesis prediction. It encompasses key tasks, including raw materials and\nequipment prediction, synthesis procedure generation, and characterization\noutcome forecasting. We propose an LLM-as-a-Judge framework that leverages\nlarge language models for automated evaluation, demonstrating strong\nstatistical agreement with expert assessments. Overall, our contributions offer\na supportive foundation for exploring the capabilities of LLMs in predicting\nand guiding materials synthesis, ultimately paving the way for more efficient\nexperimental design and accelerated innovation in materials science.\n", "date": "2025-02-23", "categories": []}
{"id": "2410.05307", "title": "Topology-Informed Machine Learning for Efficient Prediction of Solid\n  Oxide Fuel Cell Electrode Polarization", "abstract": "  Machine learning has emerged as a potent computational tool for expediting\nresearch and development in solid oxide fuel cell electrodes. The effective\napplication of machine learning for performance prediction requires\ntransforming electrode microstructure into a format compatible with artificial\nneural networks. Input data may range from a comprehensive digital material\nrepresentation of the electrode to a selected set of microstructural\nparameters. The chosen representation significantly influences the performance\nand results of the network. Here, we show a novel approach utilizing\npersistence representation derived from computational topology. Using 500\nmicrostructures and current-voltage characteristics obtained with 3D\nfirst-principles simulations, we have prepared an artificial neural network\nmodel that can replicate current-voltage characteristics of unseen\nmicrostructures based on their persistent image representation. The artificial\nneural network can accurately predict the polarization curve of solid oxide\nfuel cell electrodes. The presented method incorporates complex microstructural\ninformation from the digital material representation while requiring\nsubstantially less computational resources (preprocessing and prediction time\napproximately 1 min) compared to our high-fidelity simulations (simulation time\napproximately 1 hour) to obtain a single current-potential characteristic for\none microstructure.\n", "date": "2024-10-04", "categories": []}
{"id": "2407.20273", "title": "Learning Physics-Consistent Material Behavior from Dynamic Displacements", "abstract": "  Accurately modeling the mechanical behavior of materials is crucial for\nnumerous engineering applications. The quality of these models depends directly\non the accuracy of the constitutive law that defines the stress-strain\nrelation. However, discovering these constitutive material laws remains a\nsignificant challenge, in particular when only material deformation data is\navailable. To address this challenge, unsupervised machine learning methods\nhave been proposed to learn the constitutive law from deformation data.\nNonetheless, existing approaches have several limitations: they either fail to\nensure that the learned constitutive relations are consistent with physical\nprinciples, or they rely on boundary force data for training which are\nunavailable in many in-situ scenarios. Here, we introduce a machine learning\napproach to learn physics-consistent constitutive relations solely from\nmaterial deformation without boundary force information. This is achieved by\nconsidering a dynamic formulation rather than static equilibrium data and\napplying an input convex neural network (ICNN). We validate the effectiveness\nof the proposed method on a diverse range of hyperelastic material laws. We\ndemonstrate that it is robust to a significant level of noise and that it\nconverges to the ground truth with increasing data resolution. We also show\nthat the model can be effectively trained using a displacement field from a\nsubdomain of the test specimen and that the learned constitutive relation from\none material sample is transferable to other samples with different geometries.\nThe developed methodology provides an effective tool for discovering\nconstitutive relations. It is, due to its design based on dynamics,\nparticularly suited for applications to strain-rate-dependent materials and\nsituations where constitutive laws need to be inferred from in-situ\nmeasurements without access to global force data.\n", "date": "2024-07-25", "categories": []}
{"id": "2307.08929", "title": "Active learning of effective Hamiltonian for super-large-scale atomic\n  structures", "abstract": "  The first-principles-based effective Hamiltonian scheme provides one of the\nmost accurate modeling technique for large-scale structures, especially for\nferroelectrics. However, the parameterization of the effective Hamiltonian is\ncomplicated and can be difficult for some complex systems such as high-entropy\nperovskites. Here, we propose a general form of effective Hamiltonian and\ndevelop an active machine learning approach to parameterize the effective\nHamiltonian based on Bayesian linear regression. The parameterization is\nemployed in molecular dynamics simulations with the prediction of energy,\nforces, stress and their uncertainties at each step, which decides whether\nfirst-principles calculations are executed to retrain the parameters.\nStructures of BaTiO$_3$, Pb(Zr$_{0.75}$Ti$_{0.25}$)O$_3$ and (Pb,Sr)TiO$_3$\nsystem are taken as examples to show the accuracy of this approach, as compared\nwith conventional parametrization method and experiments. This machine learning\napproach provides a universal and automatic way to compute the effective\nHamiltonian parameters for any considered complex systems with\nsuper-large-scale (more than $10^7$ atoms) atomic structures.\n", "date": "2023-07-18", "categories": []}
{"id": "2211.11403", "title": "General time-reversal equivariant neural network potential for magnetic\n  materials", "abstract": "  This study introduces time-reversal E(3)-equivariant neural network and\nSpinGNN++ framework for constructing a comprehensive interatomic potential for\nmagnetic systems, encompassing spin-orbit coupling and noncollinear magnetic\nmoments. SpinGNN++ integrates multitask spin equivariant neural network with\nexplicit spin-lattice terms, including Heisenberg, Dzyaloshinskii-Moriya,\nKitaev, single-ion anisotropy, and biquadratic interactions, and employs\ntime-reversal equivariant neural network to learn high-order spin-lattice\ninteractions using time-reversal E(3)-equivariant convolutions. To validate\nSpinGNN++, a complex magnetic model dataset is introduced as a benchmark and\nemployed to demonstrate its capabilities. SpinGNN++ provides accurate\ndescriptions of the complex spin-lattice coupling in monolayer CrI$_3$ and\nCrTe$_2$, achieving sub-meV errors. Importantly, it facilitates large-scale\nparallel spin-lattice dynamics, thereby enabling the exploration of associated\nproperties, including the magnetic ground state and phase transition.\nRemarkably, SpinGNN++ identifies a new ferrimagnetic state as the ground\nmagnetic state for monolayer CrTe2, thereby enriching its phase diagram and\nproviding deeper insights into the distinct magnetic signals observed in\nvarious experiments.\n", "date": "2022-11-21", "categories": []}
{"id": "2503.09814", "title": "A practical guide to machine learning interatomic potentials -- Status\n  and future", "abstract": "  The rapid development and large body of literature on machine learning\ninteratomic potentials (MLIPs) can make it difficult to know how to proceed for\nresearchers who are not experts but wish to use these tools. The spirit of this\nreview is to help such researchers by serving as a practical, accessible guide\nto the state-of-the-art in MLIPs. This review paper covers a broad range of\ntopics related to MLIPs, including (i) central aspects of how and why MLIPs are\nenablers of many exciting advancements in molecular modeling, (ii) the main\nunderpinnings of different types of MLIPs, including their basic structure and\nformalism, (iii) the potentially transformative impact of universal MLIPs for\nboth organic and inorganic systems, including an overview of the most recent\nadvances, capabilities, downsides, and potential applications of this nascent\nclass of MLIPs, (iv) a practical guide for estimating and understanding the\nexecution speed of MLIPs, including guidance for users based on hardware\navailability, type of MLIP used, and prospective simulation size and time, (v)\na manual for what MLIP a user should choose for a given application by\nconsidering hardware resources, speed requirements, energy and force accuracy\nrequirements, as well as guidance for choosing pre-trained potentials or\nfitting a new potential from scratch, (vi) discussion around MLIP\ninfrastructure, including sources of training data, pre-trained potentials, and\nhardware resources for training, (vii) summary of some key limitations of\npresent MLIPs and current approaches to mitigate such limitations, including\nmethods of including long-range interactions, handling magnetic systems, and\ntreatment of excited states, and finally (viii) we finish with some more\nspeculative thoughts on what the future holds for the development and\napplication of MLIPs over the next 3-10+ years.\n", "date": "2025-03-12", "categories": []}
{"id": "2503.10073", "title": "Impact of buckypaper on the mechanical properties and failure modes of\n  composites", "abstract": "  Recently, there has been an interest in the incorporation of buckypaper (BP),\nor carbon nanotube (CNT) membranes, in composite laminates. Research has shown\nthat using BP in contrast to nanotube doped resin enables the introduction of a\nhigher CNT weight fraction which offers multiple benefits including higher\npiezo resistivity for health monitoring applications and enhanced mechanical\nresponse for structural applications. However, their impact on the deformation\nand failure mechanisms of composite laminates has not been investigated\nthoroughly. Understanding these issues experimentally would require a carefully\nexecuted test plan involving a multitude of design parameters such as BP\ngeometry and placement, material anisotropy and variability, and laminate\nstacking sequence. This paper presents a deep learning (DL)-based surrogate\nmodel for studying the mechanical response of hybrid carbon fiber reinforced\npolymer (CFRP) composite laminates with BP interleaves under various mechanical\nloads. The surrogate model utilizes a long short-term memory architecture\nimplemented within a DL framework and predicts the laminate global response for\na given configuration, geometry, and loading condition. The DL framework\ntraining and cross-validation are performed via data acquisition from a series\nof three-point bend tests conducted through finite element analysis (FEA) and\nin-house experiments, respectively. The model predictions show good agreement\nwith FEA simulations and experimental results, where CFRP with two BP\ninterleaves showed enhanced flexural strength and modulus over pristine\nsamples. This enhancement can be attributed to the excellent crack retardation\ncapabilities of CNTs, particularly in the interlaminar region.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.10102", "title": "Geometric Parameter Estimations of Perovskite Solar Cells Based on\n  Optical Simulations", "abstract": "  This paper presents a non-invasive approach to estimate the layer thicknesses\nof perovskite solar cells. The thicknesses are predicted by a convolutional\nneural network that leverages the external quantum efficiency of a perovskite\nsolar cell. The network is trained in thickness ranges where the optical\nproperties are constant, and these ranges set the constraints for the network's\napplication. Due to light sensitivity issues with opaque perovskites, the\nconvolutional neural network showed better performance with transparent\nperovskites. To optimize the performance and reduce the root mean square error,\nwe tried different sampling methods, image specifications, and Bayesian\noptimization for hyperparameter tuning. While sampling methods showed marginal\nimprovement, implementing Bayesian optimization demonstrated high accuracy.\nOther minor optimization attempts include experimenting with input\nspecifications and pre-processing approaches. The results confirm the\nfeasibility, efficiency, and effectiveness of a convolution neural network for\npredicting perovskite solar cells' layer thicknesses based on controlled\nexperiments.\n", "date": "2025-03-13", "categories": []}
{"id": "2312.00111", "title": "Multimodal Foundation Models for Material Property Prediction and\n  Discovery", "abstract": "  Artificial intelligence is transforming computational materials science,\nimproving the prediction of material properties, and accelerating the discovery\nof novel materials. Recently, publicly available material data repositories\nhave grown rapidly. This growth encompasses not only more materials but also a\ngreater variety and quantity of their associated properties. Existing machine\nlearning efforts in materials science focus primarily on single-modality tasks,\ni.e. relationships between materials and a single physical property, thus not\ntaking advantage of the rich and multimodal set of material properties. Here,\nwe introduce Multimodal Learning for Materials (MultiMat), which enables\nself-supervised multi-modality training of foundation models for materials. We\ndemonstrate our framework's potential using data from the Materials Project\ndatabase on multiple axes: (i) MultiMat achieves state-of-the-art performance\nfor challenging material property prediction tasks; (ii) MultiMat enables novel\nand accurate material discovery via latent space similarity, enabling screening\nfor stable materials with desired properties; and (iii) MultiMat encodes\ninterpretable emergent features that may provide novel scientific insights.\n", "date": "2023-11-30", "categories": []}
{"id": "2301.03570", "title": "Principal deuterium Hugoniot via Quantum Monte Carlo and\n  $\\Delta$-learning", "abstract": "  We present a study of the principal deuterium Hugoniot for pressures up to\n$150$ GPa, using Machine Learning potentials (MLPs) trained with Quantum Monte\nCarlo (QMC) energies, forces and pressures. In particular, we adopted a\nrecently proposed workflow based on the combination of Gaussian kernel\nregression and $\\Delta$-learning. By fully taking advantage of this method, we\nexplicitly considered finite-temperature electrons in the dynamics, whose\neffects are highly relevant for temperatures above $10$ kK. The Hugoniot curve\nobtained by our MLPs shows a good agreement with the most recent experiments,\nparticularly in the region below 60 GPa. At larger pressures, our Hugoniot\ncurve is slightly more compressible than the one yielded by experiments, whose\nuncertainties generally increase, however, with pressure. Our work demonstrates\nthat QMC can be successfully combined with $\\Delta$-learning to deploy reliable\nMLPs for complex extended systems across different thermodynamic conditions, by\nkeeping the QMC precision at the computational cost of a mean-field\ncalculation.\n", "date": "2023-01-09", "categories": []}
{"id": "2407.06053", "title": "Learning local equivariant representations for quantum operators", "abstract": "  Predicting quantum operator matrices such as Hamiltonian, overlap, and\ndensity matrices in the density functional theory (DFT) framework is crucial\nfor material science. Current methods often focus on individual operators and\nstruggle with efficiency and scalability for large systems. Here we introduce a\nnovel deep learning model, SLEM (strictly localized equivariant\nmessage-passing) for predicting multiple quantum operators, that achieves\nstate-of-the-art accuracy while dramatically improving computational\nefficiency. SLEM's key innovation is its strict locality-based design for\nequivariant representations of quantum tensors while preserving physical\nsymmetries. This enables complex many-body dependency without expanding the\neffective receptive field, leading to superior data efficiency and\ntransferability. Using an innovative SO(2) convolution and invariant overlap\nparameterization, SLEM reduces the computational complexity of high-order\ntensor products and is therefore capable of handling systems requiring the $f$\nand $g$ orbitals in their basis sets. We demonstrate SLEM's capabilities across\ndiverse 2D and 3D materials, achieving high accuracy even with limited training\ndata. SLEM's design facilitates efficient parallelization, potentially\nextending DFT simulations to systems with device-level sizes, opening new\npossibilities for large-scale quantum simulations and high-throughput materials\ndiscovery.\n", "date": "2024-07-08", "categories": []}
{"id": "2503.07619", "title": "Physics- and data-driven Active Learning of neural network\n  representations for free energy functions of materials from statistical\n  mechanics", "abstract": "  Accurate free energy representations are crucial for understanding phase\ndynamics in materials. We employ a scale-bridging approach to incorporate\natomistic information into our free energy model by training a neural network\non DFT-informed Monte Carlo data. To optimize sampling in the high-dimensional\nMonte Carlo space, we present an Active Learning framework that integrates\nspace-filling sampling, uncertainty-based sampling, and physics-informed\nsampling. Additionally, our approach includes methods such as hyperparameter\ntuning, dynamic sampling, and novelty enforcement. These strategies can be\ncombined to reduce MSE,either globally or in targeted regions of interest,while\nminimizing the number of required data points. The framework introduced here is\nbroadly applicable to Monte Carlo sampling of a range of materials systems.\n", "date": "2025-02-25", "categories": []}
{"id": "2503.05525", "title": "Machine Learning for Improved Density Functional Theory Thermodynamics", "abstract": "  The predictive accuracy of density functional theory (DFT) for alloy\nformation enthalpies is often limited by intrinsic energy resolution errors,\nparticularly in ternary phase stability calculations. In this work, we present\na machine learning (ML) approach to systematically correct these errors,\nimproving the reliability of first-principles predictions. A neural network\nmodel has been trained to predict the discrepancy between DFT-calculated and\nexperimentally measured enthalpies for binary and ternary alloys and compounds.\nThe model utilizes a structured feature set comprising elemental\nconcentrations, atomic numbers, and interaction terms to capture key chemical\nand structural effects. By applying supervised learning and rigorous data\ncuration we ensure a robust and physically meaningful correction. The model is\nimplemented as a multi-layer perceptron (MLP) regressor with three hidden\nlayers, optimized through leave-one-out cross-validation (LOOCV) and k-fold\ncross-validation to prevent overfitting. We illustrate the effectiveness of\nthis method by applying it to the Al-Ni-Pd and Al-Ni-Ti systems, which are of\ninterest for high-temperature applications in aerospace and protective\ncoatings.\n", "date": "2025-03-07", "categories": []}
{"id": "2503.04027", "title": "Machine learning driven search of hydrogen storage materials", "abstract": "  The transition to a low-carbon economy demands efficient and sustainable\nenergy-storage solutions, with hydrogen emerging as a promising clean-energy\ncarrier and with metal hydrides recognized for their hydrogen-storage capacity.\nHere, we leverage machine learning (ML) to predict hydrogen-to-metal (H/M)\nratios and solution energy by incorporating thermodynamic parameters and local\nlattice distortion (LLD) as key features. Our best-performing ML model provides\nimprovements to H/M ratios and solution energies over a broad class of ternary\nalloys (easily extendable to multi-principal-element alloys), such as Ti-Nb-X\n(X = Mo, Cr, Hf, Ta, V, Zr) and Co-Ni-X (X = Al, Mg, V). Ti-Nb-Mo alloys reveal\ncompositional effects in H-storage behavior, in particular Ti, Nb, and V\nenhance H-storage capacity, while Mo reduces H/M and hydrogen weight percent by\n40-50%. We attributed to slow hydrogen kinetics in molybdenum rich alloys,\nwhich is validated by our pressure-composition isotherm (PCT) experiments on\npure Ti and Ti5Mo95 alloys. Density functional theory (DFT) and molecular\nsimulations also confirm that Ti and Nb promote H diffusion, whereas Mo hinders\nit, highlighting the interplay between electronic structure, lattice\ndistortions, and hydrogen uptake. Notably, our Gradient Boosting Regression\nmodel identifies LLD as a critical factor in H/M predictions. To aid material\nselection, we present two periodic tables illustrating elemental effects on (a)\nH2 wt% and (b) solution energy, derived from ML, and provide a reference for\nidentifying alloying elements that enhance hydrogen solubility and storage.\n", "date": "2025-03-06", "categories": []}
{"id": "2503.01932", "title": "A General Neural Network Potential for Energetic Materials with C, H, N,\n  and O elements", "abstract": "  The discovery and optimization of high-energy materials (HEMs) are\nconstrained by the prohibitive computational expense and prolonged development\ncycles inherent in conventional approaches. In this work, we develop a general\nneural network potential (NNP) that efficiently predicts the structural,\nmechanical, and decomposition properties of HEMs composed of C, H, N, and O.\nOur framework leverages pre-trained NNP models, fine-tuned using transfer\nlearning on energy and force data derived from density functional theory (DFT)\ncalculations. This strategy enables rapid adaptation across 20 different HEM\nsystems while maintaining DFT-level accuracy, significantly reducing\ncomputational costs. A key aspect of this work is the ability of NNP model to\ncapture the chemical activity space of HEMs, accurately describe the key atomic\ninteractions and reaction mechanisms during thermal decomposition. The general\nNNP model has been applied in molecular dynamics (MD) simulations and validated\nwith experimental data for various HEM structures. Results show that the NNP\nmodel accurately predicts the structural, mechanical, and decomposition\nproperties of HEMs by effectively describing their chemical activity space.\nCompared to traditional force fields, it offers superior DFT-level accuracy and\ngeneralization across both microscopic and macroscopic properties, reducing the\ncomputational and experimental costs. This work provides an efficient strategy\nfor the design and development of HEMs and proposes a promising framework for\nintegrating DFT, machine learning, and experimental methods in materials\nresearch. (To facilitate further research and practical applications, we\nopen-source our NNP model on GitHub:\nhttps://github.com/MingjieWen/General-NNP-model-for-C-H-N-O-Energetic-Materials.)\n", "date": "2025-03-03", "categories": []}
{"id": "2503.02209", "title": "CrystalFramer: Rethinking the Role of Frames for SE(3)-Invariant Crystal\n  Structure Modeling", "abstract": "  Crystal structure modeling with graph neural networks is essential for\nvarious applications in materials informatics, and capturing SE(3)-invariant\ngeometric features is a fundamental requirement for these networks. A\nstraightforward approach is to model with orientation-standardized structures\nthrough structure-aligned coordinate systems, or\"frames.\" However, unlike\nmolecules, determining frames for crystal structures is challenging due to\ntheir infinite and highly symmetric nature. In particular, existing methods\nrely on a statically fixed frame for each structure, determined solely by its\nstructural information, regardless of the task under consideration. Here, we\nrethink the role of frames, questioning whether such simplistic alignment with\nthe structure is sufficient, and propose the concept of dynamic frames. While\naccommodating the infinite and symmetric nature of crystals, these frames\nprovide each atom with a dynamic view of its local environment, focusing on\nactively interacting atoms. We demonstrate this concept by utilizing the\nattention mechanism in a recent transformer-based crystal encoder, resulting in\na new architecture called CrystalFramer. Extensive experiments show that\nCrystalFramer outperforms conventional frames and existing crystal encoders in\nvarious crystal property prediction tasks.\n", "date": "2025-03-04", "categories": []}
{"id": "2410.08452", "title": "Kolmogorov-Arnold Neural Networks for High-Entropy Alloys Design", "abstract": "  A wide range of deep learning-based machine learning techniques are\nextensively applied to the design of high-entropy alloys (HEAs), yielding\nnumerous valuable insights. Kolmogorov-Arnold Networks (KAN) is a recently\ndeveloped architecture that aims to improve both the accuracy and\ninterpretability of input features. In this work, we explore three different\ndatasets for HEA design and demonstrate the application of KAN for both\nclassification and regression models. In the first example, we use a KAN\nclassification model to predict the probability of single-phase formation in\nhigh-entropy carbide ceramics based on various properties such as mixing\nenthalpy and valence electron concentration. In the second example, we employ a\nKAN regression model to predict the yield strength and ultimate tensile\nstrength of HEAs based on their chemical composition and process conditions\nincluding annealing time, cold rolling percentage, and homogenization\ntemperature. The third example involves a KAN classification model to determine\nwhether a certain composition is an HEA or non-HEA, followed by a KAN regressor\nmodel to predict the bulk modulus of the identified HEA, aiming to identify\nHEAs with high bulk modulus. In all three examples, KAN either outperform or\nmatch the performance in terms of accuracy such as F1 score for classification\nand Mean Square Error (MSE), and coefficient of determination (R2) for\nregression of the multilayer perceptron (MLP) by demonstrating the efficacy of\nKAN in handling both classification and regression tasks. We provide a\npromising direction for future research to explore advanced machine learning\ntechniques, which lead to more accurate predictions and better interpretability\nof complex materials, ultimately accelerating the discovery and optimization of\nHEAs with desirable properties.\n", "date": "2024-10-11", "categories": []}
{"id": "2503.00152", "title": "Invariant Tokenization of Crystalline Materials for Language Model\n  Enabled Generation", "abstract": "  We consider the problem of crystal materials generation using language models\n(LMs). A key step is to convert 3D crystal structures into 1D sequences to be\nprocessed by LMs. Prior studies used the crystallographic information framework\n(CIF) file stream, which fails to ensure SE(3) and periodic invariance and may\nnot lead to unique sequence representations for a given crystal structure.\nHere, we propose a novel method, known as Mat2Seq, to tackle this challenge.\nMat2Seq converts 3D crystal structures into 1D sequences and ensures that\ndifferent mathematical descriptions of the same crystal are represented in a\nsingle unique sequence, thereby provably achieving SE(3) and periodic\ninvariance. Experimental results show that, with language models, Mat2Seq\nachieves promising performance in crystal structure generation as compared with\nprior methods.\n", "date": "2025-02-28", "categories": []}
{"id": "2503.00522", "title": "Periodic Materials Generation using Text-Guided Joint Diffusion Model", "abstract": "  Equivariant diffusion models have emerged as the prevailing approach for\ngenerating novel crystal materials due to their ability to leverage the\nphysical symmetries of periodic material structures. However, current models do\nnot effectively learn the joint distribution of atom types, fractional\ncoordinates, and lattice structure of the crystal material in a cohesive\nend-to-end diffusion framework. Also, none of these models work under realistic\nsetups, where users specify the desired characteristics that the generated\nstructures must match. In this work, we introduce TGDMat, a novel text-guided\ndiffusion model designed for 3D periodic material generation. Our approach\nintegrates global structural knowledge through textual descriptions at each\ndenoising step while jointly generating atom coordinates, types, and lattice\nstructure using a periodic-E(3)-equivariant graph neural network (GNN).\nExtensive experiments using popular datasets on benchmark tasks reveal that\nTGDMat outperforms existing baseline methods by a good margin. Notably, for the\nstructure prediction task, with just one generated sample, TGDMat outperforms\nall baseline models, highlighting the importance of text-guided diffusion.\nFurther, in the generation task, TGDMat surpasses all baselines and their\ntext-fusion variants, showcasing the effectiveness of the joint diffusion\nparadigm. Additionally, incorporating textual knowledge reduces overall\ntraining and sampling computational overhead while enhancing generative\nperformance when utilizing real-world textual prompts from experts.\n", "date": "2025-03-01", "categories": []}
{"id": "2302.13537", "title": "Global optimization in the discrete and variable-dimension\n  conformational space: The case of crystal with the strongest atomic cohesion", "abstract": "  We introduce a computational method to optimize target physical properties in\nthe full configuration space regarding atomic composition, chemical\nstoichiometry, and crystal structure. The approach combines the universal\npotential of the crystal graph neural network and Bayesian optimization. The\nproposed approach effectively obtains the crystal structure with the strongest\natomic cohesion from all possible crystals. Several new crystals with high\natomic cohesion are identified and confirmed by density functional theory for\nthermodynamic and dynamic stability. Our method introduces a novel approach to\ninverse materials design with additional functional properties for practical\napplications.\n", "date": "2023-02-27", "categories": []}
{"id": "2412.12122", "title": "AI-driven Inverse Design of Band-Tunable Mechanical Metastructures for\n  Tailored Vibration Mitigation", "abstract": "  On-demand vibration mitigation in a mechanical system needs the suitable\ndesign of multiscale metastructures, involving complex unit cells. In this\nstudy, immersing in the world of patterns and examining the structural details\nof some interesting motifs are extracted from the mechanical metastructure\nperspective. Nine interlaced metastructures are fabricated using additive\nmanufacturing, and corresponding vibration characteristics are studied\nexperimentally and numerically. Further, the band-gap modulation with metallic\ninserts in the honeycomb interlaced metastructures is also studied. AI-driven\ninverse design of such complex metastructures with a desired vibration\nmitigation profile can pave the way for addressing engineering challenges in\nhigh-precision manufacturing. The current inverse design methodologies are\nlimited to designing simple periodic structures based on limited variants of\nunit cells. Therefore, a novel forward analysis model with multi-head\nFEM-inspired spatial attention (FSA) is proposed to learn the complex geometry\nof the metastructures and predict corresponding transmissibility. Subsequently,\na multiscale Gaussian self-attention (MGSA) based inverse design model with\nGaussian function for 1D spectrum position encoding is developed to produce a\nsuitable metastructure for the desired vibration transmittance. The proposed AI\nframework demonstrated outstanding performance corresponding to the expected\nlocally resonant bandgaps in a targeted frequency range.\n", "date": "2024-12-03", "categories": []}
{"id": "2502.09791", "title": "Atom identification in bilayer moire materials with Gomb-Net", "abstract": "  Moire patterns in van der Waals bilayer materials complicate the analysis of\natomic-resolution images, hindering the atomic-scale insight typically\nattainable with scanning transmission electron microscopy. Here, we report a\nmethod to detect the positions and identities of atoms in each of the\nindividual layers that compose twisted bilayer heterostructures. We developed a\ndeep learning model, Gomb-Net, which identifies the coordinates and atomic\nspecies in each layer, effectively deconvoluting the moire pattern. This\nenables layer-specific mapping of quantities like strain and dopant\ndistributions, unlike other commonly used segmentation models which struggle\nwith moire-induced complexity. Using this approach, we explored the Se atom\nsubstitutional site distribution in a twisted fractional Janus WS2-WS2(1-x)Se2x\nheterostructure and found that layer-specific implantation sites are unaffected\nby the moire pattern's local energetic or electronic modulation. This\nadvancement enables atom identification within material regimes where it was\nnot possible before, opening new insights into previously inaccessible material\nphysics.\n", "date": "2025-02-13", "categories": []}
{"id": "2502.18127", "title": "Inverse Materials Design by Large Language Model-Assisted Generative\n  Framework", "abstract": "  Deep generative models hold great promise for inverse materials design, yet\ntheir efficiency and accuracy remain constrained by data scarcity and model\narchitecture. Here, we introduce AlloyGAN, a closed-loop framework that\nintegrates Large Language Model (LLM)-assisted text mining with Conditional\nGenerative Adversarial Networks (CGANs) to enhance data diversity and improve\ninverse design. Taking alloy discovery as a case study, AlloyGAN systematically\nrefines material candidates through iterative screening and experimental\nvalidation. For metallic glasses, the framework predicts thermodynamic\nproperties with discrepancies of less than 8% from experiments, demonstrating\nits robustness. By bridging generative AI with domain knowledge and validation\nworkflows, AlloyGAN offers a scalable approach to accelerate the discovery of\nmaterials with tailored properties, paving the way for broader applications in\nmaterials science.\n", "date": "2025-02-25", "categories": []}
{"id": "2502.16665", "title": "A new framework for X-ray absorption spectroscopy data analysis based on\n  machine learning: XASDAML", "abstract": "  X-ray absorption spectroscopy (XAS) is a powerful technique to probe the\nelectronic and structural properties of materials. With the rapid growth in\nboth the volume and complexity of XAS datasets driven by advancements in\nsynchrotron radiation facilities, there is an increasing demand for advanced\ncomputational tools capable of efficiently analyzing large-scale data. To\naddress these needs, we introduce XASDAML,a flexible, machine learning based\nframework that integrates the entire data-processing workflow-including dataset\nconstruction for spectra and structural descriptors, data filtering, ML\nmodeling, prediction, and model evaluation-into a unified platform.\nAdditionally, it supports comprehensive statistical analysis, leveraging\nmethods such as principal component analysis and clustering to reveal potential\npatterns and relationships within large datasets. Each module operates\nindependently, allowing users to modify or upgrade modules in response to\nevolving research needs or technological advances. Moreover, the platform\nprovides a user-friendly interface via Jupyter Notebook, making it accessible\nto researchers at varying levels of expertise. The versatility and\neffectiveness of XASDAML are exemplified by its application to a copper\ndataset, where it efficiently manages large and complex data, supports both\nsupervised and unsupervised machine learning models, provides comprehensive\nstatistics for structural descriptors, generates spectral plots, and accurately\npredicts coordination numbers and bond lengths. Furthermore, the platform\nstreamlining the integration of XAS with machine learning and lowering the\nbarriers to entry for new users.\n", "date": "2025-02-23", "categories": []}
{"id": "2502.16984", "title": "Active Learning for Conditional Inverse Design with Crystal Generation\n  and Foundation Atomic Models", "abstract": "  Artificial intelligence (AI) is transforming materials science, enabling both\ntheoretical advancements and accelerated materials discovery. Recent progress\nin crystal generation models, which design crystal structures for targeted\nproperties, and foundation atomic models (FAMs), which capture interatomic\ninteractions across the periodic table, has significantly improved inverse\nmaterials design. However, an efficient integration of these two approaches\nremains an open challenge. Here, we present an active learning framework that\ncombines crystal generation models and foundation atomic models to enhance the\naccuracy and efficiency of inverse design. As a case study, we employ Con-CDVAE\nto generate candidate crystal structures and MACE-MP-0 FAM as one of the\nhigh-throughput screeners for bulk modulus evaluation. Through iterative active\nlearning, we demonstrate that Con-CDVAE progressively improves its accuracy in\ngenerating crystals with target properties, highlighting the effectiveness of a\nproperty-driven fine-tuning process. Our framework is general to accommodate\ndifferent crystal generation and foundation atomic models, and establishes a\nscalable approach for AI-driven materials discovery. By bridging generative\nmodeling with atomic-scale simulations, this work paves the way for more\naccurate and efficient inverse materials design.\n", "date": "2025-02-24", "categories": []}
{"id": "2411.09429", "title": "AI-driven inverse design of materials: Past, present and future", "abstract": "  The discovery of advanced materials is the cornerstone of human technological\ndevelopment and progress. The structures of materials and their corresponding\nproperties are essentially the result of a complex interplay of multiple\ndegrees of freedom such as lattice, charge, spin, symmetry, and topology. This\nposes significant challenges for the inverse design methods of materials.\nHumans have long explored new materials through a large number of experiments\nand proposed corresponding theoretical systems to predict new material\nproperties and structures. With the improvement of computational power,\nresearchers have gradually developed various electronic structure calculation\nmethods, such as the density functional theory and high-throughput\ncomputational methods. Recently, the rapid development of artificial\nintelligence technology in the field of computer science has enabled the\neffective characterization of the implicit association between material\nproperties and structures, thus opening up an efficient paradigm for the\ninverse design of functional materials. A significant progress has been made in\ninverse design of materials based on generative and discriminative models,\nattracting widespread attention from researchers. Considering this rapid\ntechnological progress, in this survey, we look back on the latest advancements\nin AI-driven inverse design of materials by introducing the background, key\nfindings, and mainstream technological development routes. In addition, we\nsummarize the remaining issues for future directions. This survey provides the\nlatest overview of AI-driven inverse design of materials, which can serve as a\nuseful resource for researchers.\n", "date": "2024-11-14", "categories": []}
{"id": "2502.14631", "title": "Synergistic Fusion of Multi-Source Knowledge via Evidence Theory for\n  High-Entropy Alloy Discovery", "abstract": "  Discovering novel high-entropy alloys (HEAs) with desirable properties is\nchallenging due to the vast compositional space and complex phase formation\nmechanisms. Efficient exploration of this space requires a strategic approach\nthat integrates heterogeneous knowledge sources. Here, we propose a framework\nthat systematically combines knowledge extracted from computational material\ndatasets with domain knowledge distilled from scientific literature using large\nlanguage models (LLMs). A central feature of this approach is the explicit\nconsideration of element substitutability, identifying chemically similar\nelements that can be interchanged to potentially stabilize desired HEAs.\nDempster-Shafer theory, a mathematical framework for reasoning under\nuncertainty, is employed to model and combine substitutabilities based on\naggregated evidence from multiple sources. The framework predicts the phase\nstability of candidate HEA compositions and is systematically evaluated on both\nquaternary alloy systems, demonstrating superior performance compared to\nbaseline machine learning models and methods reliant on single-source evidence\nin cross-validation experiments. By leveraging multi-source knowledge, the\nframework retains robust predictive power even when key elements are absent\nfrom the training data, underscoring its potential for knowledge transfer and\nextrapolation. Furthermore, the enhanced interpretability of the methodology\noffers insights into the fundamental factors governing HEA formation. Overall,\nthis work provides a promising strategy for accelerating HEA discovery by\nintegrating computational and textual knowledge sources, enabling efficient\nexploration of vast compositional spaces with improved generalization and\ninterpretability.\n", "date": "2025-02-20", "categories": []}
{"id": "2502.13886", "title": "Refining embeddings with fill-tuning: data-efficient generalised\n  performance improvements for materials foundation models", "abstract": "  Pretrained foundation models learn embeddings that can be used for a wide\nrange of downstream tasks. These embeddings optimise general performance, and\nif insufficiently accurate at a specific task the model can be fine-tuned to\nimprove performance. For all current methodologies this operation necessarily\ndegrades performance on all out-of-distribution tasks. In this work we present\n'fill-tuning', a novel methodology to generate datasets for continued\npretraining of foundation models that are not suited to a particular downstream\ntask, but instead aim to correct poor regions of the embedding. We present the\napplication of roughness analysis to latent space topologies and illustrate how\nit can be used to propose data that will be most valuable to improving the\nembedding. We apply fill-tuning to a set of state-of-the-art materials\nfoundation models trained on $O(10^9)$ data points and show model improvement\nof almost 1% in all downstream tasks with the addition of only 100 data points.\nThis method provides a route to the general improvement of foundation models at\nthe computational cost of fine-tuning.\n", "date": "2025-02-19", "categories": []}
{"id": "2502.12548", "title": "Improving the Stability of GNN Force Field Models by Reducing Feature\n  Correlation", "abstract": "  Recently, Graph Neural Network based Force Field (GNNFF) models are widely\nused in Molecular Dynamics (MD) simulation, which is one of the most\ncost-effective means in semiconductor material research. However, even such\nmodels provide high accuracy in energy and force Mean Absolute Error (MAE) over\ntrained (in-distribution) datasets, they often become unstable during long-time\nMD simulation when used for out-of-distribution datasets. In this paper, we\npropose a feature correlation based method for GNNFF models to enhance the\nstability of MD simulation. We reveal the negative relationship between feature\ncorrelation and the stability of GNNFF models, and design a loss function with\na dynamic loss coefficient scheduler to reduce edge feature correlation that\ncan be applied in general GNNFF training. We also propose an empirical metric\nto evaluate the stability in MD simulation. Experiments show our method can\nsignificantly improve stability for GNNFF models especially in\nout-of-distribution data with less than 3% computational overhead. For example,\nwe can ensure the stable MD simulation time from 0.03ps to 10ps for Allegro\nmodel.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.11369", "title": "Physics-Informed Gaussian Process Classification for Constraint-Aware\n  Alloy Design", "abstract": "  Alloy design can be framed as a constraint-satisfaction problem. Building on\nprevious methodologies, we propose equipping Gaussian Process Classifiers\n(GPCs) with physics-informed prior mean functions to model the boundaries of\nfeasible design spaces. Through three case studies, we highlight the utility of\ninformative priors for handling constraints on continuous and categorical\nproperties. (1) Phase Stability: By incorporating CALPHAD predictions as priors\nfor solid-solution phase stability, we enhance model validation using a\npublicly available XRD dataset. (2) Phase Stability Prediction Refinement: We\ndemonstrate an in silico active learning approach to efficiently correct phase\ndiagrams. (3) Continuous Property Thresholds: By embedding priors into\ncontinuous property models, we accelerate the discovery of alloys meeting\nspecific property thresholds via active learning. In each case, integrating\nphysics-based insights into the classification framework substantially improved\nmodel performance, demonstrating an efficient strategy for constraint-aware\nalloy design.\n", "date": "2025-02-17", "categories": []}
{"id": "2502.03146", "title": "Symmetry-Aware Bayesian Flow Networks for Crystal Generation", "abstract": "  The discovery of new crystalline materials is essential to scientific and\ntechnological progress. However, traditional trial-and-error approaches are\ninefficient due to the vast search space. Recent advancements in machine\nlearning have enabled generative models to predict new stable materials by\nincorporating structural symmetries and to condition the generation on desired\nproperties. In this work, we introduce SymmBFN, a novel symmetry-aware Bayesian\nFlow Network (BFN) for crystalline material generation that accurately\nreproduces the distribution of space groups found in experimentally observed\ncrystals. SymmBFN substantially improves efficiency, generating stable\nstructures at least 50 times faster than the next-best method. Furthermore, we\ndemonstrate its capability for property-conditioned generation, enabling the\ndesign of materials with tailored properties. Our findings establish BFNs as an\neffective tool for accelerating the discovery of crystalline materials.\n", "date": "2025-02-05", "categories": []}
{"id": "2502.09970", "title": "Universal Machine Learning Interatomic Potentials are Ready for Solid\n  Ion Conductors", "abstract": "  With the rapid development of energy storage technology, high-performance\nsolid-state electrolytes (SSEs) have become critical for next-generation\nlithium-ion batteries. These materials require high ionic conductivity,\nexcellent electrochemical stability, and good mechanical properties to meet the\ndemands of electric vehicles and portable electronics. However, traditional\nmethods like density functional theory (DFT) and empirical force fields face\nchallenges such as high computational costs, poor scalability, and limited\naccuracy across material systems. Universal machine learning interatomic\npotentials (uMLIPs) offer a promising solution with their efficiency and\nnear-DFT-level accuracy.This study systematically evaluates six advanced uMLIP\nmodels (MatterSim, MACE, SevenNet, CHGNet, M3GNet, and ORBFF) in terms of\nenergy, forces, thermodynamic properties, elastic moduli, and lithium-ion\ndiffusion behavior. The results show that MatterSim outperforms others in\nnearly all metrics, particularly in complex material systems, demonstrating\nsuperior accuracy and physical consistency. Other models exhibit significant\ndeviations due to issues like energy inconsistency or insufficient training\ndata coverage.Further analysis reveals that MatterSim achieves excellent\nagreement with reference values in lithium-ion diffusivity calculations,\nespecially at room temperature. Studies on Li3YCl6 and Li6PS5Cl uncover how\ncrystal structure, anion disorder levels, and Na/Li arrangements influence\nionic conductivity. Appropriate S/Cl disorder levels and optimized Na/Li\narrangements enhance diffusion pathway connectivity, improving overall ionic\ntransport performance.\n", "date": "2025-02-14", "categories": []}
{"id": "2502.09423", "title": "Transformer-Enhanced Variational Autoencoder for Crystal Structure\n  Prediction", "abstract": "  Crystal structure forms the foundation for understanding the physical and\nchemical properties of materials. Generative models have emerged as a new\nparadigm in crystal structure prediction(CSP), however, accurately capturing\nkey characteristics of crystal structures, such as periodicity and symmetry,\nremains a significant challenge. In this paper, we propose a\nTransformer-Enhanced Variational Autoencoder for Crystal Structure Prediction\n(TransVAE-CSP), who learns the characteristic distribution space of stable\nmaterials, enabling both the reconstruction and generation of crystal\nstructures. TransVAE-CSP integrates adaptive distance expansion with\nirreducible representation to effectively capture the periodicity and symmetry\nof crystal structures, and the encoder is a transformer network based on an\nequivariant dot product attention mechanism. Experimental results on the\ncarbon_24, perov_5, and mp_20 datasets demonstrate that TransVAE-CSP\noutperforms existing methods in structure reconstruction and generation tasks\nunder various modeling metrics, offering a powerful tool for crystal structure\ndesign and optimization.\n", "date": "2025-02-13", "categories": []}
{"id": "2409.05169", "title": "Stress Predictions in Polycrystal Plasticity using Graph Neural Networks\n  with Subgraph Training", "abstract": "  Numerical modeling of polycrystal plasticity is computationally intensive. We\nemploy Graph Neural Networks (GNN) to predict stresses on complex geometries\nfor polycrystal plasticity from Finite Element Method (FEM) simulations. We\npresent a novel message-passing GNN that encodes nodal strain and edge\ndistances between FEM mesh cells, and aggregates to obtain embeddings and\ncombines the decoded embeddings with the nodal strains to predict stress\ntensors on graph nodes. The GNN is trained on subgraphs generated from FEM mesh\ngraphs, in which the mesh cells are converted to nodes and edges are created\nbetween adjacent cells. We apply the trained GNN to periodic polycrystals with\ncomplex geometries and learn the strain-stress maps based on crystal plasticity\ntheory. The GNN is accurately trained on FEM graphs, in which the $R^2$ for\nboth training and testing sets are larger than 0.99. The proposed GNN approach\nspeeds up more than 150 times compared with FEM on stress predictions. We also\napply the trained GNN to unseen simulations for validations and the GNN\ngeneralizes well with an overall $R^2$ of 0.992. The GNN accurately predicts\nthe von Mises stress on polycrystals. The proposed model does not overfit and\ngeneralizes well beyond the training data, as the error distributions\ndemonstrate. This work outlooks surrogating crystal plasticity simulations\nusing graph data.\n", "date": "2024-09-08", "categories": []}
{"id": "2502.07293", "title": "Global Universal Scaling and Ultra-Small Parameterization in Machine\n  Learning Interatomic Potentials with Super-Linearity", "abstract": "  Using machine learning (ML) to construct interatomic interactions and thus\npotential energy surface (PES) has become a common strategy for materials\ndesign and simulations. However, those current models of machine learning\ninteratomic potential (MLIP) provide no relevant physical constrains, and thus\nmay owe intrinsic out-of-domain difficulty which underlies the challenges of\nmodel generalizability and physical scalability. Here, by incorporating\nphysics-informed Universal-Scaling law and nonlinearity-embedded interaction\nfunction, we develop a Super-linear MLIP with both Ultra-Small parameterization\nand greatly expanded expressive capability, named SUS2-MLIP. Due to the global\nscaling rooting in universal equation of state (UEOS), SUS2-MLIP not only has\nsignificantly-reduced parameters by decoupling the element space from\ncoordinate space, but also naturally outcomes the out-of-domain difficulty and\nendows the potentials with inherent generalizability and scalability even with\nrelatively small training dataset. The nonlinearity-enbeding transformation for\ninteraction function expands the expressive capability and make the potentials\nsuper-linear. The SUS2-MLIP outperforms the state-of-the-art MLIP models with\nits exceptional computational efficiency especially for multiple-element\nmaterials and physical scalability in property prediction. This work not only\npresents a highly-efficient universal MLIP model but also sheds light on\nincorporating physical constraints into artificial-intelligence-aided materials\nsimulation.\n", "date": "2025-02-11", "categories": []}
{"id": "2401.03862", "title": "End-to-End Crystal Structure Prediction from Powder X-Ray Diffraction", "abstract": "  Powder X-ray diffraction (PXRD) is a prevalent technique in materials\ncharacterization. While the analysis of PXRD often requires extensive human\nmanual intervention, and most automated method only achieved at coarse-grained\nlevel. The more difficult and important task of fine-grained crystal structure\nprediction from PXRD remains unaddressed. This study introduces XtalNet, the\nfirst equivariant deep generative model for end-to-end crystal structure\nprediction from PXRD. Unlike previous crystal structure prediction methods that\nrely solely on composition, XtalNet leverages PXRD as an additional condition,\neliminating ambiguity and enabling the generation of complex organic structures\nwith up to 400 atoms in the unit cell. XtalNet comprises two modules: a\nContrastive PXRD-Crystal Pretraining (CPCP) module that aligns PXRD space with\ncrystal structure space, and a Conditional Crystal Structure Generation (CCSG)\nmodule that generates candidate crystal structures conditioned on PXRD\npatterns. Evaluation on two MOF datasets (hMOF-100 and hMOF-400) demonstrates\nXtalNet's effectiveness. XtalNet achieves a top-10 Match Rate of 90.2% and 79%\nfor hMOF-100 and hMOF-400 in conditional crystal structure prediction task,\nrespectively. XtalNet enables the direct prediction of crystal structures from\nexperimental measurements, eliminating the need for manual intervention and\nexternal databases. This opens up new possibilities for automated crystal\nstructure determination and the accelerated discovery of novel materials.\n", "date": "2024-01-08", "categories": []}
{"id": "2502.04289", "title": "Retro-Rank-In: A Ranking-Based Approach for Inorganic Materials\n  Synthesis Planning", "abstract": "  Retrosynthesis strategically plans the synthesis of a chemical target\ncompound from simpler, readily available precursor compounds. This process is\ncritical for synthesizing novel inorganic materials, yet traditional methods in\ninorganic chemistry continue to rely on trial-and-error experimentation.\nEmerging machine-learning approaches struggle to generalize to entirely new\nreactions due to their reliance on known precursors, as they frame\nretrosynthesis as a multi-label classification task. To address these\nlimitations, we propose Retro-Rank-In, a novel framework that reformulates the\nretrosynthesis problem by embedding target and precursor materials into a\nshared latent space and learning a pairwise ranker on a bipartite graph of\ninorganic compounds. We evaluate Retro-Rank-In's generalizability on\nchallenging retrosynthesis dataset splits designed to mitigate data duplicates\nand overlaps. For instance, for Cr2AlB2, it correctly predicts the verified\nprecursor pair CrB + Al despite never seeing them in training, a capability\nabsent in prior work. Extensive experiments show that Retro-Rank-In sets a new\nstate-of-the-art, particularly in out-of-distribution generalization and\ncandidate set ranking, offering a powerful tool for accelerating inorganic\nmaterial synthesis.\n", "date": "2025-02-06", "categories": []}
{"id": "2502.05735", "title": "Towards Autonomous Experimentation: Bayesian Optimization over Problem\n  Formulation Space for Accelerated Alloy Development", "abstract": "  Accelerated discovery in materials science demands autonomous systems capable\nof dynamically formulating and solving design problems. In this work, we\nintroduce a novel framework that leverages Bayesian optimization over a problem\nformulation space to identify optimal design formulations in line with\ndecision-maker preferences. By mapping various design scenarios to a multi\nattribute utility function, our approach enables the system to balance\nconflicting objectives such as ductility, yield strength, density, and\nsolidification range without requiring an exact problem definition at the\noutset. We demonstrate the efficacy of our method through an in silico case\nstudy on a Mo-Nb-Ti-V-W alloy system targeted for gas turbine engine blade\napplications. The framework converges on a sweet spot that satisfies critical\nperformance thresholds, illustrating that integrating problem formulation\ndiscovery into the autonomous design loop can significantly streamline the\nexperimental process. Future work will incorporate human feedback to further\nenhance the adaptability of the system in real-world experimental settings.\n", "date": "2025-02-09", "categories": []}
{"id": "2502.05970", "title": "Known Unknowns: Out-of-Distribution Property Prediction in Materials and\n  Molecules", "abstract": "  Discovery of high-performance materials and molecules requires identifying\nextremes with property values that fall outside the known distribution.\nTherefore, the ability to extrapolate to out-of-distribution (OOD) property\nvalues is critical for both solid-state materials and molecular design. Our\nobjective is to train predictor models that extrapolate zero-shot to higher\nranges than in the training data, given the chemical compositions of solids or\nmolecular graphs and their property values. We propose using a transductive\napproach to OOD property prediction, achieving improvements in prediction\naccuracy. In particular, the True Positive Rate (TPR) of OOD classification of\nmaterials and molecules improved by 3x and 2.5x, respectively, and precision\nimproved by 2x and 1.5x compared to non-transductive baselines. Our method\nleverages analogical input-target relations in the training and test sets,\nenabling generalization beyond the training target support, and can be applied\nto any other material and molecular tasks.\n", "date": "2025-02-09", "categories": []}
{"id": "2403.15073", "title": "Broadening the Scope of Neural Network Potentials through Direct\n  Inclusion of Additional Molecular Attributes", "abstract": "  Most state-of-the-art neural network potentials do not account for molecular\nattributes other than atomic numbers and positions, which limits its range of\napplicability by design. In this work, we demonstrate the importance of\nincluding additional electronic attributes in neural network potential\nrepresentations with a minimal architectural change to TensorNet, a\nstate-of-the-art equivariant model based on Cartesian rank-2 tensor\nrepresentations. By performing experiments on both custom-made and public\nbenchmarking datasets, we show that this modification resolves the input\ndegeneracy issues stemming from the use of atomic numbers and positions alone,\nwhile enhancing the model's predictive accuracy across diverse chemical systems\nwith different charge or spin states. This is accomplished without tailored\nstrategies or the inclusion of physics-based energy terms, while maintaining\nefficiency and accuracy. These findings should furthermore encourage\nresearchers to train and use models incorporating these additional\nrepresentations.\n", "date": "2024-03-22", "categories": []}
{"id": "2406.17119", "title": "Accelerating Phase Field Simulations Through a Hybrid Adaptive Fourier\n  Neural Operator with U-Net Backbone", "abstract": "  Prolonged contact between a corrosive liquid and metal alloys can cause\nprogressive dealloying. For such liquid-metal dealloying (LMD) process, phase\nfield models have been developed. However, the governing equations often\ninvolve coupled non-linear partial differential equations (PDE), which are\nchallenging to solve numerically. In particular, stiffness in the PDEs requires\nan extremely small time steps (e.g. $10^{-12}$ or smaller). This computational\nbottleneck is especially problematic when running LMD simulation until a late\ntime horizon is required. This motivates the development of surrogate models\ncapable of leaping forward in time, by skipping several consecutive time steps\nat-once. In this paper, we propose U-Shaped Adaptive Fourier Neural Operators\n(U-AFNO), a machine learning (ML) model inspired by recent advances in neural\noperator learning. U-AFNO employs U-Nets for extracting and reconstructing\nlocal features within the physical fields, and passes the latent space through\na vision transformer (ViT) implemented in the Fourier space (AFNO). We use\nU-AFNOs to learn the dynamics mapping the field at a current time step into a\nlater time step. We also identify global quantities of interest (QoI)\ndescribing the corrosion process (e.g. the deformation of the liquid-metal\ninterface) and show that our proposed U-AFNO model is able to accurately\npredict the field dynamics, in-spite of the chaotic nature of LMD. Our model\nreproduces the key micro-structure statistics and QoIs with a level of accuracy\non-par with the high-fidelity numerical solver. We also investigate the\nopportunity of using hybrid simulations, in which we alternate forward leap in\ntime using the U-AFNO with high-fidelity time stepping. We demonstrate that\nwhile advantageous for some surrogate model design choices, our proposed U-AFNO\nmodel in fully auto-regressive settings consistently outperforms hybrid\nschemes.\n", "date": "2024-06-24", "categories": []}
{"id": "2502.02607", "title": "MIND: Microstructure INverse Design with Generative Hybrid Neural\n  Representation", "abstract": "  The inverse design of microstructures plays a pivotal role in optimizing\nmetamaterials with specific, targeted physical properties. While traditional\nforward design methods are constrained by their inability to explore the vast\ncombinatorial design space, inverse design offers a compelling alternative by\ndirectly generating structures that fulfill predefined performance criteria.\nHowever, achieving precise control over both geometry and material properties\nremains a significant challenge due to their intricate interdependence.\nExisting approaches, which typically rely on voxel or parametric\nrepresentations, often limit design flexibility and structural diversity. In\nthis work, we present a novel generative model that integrates latent diffusion\nwith Holoplane, an advanced hybrid neural representation that simultaneously\nencodes both geometric and physical properties. This combination ensures\nsuperior alignment between geometry and properties. Our approach generalizes\nacross multiple microstructure classes, enabling the generation of diverse,\ntileable microstructures with significantly improved property accuracy and\nenhanced control over geometric validity, surpassing the performance of\nexisting methods. We introduce a multi-class dataset encompassing a variety of\ngeometric morphologies, including truss, shell, tube, and plate structures, to\ntrain and validate our model. Experimental results demonstrate the model's\nability to generate microstructures that meet target properties, maintain\ngeometric validity, and integrate seamlessly into complex assemblies.\nAdditionally, we explore the potential of our framework through the generation\nof new microstructures, cross-class interpolation, and the infilling of\nheterogeneous microstructures. The dataset and source code will be open-sourced\nupon publication.\n", "date": "2025-02-01", "categories": []}
{"id": "2502.02905", "title": "AI-driven materials design: a mini-review", "abstract": "  Materials design is an important component of modern science and technology,\nyet traditional approaches rely heavily on trial-and-error and can be\ninefficient. Computational techniques, enhanced by modern artificial\nintelligence (AI), have greatly accelerated the design of new materials. Among\nthese approaches, inverse design has shown great promise in designing materials\nthat meet specific property requirements. In this mini-review, we summarize key\ncomputational advancements for materials design over the past few decades. We\nfollow the evolution of relevant materials design techniques, from\nhigh-throughput forward machine learning (ML) methods and evolutionary\nalgorithms, to advanced AI strategies like reinforcement learning (RL) and deep\ngenerative models. We highlight the paradigm shift from conventional screening\napproaches to inverse generation driven by deep generative models. Finally, we\ndiscuss current challenges and future perspectives of materials inverse design.\nThis review may serve as a brief guide to the approaches, progress, and outlook\nof designing future functional materials with technological relevance.\n", "date": "2025-02-05", "categories": []}
{"id": "2502.02016", "title": "A Periodic Bayesian Flow for Material Generation", "abstract": "  Generative modeling of crystal data distribution is an important yet\nchallenging task due to the unique periodic physical symmetry of crystals.\nDiffusion-based methods have shown early promise in modeling crystal\ndistribution. More recently, Bayesian Flow Networks were introduced to\naggregate noisy latent variables, resulting in a variance-reduced parameter\nspace that has been shown to be advantageous for modeling Euclidean data\ndistributions with structural constraints (Song et al., 2023). Inspired by\nthis, we seek to unlock its potential for modeling variables located in\nnon-Euclidean manifolds e.g. those within crystal structures, by overcoming\nchallenging theoretical issues. We introduce CrysBFN, a novel crystal\ngeneration method by proposing a periodic Bayesian flow, which essentially\ndiffers from the original Gaussian-based BFN by exhibiting non-monotonic\nentropy dynamics. To successfully realize the concept of periodic Bayesian\nflow, CrysBFN integrates a new entropy conditioning mechanism and empirically\ndemonstrates its significance compared to time-conditioning. Extensive\nexperiments over both crystal ab initio generation and crystal structure\nprediction tasks demonstrate the superiority of CrysBFN, which consistently\nachieves new state-of-the-art on all benchmarks. Surprisingly, we found that\nCrysBFN enjoys a significant improvement in sampling efficiency, e.g., ~100x\nspeedup 10 v.s. 2000 steps network forwards) compared with previous\ndiffusion-based methods on MP-20 dataset. Code is available at\nhttps://github.com/wu-han-lin/CrysBFN.\n", "date": "2025-02-04", "categories": []}
{"id": "2502.02026", "title": "ContinuouSP: Generative Model for Crystal Structure Prediction with\n  Invariance and Continuity", "abstract": "  The discovery of new materials using crystal structure prediction (CSP) based\non generative machine learning models has become a significant research topic\nin recent years. In this paper, we study invariance and continuity in the\ngenerative machine learning for CSP. We propose a new model, called\nContinuouSP, which effectively handles symmetry and periodicity in crystals. We\nclearly formulate the invariance and the continuity, and construct a model\nbased on the energy-based model. Our preliminary evaluation demonstrates the\neffectiveness of this model with the CSP task.\n", "date": "2025-02-04", "categories": []}
{"id": "2502.02457", "title": "Orientation-aware interaction-based deep material network in\n  polycrystalline materials modeling", "abstract": "  Multiscale simulations are indispensable for connecting microstructural\nfeatures to the macroscopic behavior of polycrystalline materials, but their\nhigh computational demands limit their practicality. Deep material networks\n(DMNs) have been proposed as efficient surrogate models, yet they fall short of\ncapturing texture evolution. To address this limitation, we propose the\norientation-aware interaction-based deep material network (ODMN), which\nincorporates an orientation-aware mechanism and an interaction mechanism\ngrounded in the Hill-Mandel principle. The orientation-aware mechanism learns\nthe crystallographic textures, while the interaction mechanism captures\nstress-equilibrium directions among representative volume element (RVE)\nsubregions, offering insight into internal microstructural mechanics. Notably,\nODMN requires only linear elastic data for training yet generalizes effectively\nto complex nonlinear and anisotropic responses. Our results show that ODMN\naccurately predicts both mechanical responses and texture evolution under\ncomplex plastic deformation, thus expanding the applicability of DMNs to\npolycrystalline materials. By balancing computational efficiency with\npredictive fidelity, ODMN provides a robust framework for multiscale\nsimulations of polycrystalline materials.\n", "date": "2025-02-04", "categories": []}
{"id": "2501.18189", "title": "Neural Network Modeling of Microstructure Complexity Using Digital\n  Libraries", "abstract": "  Microstructure evolution in matter is often modeled numerically using field\nor level-set solvers, mirroring the dual representation of spatiotemporal\ncomplexity in terms of pixel or voxel data, and geometrical forms in vector\ngraphics. Motivated by this analog, as well as the structural and event-driven\nnature of artificial and spiking neural networks, respectively, we evaluate\ntheir performance in learning and predicting fatigue crack growth and Turing\npattern development. Predictions are made based on digital libraries\nconstructed from computer simulations, which can be replaced by experimental\ndata to lift the mathematical overconstraints of physics. Our assessment\nsuggests that the leaky integrate-and-fire neuron model offers superior\npredictive accuracy with fewer parameters and less memory usage, alleviating\nthe accuracy-cost tradeoff in contrast to the common practices in computer\nvision tasks. Examination of network architectures shows that these benefits\narise from its reduced weight range and sparser connections. The study\nhighlights the capability of event-driven models in tackling problems with\nevolutionary bulk-phase and interface behaviors using the digital library\napproach.\n", "date": "2025-01-30", "categories": []}
{"id": "2406.02457", "title": "Machine learning Hubbard parameters with equivariant neural networks", "abstract": "  Density-functional theory with extended Hubbard functionals (DFT+$U$+$V$)\nprovides a robust framework to accurately describe complex materials containing\ntransition-metal or rare-earth elements. It does so by mitigating\nself-interaction errors inherent to semi-local functionals which are\nparticularly pronounced in systems with partially-filled d and f electronic\nstates. However, achieving accuracy in this approach hinges upon the accurate\ndetermination of the on-site $U$ and inter-site $V$ Hubbard parameters. In\npractice, these are obtained either by semi-empirical tuning, requiring prior\nknowledge, or, more correctly, by using predictive but expensive\nfirst-principles calculations. Here, we present a machine learning model based\non equivariant neural networks which uses atomic occupation matrices as\ndescriptors, directly capturing the electronic structure, local chemical\nenvironment, and oxidation states of the system at hand. We target here the\nprediction of Hubbard parameters computed self-consistently with iterative\nlinear-response calculations, as implemented in density-functional perturbation\ntheory (DFPT), and structural relaxations. Remarkably, when trained on data\nfrom 12 materials spanning various crystal structures and compositions, our\nmodel achieves mean absolute relative errors of 3% and 5% for Hubbard $U$ and\n$V$ parameters, respectively. By circumventing computationally expensive DFT or\nDFPT self-consistent protocols, our model significantly expedites the\nprediction of Hubbard parameters with negligible computational overhead, while\napproaching the accuracy of DFPT. Moreover, owing to its robust\ntransferability, the model facilitates accelerated materials discovery and\ndesign via high-throughput calculations, with relevance for various\ntechnological applications.\n", "date": "2024-06-04", "categories": []}
{"id": "2501.16358", "title": "The OpenLAM Challenges", "abstract": "  Inspired by the success of Large Language Models (LLMs), the development of\nLarge Atom Models (LAMs) has gained significant momentum in scientific\ncomputation. Since 2022, the Deep Potential team has been actively pretraining\nLAMs and launched the OpenLAM Initiative to develop an open-source foundation\nmodel spanning the periodic table. A core objective is establishing\ncomprehensive benchmarks for reliable LAM evaluation, addressing limitations in\nexisting datasets. As a first step, the LAM Crystal Philately competition has\ncollected over 19.8 million valid structures, including 1 million on the\nOpenLAM convex hull, driving advancements in generative modeling and materials\nscience applications.\n", "date": "2025-01-20", "categories": []}
{"id": "2402.03753", "title": "Enhanced sampling of robust molecular datasets with uncertainty-based\n  collective variables", "abstract": "  Generating a data set that is representative of the accessible configuration\nspace of a molecular system is crucial for the robustness of machine learned\ninteratomic potentials (MLIP). However, the complexity of molecular systems,\ncharacterized by intricate potential energy surfaces (PESs) with numerous local\nminima and energy barriers, presents a significant challenge. Traditional\nmethods of data generation, such as random sampling or exhaustive exploration,\nare either intractable or may not capture rare, but highly informative\nconfigurations. In this study, we propose a method that leverages uncertainty\nas the collective variable (CV) to guide the acquisition of chemically-relevant\ndata points, focusing on regions of the configuration space where ML model\npredictions are most uncertain. This approach employs a Gaussian Mixture\nModel-based uncertainty metric from a single model as the CV for biased\nmolecular dynamics simulations. The effectiveness of our approach in overcoming\nenergy barriers and exploring unseen energy minima, thereby enhancing the data\nset in an active learning framework, is demonstrated on the alanine dipeptide\nbenchmark system.\n", "date": "2024-02-06", "categories": []}
{"id": "2403.07526", "title": "Discovering High-Strength Alloys via Physics-Transfer Learning", "abstract": "  Predicting the strength of materials requires considering various length and\ntime scales, striking a balance between accuracy and efficiency. Peierls stress\nmeasures material strength by evaluating dislocation resistance to plastic\nflow, reliant on elastic lattice responses and crystal slip energy landscape.\nComputational challenges due to the non-local and non-equilibrium nature of\ndislocations prohibit Peierls stress evaluation from state-of-the-art material\ndatabases. We propose a data-driven framework that leverages neural networks\ntrained on force field simulations to understand crystal plasticity physics,\npredicting Peierls stress from material parameters derived via density\nfunctional theory computations, which are otherwise computationally intensive\nfor direct dislocation modeling. This physics transfer approach successfully\nscreen the strength of metallic alloys from a limited number of single-point\ncalculations with chemical accuracy. Guided by these predictions, we fabricate\nhigh-strength binary alloys previously unexplored, utilizing high-throughput\nion beam deposition techniques. The framework extends to problems facing the\naccuracy-performance dilemma in general by harnessing the hierarchy of physics\nof multiscale models in materials sciences.\n", "date": "2024-03-12", "categories": []}
{"id": "2501.14813", "title": "Dissertation Machine Learning in Materials Science -- A case study in\n  Carbon Nanotube field effect transistors", "abstract": "  In this thesis, I explored the use of several machine learning techniques,\nincluding neural networks, simulation-based inference, and generative flow\nnetworks, on predicting CNTFETs performance, probing the conductivity\nproperties of CNT network, and generating CNTFETs processing information for\ntarget performance.\n", "date": "2025-01-19", "categories": []}
{"id": "2501.15057", "title": "Predictive Modeling and Uncertainty Quantification of Fatigue Life in\n  Metal Alloys using Machine Learning", "abstract": "  Recent advancements in machine learning-based methods have demonstrated great\npotential for improved property prediction in material science. However,\nreliable estimation of the confidence intervals for the predicted values\nremains a challenge, due to the inherent complexities in material modeling.\nThis study introduces a novel approach for uncertainty quantification in\nfatigue life prediction of metal materials based on integrating knowledge from\nphysics-based fatigue life models and machine learning models. The proposed\napproach employs physics-based input features estimated using the Basquin\nfatigue model to augment the experimentally collected data of fatigue life.\nFurthermore, a physics-informed loss function that enforces boundary\nconstraints for the estimated fatigue life of considered materials is\nintroduced for the neural network models. Experimental validation on datasets\ncomprising collected data from fatigue life tests for Titanium alloys and\nCarbon steel alloys demonstrates the effectiveness of the proposed approach.\nThe synergy between physics-based models and data-driven models enhances the\nconsistency in predicted values and improves uncertainty interval estimates.\n", "date": "2025-01-25", "categories": []}
{"id": "2501.15838", "title": "CrySPAI: A new Crystal Structure Prediction Software Based on Artificial\n  Intelligence", "abstract": "  Crystal structure predictions based on the combination of first-principles\ncalculations and machine learning have achieved significant success in\nmaterials science. However, most of these approaches are limited to predicting\nspecific systems, which hinders their application to unknown or unexplored\ndomains. In this paper, we present CrySPAI, a crystal structure prediction\npackage developed using artificial intelligence (AI) to predict energetically\nstable crystal structures of inorganic materials given their chemical\ncompositions. The software consists of three key modules, an evolutionary\noptimization algorithm (EOA) that searches for all possible crystal structure\nconfigurations, density functional theory (DFT) that provides the accurate\nenergy values for these structures, and a deep neural network (DNN) that learns\nthe relationship between crystal structures and their corresponding energies.\nTo optimize the process across these modules, a distributed framework is\nimplemented to parallelize tasks, and an automated workflow has been integrated\ninto CrySPAI for seamless execution. This paper reports the development and\nimplementation of AI AI-based CrySPAI Crystal Prediction Software tool and its\nunique features.\n", "date": "2025-01-27", "categories": []}
{"id": "2410.05604", "title": "Accelerating the discovery of low-energy structure configurations: a\n  computational approach that integrates first-principles calculations, Monte\n  Carlo sampling, and Machine Learning", "abstract": "  Finding Minimum Energy Configurations (MECs) is essential in fields such as\nphysics, chemistry, and materials science, as they represent the most stable\nstates of the systems. In particular, identifying such MECs in multi-component\nalloys considered candidate PFMs is key because it determines the most stable\narrangement of atoms within the alloy, directly influencing its phase\nstability, structural integrity, and thermo-mechanical properties. However,\nsince the search space grows exponentially with the number of atoms considered,\nobtaining such MECs using computationally expensive first-principles DFT\ncalculations often results in a cumbersome task. To escape the above compromise\nbetween physical fidelity and computational efficiency, we have developed a\nnovel physics-based data-driven approach that combines Monte Carlo sampling,\nfirst-principles DFT calculations, and Machine Learning to accelerate the\ndiscovery of MECs in multi-component alloys. More specifically, we have\nleveraged well-established Cluster Expansion (CE) techniques with Local Outlier\nFactor models to establish strategies that enhance the reliability of the CE\nmethod. In this work, we demonstrated the capabilities of the proposed approach\nfor the particular case of a tungsten-based quaternary high-entropy alloy.\nHowever, the method is applicable to other types of alloys and enables a wide\nrange of applications.\n", "date": "2024-10-08", "categories": []}
{"id": "2501.12558", "title": "Structural and mechanical properties of W-Cu compounds characterized by\n  a neural-network-based potential", "abstract": "  Tungsten-copper (W-Cu) compounds are widely utilized in various industrial\nfields due to their exceptional mechanical properties. In this study, we have\ndeveloped a neural-network-based deep potential (DP) model that covers a wide\nrange of temperatures, ranging from 0 to 3,000 K, and pressures, varying from 0\nto 10 GPa. This study presents a model trained using density functional theory\ndata for full concentration CuxW100-x compounds. Through this model, we\nsystematically investigate the structural and mechanical properties of W-Cu\nalloys and have the following findings. First, the bulk modulus (B) and Young's\nmodulus (E) of W-Cu alloys exhibit a linear decline as the Cu content\nincreases, indicating a softening trend in the CuxW100-x compounds as the Cu\nconcentration rises. Second, a higher Cu content results in higher critical\nstrain and lower critical stress for these compounds. A brittle-to-ductile\ntransition in the deformation mode predicted is predicted at around 37.5 at. %\nCu content. Third, tensile loading tests in the W-Cu gradient structure reveal\nthat Cu-poor region serves as a barrier, hindering shear band propagation while\npromoting new shear band formation in the Cu-rich region. The above results\nfrom the DP model are anticipated to aid in exploring the physical mechanisms\nunderlying the complex phenomena of W-Cu systems and contribute to the\nadvancement of methodologies for materials simulation.\n", "date": "2025-01-22", "categories": []}
{"id": "2401.16611", "title": "Accelerating superconductor discovery through tempered deep learning of\n  the electron-phonon spectral function", "abstract": "  Integrating deep learning with the search for new electron-phonon\nsuperconductors represents a burgeoning field of research, where the primary\nchallenge lies in the computational intensity of calculating the\nelectron-phonon spectral function, $\\alpha^2F(\\omega)$, the essential\ningredient of Midgal-Eliashberg theory of superconductivity. To overcome this\nchallenge, we adopt a two-step approach. First, we compute $\\alpha^2F(\\omega)$\nfor 818 dynamically stable materials. We then train a deep-learning model to\npredict $\\alpha^2F(\\omega)$, using an unconventional training strategy to\ntemper the model's overfitting, enhancing predictions. Specifically, we train a\nBootstrapped Ensemble of Tempered Equivariant graph neural NETworks (BETE-NET),\nobtaining an MAE of 0.21, 45 K, and 43 K for the Eliashberg moments derived\nfrom $\\alpha^2F(\\omega)$: $\\lambda$, $\\omega_{\\log}$, and $\\omega_{2}$,\nrespectively, yielding an MAE of 2.5 K for the critical temperature, $T_c$.\nFurther, we incorporate domain knowledge of the site-projected phonon density\nof states to impose inductive bias into the model's node attributes and enhance\npredictions. This methodological innovation decreases the MAE to 0.18, 29 K,\nand 28 K, respectively, yielding an MAE of 2.1 K for $T_c$. We illustrate the\npractical application of our model in high-throughput screening for high-$T_c$\nmaterials. The model demonstrates an average precision nearly five times higher\nthan random screening, highlighting the potential of ML in accelerating\nsuperconductor discovery. BETE-NET accelerates the search for high-$T_c$\nsuperconductors while setting a precedent for applying ML in materials\ndiscovery, particularly when data is limited.\n", "date": "2024-01-29", "categories": []}
{"id": "2501.10651", "title": "MOFA: Discovering Materials for Carbon Capture with a GenAI- and\n  Simulation-Based Workflow", "abstract": "  We present MOFA, an open-source generative AI (GenAI) plus simulation\nworkflow for high-throughput generation of metal-organic frameworks (MOFs) on\nlarge-scale high-performance computing (HPC) systems. MOFA addresses key\nchallenges in integrating GPU-accelerated computing for GPU-intensive GenAI\ntasks, including distributed training and inference, alongside CPU- and\nGPU-optimized tasks for screening and filtering AI-generated MOFs using\nmolecular dynamics, density functional theory, and Monte Carlo simulations.\nThese heterogeneous tasks are unified within an online learning framework that\noptimizes the utilization of available CPU and GPU resources across HPC\nsystems. Performance metrics from a 450-node (14,400 AMD Zen 3 CPUs + 1800\nNVIDIA A100 GPUs) supercomputer run demonstrate that MOFA achieves\nhigh-throughput generation of novel MOF structures, with CO$_2$ adsorption\ncapacities ranking among the top 10 in the hypothetical MOF (hMOF) dataset.\nFurthermore, the production of high-quality MOFs exhibits a linear relationship\nwith the number of nodes utilized. The modular architecture of MOFA will\nfacilitate its integration into other scientific applications that dynamically\ncombine GenAI with large-scale simulations.\n", "date": "2025-01-18", "categories": []}
{"id": "2501.11154", "title": "Modelling of automotive steel fatigue lifetime by machine learning\n  method", "abstract": "  In the current study, the fatigue life of QSTE340TM steel was modelled using\na machine learning method, namely, a neural network. This problem was solved by\na Multi-Layer Perceptron (MLP) neural network with a 3-75-1 architecture, which\nallows the prediction of the crack length based on the number of load cycles N,\nthe stress ratio R, and the overload ratio Rol. The proposed model showed high\naccuracy, with mean absolute percentage error (MAPE) ranging from 0.02% to\n4.59% for different R and Rol. The neural network effectively reveals the\nnonlinear relationships between input parameters and fatigue crack growth,\nproviding reliable predictions for different loading conditions.\n", "date": "2025-01-19", "categories": []}
{"id": "2501.06233", "title": "Mechanics and Design of Metastructured Auxetic Patches with Bio-inspired\n  Materials", "abstract": "  Metastructured auxetic patches, characterized by negative Poisson's ratios,\noffer unique mechanical properties that closely resemble the behavior of human\ntissues and organs. As a result, these patches have gained significant\nattention for their potential applications in organ repair and tissue\nregeneration. This study focuses on neural networks-based computational\nmodeling of auxetic patches with a sinusoidal metastructure fabricated from\nsilk fibroin, a bio-inspired material known for its biocompatibility and\nstrength. The primary objective of this research is to introduce a novel,\ndata-driven framework for patch design. To achieve this, we conducted\nexperimental fabrication and mechanical testing to determine material\nproperties and validate the corresponding finite element models. Finite element\nsimulations were then employed to generate the necessary data, while greedy\nsampling, an active learning technique, was utilized to reduce the\ncomputational cost associated with data labeling. Two neural networks were\ntrained to accurately predict Poisson's ratios and stresses for strains up to\n15\\%, respectively. Both models achieved $R^2$ scores exceeding 0.995, which\nindicates highly reliable predictions. Building on this, we developed a neural\nnetwork-based design model capable of tailoring patch designs to achieve\nspecific mechanical properties. This model demonstrated superior performance\nwhen compared to traditional optimization methods, such as genetic algorithms,\nby providing more efficient and precise design solutions. The proposed\nframework represents a significant advancement in the design of bio-inspired\nmetastructures for medical applications, paving the way for future innovations\nin tissue engineering and regenerative medicine.\n", "date": "2025-01-08", "categories": []}
{"id": "2501.05762", "title": "Development and Comparison of Model-Based and Data-Driven Approaches for\n  the Prediction of the Mechanical Properties of Lattice Structures", "abstract": "  Lattice structures have great potential for several application fields\nranging from medical and tissue engineering to aeronautical one. Their\ndevelopment is further speeded up by the continuing advances in additive\nmanufacturing technologies that allow to overcome issues typical of standard\nprocesses and to propose tailored designs. However, the design of lattice\nstructures is still challenging since their properties are considerably\naffected by numerous factors. The present paper aims to propose, discuss, and\ncompare various modeling approaches to describe, understand, and predict the\ncorrelations between the mechanical properties and the void volume fraction of\ndifferent types of lattice structures fabricated by fused deposition modeling\n3D printing. Particularly, four approaches are proposed: (i) a simplified\nanalytical model; (ii) a semi-empirical model combining analytical equations\nwith experimental correction factors; (iii) an artificial neural network\ntrained on experimental data; (iv) numerical simulations by finite element\nanalyses. The comparison among the various approaches, and with experimental\ndata, allows to identify the performances, advantages, and disadvantages of\neach approach, thus giving important guidelines for choosing the right design\nmethodology based on the needs and available data.\n", "date": "2025-01-10", "categories": []}
{"id": "2501.04172", "title": "Machine Learning for Identifying Grain Boundaries in Scanning Electron\n  Microscopy (SEM) Images of Nanoparticle Superlattices", "abstract": "  Nanoparticle superlattices consisting of ordered arrangements of\nnanoparticles exhibit unique optical, magnetic, and electronic properties\narising from nanoparticle characteristics as well as their collective\nbehaviors. Understanding how processing conditions influence the nanoscale\narrangement and microstructure is critical for engineering materials with\ndesired macroscopic properties. Microstructural features such as grain\nboundaries, lattice defects, and pores significantly affect these properties\nbut are challenging to quantify using traditional manual analyses as they are\nlabor-intensive and prone to errors. In this work, we present a machine\nlearning workflow for automating grain segmentation in scanning electron\nmicroscopy (SEM) images of nanoparticle superlattices. This workflow integrates\nsignal processing techniques, such as Radon transforms, with unsupervised\nlearning methods like agglomerative hierarchical clustering to identify and\nsegment grains without requiring manually annotated data. In the workflow we\ntransform the raw pixel data into explainable numerical representation of\nsuperlattice orientations for clustering. Benchmarking results demonstrate the\nworkflow's robustness against noisy images and edge cases, with a processing\nspeed of four images per minute on standard computational hardware. This\nefficiency makes the workflow scalable to large datasets and makes it a\nvaluable tool for integrating data-driven models into decision-making processes\nfor material design and analysis. For example, one can use this workflow to\nquantify grain size distributions at varying processing conditions like\ntemperature and pressure and using that knowledge adjust processing conditions\nto achieve desired superlattice orientations and grain sizes.\n", "date": "2025-01-07", "categories": []}
{"id": "2501.03254", "title": "Advanced Displacement Magnitude Prediction in Multi-Material Architected\n  Lattice Structure Beams Using Physics Informed Neural Network Architecture", "abstract": "  This paper proposes an innovative method for predicting deformation in\narchitected lattice structures that combines Physics-Informed Neural Networks\n(PINNs) with finite element analysis. A thorough study was carried out on\nFCC-based lattice beams utilizing five different materials (Structural Steel,\nAA6061, AA7075, Ti6Al4V, and Inconel 718) under varied edge loads (1000-10000\nN). The PINN model blends data-driven learning with physics-based limitations\nvia a proprietary loss function, resulting in much higher prediction accuracy\nthan linear regression. PINN outperforms linear regression, achieving greater\nR-square (0.7923 vs 0.5686) and lower error metrics (MSE: 0.00017417 vs\n0.00036187). Among the materials examined, AA6061 had the highest displacement\nsensitivity (0.1014 mm at maximum load), while Inconel718 had better structural\nstability.\n", "date": "2024-12-31", "categories": []}
{"id": "2501.02932", "title": "Predicting band gap from chemical composition: A simple learned model\n  for a material property with atypical statistics", "abstract": "  In solid-state materials science, substantial efforts have been devoted to\nthe calculation and modeling of the electronic band gap. While a wide range of\nab initio methods and machine learning algorithms have been created that can\npredict this quantity, the development of new computational approaches for\nstudying the band gap remains an active area of research. Here we introduce a\nsimple machine learning model for predicting the band gap using only the\nchemical composition of the crystalline material. To motivate the form of the\nmodel, we first analyze the empirical distribution of the band gap, which sheds\nnew light on its atypical statistics. Specifically, our analysis enables us to\nframe band gap prediction as a task of modeling a mixed random variable, and we\ndesign our model accordingly. Our model formulation incorporates thematic ideas\nfrom chemical heuristic models for other material properties in a manner that\nis suited towards the band gap modeling task. The model has exactly one\nparameter corresponding to each element, which is fit using data. To predict\nthe band gap for a given material, the model computes a weighted average of the\nparameters associated with its constituent elements and then takes the maximum\nof this quantity and zero. The model provides heuristic chemical\ninterpretability by intuitively capturing the associations between the band gap\nand individual chemical elements.\n", "date": "2025-01-06", "categories": []}
{"id": "2312.03690", "title": "AI-guided inverse design and discovery of recyclable vitrimeric polymers", "abstract": "  Vitrimer is a new, exciting class of sustainable polymers with the ability to\nheal due to their dynamic covalent adaptive network that can go through\nassociative rearrangement reactions. However, a limited choice of constituent\nmolecules restricts their property space, prohibiting full realization of their\npotential applications. To overcome this challenge, we couple molecular\ndynamics (MD) simulations and a novel graph variational autoencoder (VAE)\nmachine learning model for inverse design of vitrimer chemistries with desired\nglass transition temperature (Tg) and synthesize a novel vitrimer polymer. We\nbuild the first vitrimer dataset of one million chemistries and calculate Tg on\n8,424 of them by high-throughput MD simulations calibrated by a Gaussian\nprocess model. The proposed novel VAE employs dual graph encoders and a latent\ndimension overlapping scheme which allows for individual representation of\nmulti-component vitrimers. By constructing a continuous latent space containing\nnecessary information of vitrimers, we demonstrate high accuracy and efficiency\nof our framework in discovering novel vitrimers with desirable Tg beyond the\ntraining regime. To validate the effectiveness of our framework in experiments,\nwe generate novel vitrimer chemistries with a target Tg = 323 K. By\nincorporating chemical intuition, we synthesize a vitrimer with Tg of 311-317\nK, and experimentally demonstrate healability and flowability. The proposed\nframework offers an exciting tool for polymer chemists to design and synthesize\nnovel, sustainable vitrimer polymers for a facet of applications.\n", "date": "2023-12-06", "categories": []}
{"id": "2412.16270", "title": "MetaScientist: A Human-AI Synergistic Framework for Automated Mechanical\n  Metamaterial Design", "abstract": "  The discovery of novel mechanical metamaterials, whose properties are\ndominated by their engineered structures rather than chemical composition, is a\nknowledge-intensive and resource-demanding process. To accelerate the design of\nnovel metamaterials, we present MetaScientist, a human-in-the-loop system that\nintegrates advanced AI capabilities with expert oversight with two primary\nphases: (1) hypothesis generation, where the system performs complex reasoning\nto generate novel and scientifically sound hypotheses, supported with\ndomain-specific foundation models and inductive biases retrieved from existing\nliterature; (2) 3D structure synthesis, where a 3D structure is synthesized\nwith a novel 3D diffusion model based on the textual hypothesis and refined it\nwith a LLM-based refinement model to achieve better structure properties. At\neach phase, domain experts iteratively validate the system outputs, and provide\nfeedback and supplementary materials to ensure the alignment of the outputs\nwith scientific principles and human preferences. Through extensive evaluation\nfrom human scientists, MetaScientist is able to deliver novel and valid\nmechanical metamaterial designs that have the potential to be highly impactful\nin the metamaterial field.\n", "date": "2024-12-20", "categories": []}
{"id": "2403.09549", "title": "Generalizing Denoising to Non-Equilibrium Structures Improves\n  Equivariant Force Fields", "abstract": "  Understanding the interactions of atoms such as forces in 3D atomistic\nsystems is fundamental to many applications like molecular dynamics and\ncatalyst design. However, simulating these interactions requires\ncompute-intensive ab initio calculations and thus results in limited data for\ntraining neural networks. In this paper, we propose to use denoising\nnon-equilibrium structures (DeNS) as an auxiliary task to better leverage\ntraining data and improve performance. For training with DeNS, we first corrupt\na 3D structure by adding noise to its 3D coordinates and then predict the\nnoise. Different from previous works on denoising, which are limited to\nequilibrium structures, the proposed method generalizes denoising to a much\nlarger set of non-equilibrium structures. The main difference is that a\nnon-equilibrium structure does not correspond to local energy minima and has\nnon-zero forces, and therefore it can have many possible atomic positions\ncompared to an equilibrium structure. This makes denoising non-equilibrium\nstructures an ill-posed problem since the target of denoising is not uniquely\ndefined. Our key insight is to additionally encode the forces of the original\nnon-equilibrium structure to specify which non-equilibrium structure we are\ndenoising. Concretely, given a corrupted non-equilibrium structure and the\nforces of the original one, we predict the non-equilibrium structure satisfying\nthe input forces instead of any arbitrary structures. Since DeNS requires\nencoding forces, DeNS favors equivariant networks, which can easily incorporate\nforces and other higher-order tensors in node embeddings. We study the\neffectiveness of training equivariant networks with DeNS on OC20, OC22 and MD17\ndatasets and demonstrate that DeNS can achieve new state-of-the-art results on\nOC20 and OC22 and significantly improve training efficiency on MD17.\n", "date": "2024-03-14", "categories": []}
{"id": "2408.15165", "title": "Latent Ewald summation for machine learning of long-range interactions", "abstract": "  Machine learning interatomic potentials (MLIPs) often neglect long-range\ninteractions, such as electrostatic and dispersion forces. In this work, we\nintroduce a straightforward and efficient method to account for long-range\ninteractions by learning a latent variable from local atomic descriptors and\napplying an Ewald summation to this variable. We demonstrate that in systems\nincluding charged and polar molecular dimers, bulk water, and water-vapor\ninterface, standard short-ranged MLIPs can lead to unphysical predictions even\nwhen employing message passing. The long-range models effectively eliminate\nthese artifacts, with only about twice the computational cost of short-range\nMLIPs.\n", "date": "2024-08-27", "categories": []}
{"id": "2412.14109", "title": "Machine Learning Co-pilot for Screening of Organic Molecular Additives\n  for Perovskite Solar Cells", "abstract": "  Machine learning (ML) has been extensively employed in planar perovskite\nphotovoltaics to screen effective organic molecular additives, while\nencountering predictive biases for novel materials due to small datasets and\nreliance on predefined descriptors. Present work thus proposes an effective\napproach, Co-Pilot for Perovskite Additive Screener (Co-PAS), an ML-driven\nframework designed to accelerate additive screening for perovskite solar cells\n(PSCs). Co-PAS overcomes predictive biases by integrating the Molecular\nScaffold Classifier (MSC) for scaffold-based pre-screening and utilizing\nJunction Tree Variational Autoencoder (JTVAE) latent vectors to enhance\nmolecular structure representation, thereby enhancing the accuracy of power\nconversion efficiency (PCE) predictions. Leveraging Co-PAS, we integrate domain\nknowledge to screen an extensive dataset of 250,000 molecules from PubChem,\nprioritizing candidates based on predicted PCE values and key molecular\nproperties such as donor number, dipole moment, and hydrogen bond acceptor\ncount. This workflow leads to the identification of several promising\npassivating molecules, including the novel Boc-L-threonine N-hydroxysuccinimide\nester (BTN), which, to our knowledge, has not been explored as an additive in\nPSCs and achieves a device PCE of 25.20%. Our results underscore the potential\nof Co-PAS in advancing additive discovery for high-performance PSCs.\n", "date": "2024-12-18", "categories": []}
{"id": "2412.10405", "title": "Elucidating microstructural influences on fatigue behavior for\n  additively manufactured Hastelloy X using Bayesian-calibrated crystal\n  plasticity model", "abstract": "  Crystal plasticity (CP) modeling is a vital tool for predicting the\nmechanical behavior of materials, but its calibration involves numerous (>8)\nconstitutive parameters, often requiring time-consuming trial-and-error\nmethods. This paper proposes a robust calibration approach using Bayesian\noptimization (BO) to identify optimal CP model parameters under fatigue loading\nconditions. Utilizing cyclic data from additively manufactured Hastelloy X\nspecimens at 500 degree-F, the BO framework, integrated with a Gaussian process\nsurrogate model, significantly reduces the number of required simulations. A\nnovel objective function is developed to match experimental stress-strain data\nacross different strain amplitudes. Results demonstrate that effective CP model\ncalibration is achieved within 75 iterations, with as few as 50 initial\nsimulations. Sensitivity analysis reveals the influence of CP parameters at\nvarious loading points on the stress-strain curve. The results show that the\nstress-strain response is predominantly controlled by parameters related to\nyield, with increased influence from backstress parameters during compressive\nloading. In addition, the effect of introducing twins into the synthetic\nmicrostructure on fatigue behavior is studied, and a relationship between\nmicrostructural features and the fatigue indicator parameter is established.\nResults show that larger diameter grains, which exhibit a higher Schmid factor\nand an average misorientation of approximately 42 degrees +/- 1.67 degree, are\nidentified as probable sites for failure. The proposed optimization framework\ncan be applied to any material system or CP model, streamlining the calibration\nprocess and improving the predictive accuracy of such models.\n", "date": "2024-12-06", "categories": []}
{"id": "2412.10838", "title": "Deep Learning Models for Colloidal Nanocrystal Synthesis", "abstract": "  Colloidal synthesis of nanocrystals usually includes complex chemical\nreactions and multi-step crystallization processes. Despite the great success\nin the past 30 years, it remains challenging to clarify the correlations\nbetween synthetic parameters of chemical reaction and physical properties of\nnanocrystals. Here, we developed a deep learning-based nanocrystal synthesis\nmodel that correlates synthetic parameters with the final size and shape of\ntarget nanocrystals, using a dataset of 3500 recipes covering 348 distinct\nnanocrystal compositions. The size and shape labels were obtained from\ntransmission electron microscope images using a segmentation model trained with\na semi-supervised algorithm on a dataset comprising 1.2 million nanocrystals.\nBy applying the reaction intermediate-based data augmentation method and\nelaborated descriptors, the synthesis model was able to predict nanocrystal's\nsize with a mean absolute error of 1.39 nm, while reaching an 89% average\naccuracy for shape classification. The synthesis model shows knowledge transfer\ncapabilities across different nanocrystals with inputs of new recipes. With\nthat, the influence of chemicals on the final size of nanocrystals was further\nevaluated, revealing the importance order of nanocrystal composition, precursor\nor ligand, and solvent. Overall, the deep learning-based nanocrystal synthesis\nmodel offers a powerful tool to expedite the development of high-quality\nnanocrystals.\n", "date": "2024-12-14", "categories": []}
{"id": "2410.08833", "title": "Symmetry-Constrained Generation of Diverse Low-Bandgap Molecules with\n  Monte Carlo Tree Search", "abstract": "  Organic optoelectronic materials are a promising avenue for next-generation\nelectronic devices due to their solution processability, mechanical\nflexibility, and tunable electronic properties. In particular, near-infrared\n(NIR) sensitive molecules have unique applications in night-vision equipment\nand biomedical imaging. Molecular engineering has played a crucial role in\ndeveloping non-fullerene acceptors (NFAs) such as the Y-series molecules, which\nhave significantly improved the power conversion efficiency (PCE) of solar\ncells and enhanced spectral coverage in the NIR region. However, systematically\ndesigning molecules with targeted optoelectronic properties while ensuring\nsynthetic accessibility remains a challenge. To address this, we leverage\nstructural priors from domain-focused, patent-mined datasets of organic\nelectronic molecules using a symmetry-aware fragment decomposition algorithm\nand a fragment-constrained Monte Carlo Tree Search (MCTS) generator. Our\napproach generates candidates that retain symmetry constraints from the patent\ndataset, while also exhibiting red-shifted absorption, as validated by TD-DFT\ncalculations.\n", "date": "2024-10-11", "categories": []}
{"id": "2308.14920", "title": "Matbench Discovery -- A framework to evaluate machine learning crystal\n  stability predictions", "abstract": "  The rapid adoption of machine learning (ML) in domain sciences necessitates\nbest practices and standardized benchmarking for performance evaluation. We\npresent Matbench Discovery, an evaluation framework for ML energy models,\napplied as pre-filters for high-throughput searches of stable inorganic\ncrystals. This framework addresses the disconnect between thermodynamic\nstability and formation energy, as well as retrospective vs. prospective\nbenchmarking in materials discovery. We release a Python package to support\nmodel submissions and maintain an online leaderboard, offering insights into\nperformance trade-offs. To identify the best-performing ML methodologies for\nmaterials discovery, we benchmarked various approaches, including random\nforests, graph neural networks (GNNs), one-shot predictors, iterative Bayesian\noptimizers, and universal interatomic potentials (UIP). Our initial results\nrank models by test set F1 scores for thermodynamic stability prediction:\nEquiformerV2 + DeNS > Orb > SevenNet > MACE > CHGNet > M3GNet > ALIGNN > MEGNet\n> CGCNN > CGCNN+P > Wrenformer > BOWSR > Voronoi fingerprint random forest.\nUIPs emerge as the top performers, achieving F1 scores of 0.57-0.82 and\ndiscovery acceleration factors (DAF) of up to 6x on the first 10k stable\npredictions compared to random selection. We also identify a misalignment\nbetween regression metrics and task-relevant classification metrics. Accurate\nregressors can yield high false-positive rates near the decision boundary at 0\neV/atom above the convex hull. Our results demonstrate UIPs' ability to\noptimize computational budget allocation for expanding materials databases.\nHowever, their limitations remain underexplored in traditional benchmarks. We\nadvocate for task-based evaluation frameworks, as implemented here, to address\nthese limitations and advance ML-guided materials discovery.\n", "date": "2023-08-28", "categories": []}
{"id": "2312.13289", "title": "Compositional Representation of Polymorphic Crystalline Materials", "abstract": "  Machine learning (ML) has seen promising developments in materials science,\nyet its efficacy largely depends on detailed crystal structural data, which are\noften complex and hard to obtain, limiting their applicability in real-world\nmaterial synthesis processes. An alternative, using compositional descriptors,\noffers a simpler approach by indicating the elemental ratios of compounds\nwithout detailed structural insights. However, accurately representing\nmaterials solely with compositional descriptors presents challenges due to\npolymorphism, where a single composition can correspond to various structural\narrangements, creating ambiguities in its representation. To this end, we\nintroduce PCRL, a novel approach that employs probabilistic modeling of\ncomposition to capture the diverse polymorphs from available structural\ninformation. Extensive evaluations on sixteen datasets demonstrate the\neffectiveness of PCRL in learning compositional representation, and our\nanalysis highlights its potential applicability of PCRL in material discovery.\nThe source code for PCRL is available at https://github.com/Namkyeong/PCRL.\n", "date": "2023-11-17", "categories": []}
{"id": "2412.05947", "title": "Materials-Discovery Workflows Guided by Symbolic Regression: Identifying\n  Acid-Stable Oxides for Electrocatalysis", "abstract": "  The efficiency of active learning (AL) approaches to identify materials with\ndesired properties relies on the knowledge of a few parameters describing the\nproperty. However, these parameters are unknown if the property is governed by\na high intricacy of many atomistic processes. Here, we develop an AL workflow\nbased on the sure-independence screening and sparsifying operator (SISSO)\nsymbolic-regression approach. SISSO identifies the few, key parameters\ncorrelated with a given materials property via analytical expressions, out of\nmany offered primary features. Crucially, we train ensembles of SISSO models in\norder to quantify mean predictions and their uncertainty, enabling the use of\nSISSO in AL. By combining bootstrap sampling to obtain training datasets with\nMonte-Carlo feature dropout, the high prediction errors observed by a single\nSISSO model are improved. Besides, the feature dropout procedure alleviates the\noverconfidence issues observed in the widely used bagging approach. We\ndemonstrate the SISSO-guided AL workflow by identifying acid-stable oxides for\nwater splitting using high-quality DFT-HSE06 calculations. From a pool of 1470\nmaterials, 12 acid-stable materials are identified in only 30 AL iterations.\nThe materials property maps provided by SISSO along with the uncertainty\nestimates reduce the risk of missing promising portions of the materials space\nthat were overlooked in the initial, possibly biased dataset.\n", "date": "2024-12-08", "categories": []}
{"id": "2412.06064", "title": "Implicit Delta Learning of High Fidelity Neural Network Potentials", "abstract": "  Neural network potentials (NNPs) offer a fast and accurate alternative to\nab-initio methods for molecular dynamics (MD) simulations but are hindered by\nthe high cost of training data from high-fidelity Quantum Mechanics (QM)\nmethods. Our work introduces the Implicit Delta Learning (IDLe) method, which\nreduces the need for high-fidelity QM data by leveraging cheaper semi-empirical\nQM computations without compromising NNP accuracy or inference cost. IDLe\nemploys an end-to-end multi-task architecture with fidelity-specific heads that\ndecode energies based on a shared latent representation of the input atomistic\nsystem. In various settings, IDLe achieves the same accuracy as single\nhigh-fidelity baselines while using up to 50x less high-fidelity data. This\nresult could significantly reduce data generation cost and consequently enhance\naccuracy and generalization, and expand chemical coverage for NNPs, advancing\nMD simulations for material science and drug discovery. Additionally, we\nprovide a novel set of 11 million semi-empirical QM calculations to support\nfuture multi-fidelity NNP modeling.\n", "date": "2024-12-08", "categories": []}
{"id": "2405.14776", "title": "Kinetics of orbital ordering in cooperative Jahn-Teller models:\n  Machine-learning enabled large-scale simulations", "abstract": "  We present a scalable machine learning (ML) force-field model for the\nadiabatic dynamics of cooperative Jahn-Teller (JT) systems. Large scale\ndynamical simulations of the JT model also shed light on the orbital ordering\ndynamics in colossal magnetoresistance manganites. The JT effect in these\nmaterials describes the distortion of local oxygen octahedra driven by a\ncoupling to the orbital degrees of freedom of $e_g$ electrons. An effective\nelectron-mediated interaction between the local JT modes leads to a structural\ntransition and the emergence of long-range orbital order at low temperatures.\nAssuming the principle of locality, a deep-learning neural-network model is\ndeveloped to accurately and efficiently predict the electron-induced forces\nthat drive the dynamical evolution of JT phonons. A group-theoretical method is\nutilized to develop a descriptor that incorporates the combined orbital and\nlattice symmetry into the ML model. Large-scale Langevin dynamics simulations,\nenabled by the ML force-field models, are performed to investigate the\ncoarsening dynamics of the composite JT distortion and orbital order after a\nthermal quench. The late-stage coarsening of orbital domains exhibits\npronounced freezing behaviors which are likely related to the unusual\nmorphology of the domain structures. Our work highlights a promising avenue for\nmulti-scale dynamical modeling of correlated electron systems.\n", "date": "2024-05-23", "categories": []}
{"id": "2405.20397", "title": "Explainable Data-driven Modeling of Adsorption Energy in Heterogeneous\n  Catalysis", "abstract": "  The increasing popularity of machine learning (ML) in catalysis has spurred\ninterest in leveraging these techniques to enhance catalyst design. Our study\naims to bridge the gap between physics-based studies and data-driven\nmethodologies by integrating ML techniques with eXplainable AI (XAI).\nSpecifically, we employ two XAI techniques: Post-hoc XAI analysis and Symbolic\nRegression. These techniques help us unravel the correlation between adsorption\nenergy and the properties of the adsorbate-catalyst system. Leveraging a large\ndataset such as the Open Catalyst Dataset (OC20), we employ a combination of\nshallow ML techniques and XAI methodologies. Our investigation involves\nutilizing multiple shallow machine learning techniques to predict adsorption\nenergy, followed by post-hoc analysis for feature importance, inter-feature\ncorrelations, and the influence of various feature values on the prediction of\nadsorption energy. The post-hoc analysis reveals that adsorbate properties\nexert a greater influence than catalyst properties in our dataset. The top five\nfeatures based on higher Shapley values are adsorbate electronegativity, the\nnumber of adsorbate atoms, catalyst electronegativity, effective coordination\nnumber, and the sum of atomic numbers of the adsorbate molecule. There is a\npositive correlation between catalyst and adsorbate electronegativity with the\nprediction of adsorption energy. Additionally, symbolic regression yields\nresults consistent with SHAP analysis. It deduces a mathematical relationship\nindicating that the square of the catalyst electronegativity is directly\nproportional to the adsorption energy. These consistent correlations resemble\nthose derived from physics-based equations in previous research. Our work\nestablishes a robust framework that integrates ML techniques with XAI,\nleveraging large datasets like OC20 to enhance catalyst design through model\nexplainability.\n", "date": "2024-05-30", "categories": []}
{"id": "2404.10746", "title": "Interpolation and differentiation of alchemical degrees of freedom in\n  machine learning interatomic potentials", "abstract": "  Machine learning interatomic potentials (MLIPs) have become a workhorse of\nmodern atomistic simulations, and recently published universal MLIPs,\npre-trained on large datasets, have demonstrated remarkable accuracy and\ngeneralizability. However, the computational cost of MLIPs limits their\napplicability to chemically disordered systems requiring large simulation cells\nor to sample-intensive statistical methods. Here, we report the use of\ncontinuous and differentiable alchemical degrees of freedom in atomistic\nmaterials simulations, exploiting the fact that graph neural network MLIPs\nrepresent discrete elements as real-valued tensors. The proposed method\nintroduces alchemical atoms with corresponding weights into the input graph,\nalongside modifications to the message-passing and readout mechanisms of MLIPs,\nand allows smooth interpolation between the compositional states of materials.\nThe end-to-end differentiability of MLIPs enables efficient calculation of the\ngradient of energy with respect to the compositional weights. With this\nmodification, we propose methodologies for optimizing the composition of solid\nsolutions towards target macroscopic properties, characterizing order and\ndisorder in multicomponent oxides, and conducting alchemical free energy\nsimulations to quantify the free energy of vacancy formation and composition\nchanges. The approach offers an avenue for extending the capabilities of\nuniversal MLIPs in the modeling of compositional disorder and characterizing\nthe phase stability of complex materials systems.\n", "date": "2024-04-16", "categories": []}
{"id": "2411.00177", "title": "LLM4Mat-Bench: Benchmarking Large Language Models for Materials Property\n  Prediction", "abstract": "  Large language models (LLMs) are increasingly being used in materials\nscience. However, little attention has been given to benchmarking and\nstandardized evaluation for LLM-based materials property prediction, which\nhinders progress. We present LLM4Mat-Bench, the largest benchmark to date for\nevaluating the performance of LLMs in predicting the properties of crystalline\nmaterials. LLM4Mat-Bench contains about 1.9M crystal structures in total,\ncollected from 10 publicly available materials data sources, and 45 distinct\nproperties. LLM4Mat-Bench features different input modalities: crystal\ncomposition, CIF, and crystal text description, with 4.7M, 615.5M, and 3.1B\ntokens in total for each modality, respectively. We use LLM4Mat-Bench to\nfine-tune models with different sizes, including LLM-Prop and MatBERT, and\nprovide zero-shot and few-shot prompts to evaluate the property prediction\ncapabilities of LLM-chat-like models, including Llama, Gemma, and Mistral. The\nresults highlight the challenges of general-purpose LLMs in materials science\nand the need for task-specific predictive models and task-specific\ninstruction-tuned LLMs in materials property prediction.\n", "date": "2024-10-31", "categories": []}
{"id": "2411.19617", "title": "Materials Learning Algorithms (MALA): Scalable Machine Learning for\n  Electronic Structure Calculations in Large-Scale Atomistic Simulations", "abstract": "  We present the Materials Learning Algorithms (MALA) package, a scalable\nmachine learning framework designed to accelerate density functional theory\n(DFT) calculations suitable for large-scale atomistic simulations. Using local\ndescriptors of the atomic environment, MALA models efficiently predict key\nelectronic observables, including local density of states, electronic density,\ndensity of states, and total energy. The package integrates data sampling,\nmodel training and scalable inference into a unified library, while ensuring\ncompatibility with standard DFT and molecular dynamics codes. We demonstrate\nMALA's capabilities with examples including boron clusters, aluminum across its\nsolid-liquid phase boundary, and predicting the electronic structure of a\nstacking fault in a large beryllium slab. Scaling analyses reveal MALA's\ncomputational efficiency and identify bottlenecks for future optimization. With\nits ability to model electronic structures at scales far beyond standard DFT,\nMALA is well suited for modeling complex material systems, making it a\nversatile tool for advanced materials research.\n", "date": "2024-11-29", "categories": []}
{"id": "2411.19629", "title": "OpenQDC: Open Quantum Data Commons", "abstract": "  Machine Learning Interatomic Potentials (MLIPs) are a highly promising\nalternative to force-fields for molecular dynamics (MD) simulations, offering\nprecise and rapid energy and force calculations. However, Quantum-Mechanical\n(QM) datasets, crucial for MLIPs, are fragmented across various repositories,\nhindering accessibility and model development. We introduce the openQDC\npackage, consolidating 37 QM datasets from over 250 quantum methods and 400\nmillion geometries into a single, accessible resource. These datasets are\nmeticulously preprocessed, and standardized for MLIP training, covering a wide\nrange of chemical elements and interactions relevant in organic chemistry.\nOpenQDC includes tools for normalization and integration, easily accessible via\nPython. Experiments with well-known architectures like SchNet, TorchMD-Net, and\nDimeNet reveal challenges for those architectures and constitute a leaderboard\nto accelerate benchmarking and guide novel algorithms development. Continuously\nadding datasets to OpenQDC will democratize QM dataset access, foster more\ncollaboration and innovation, enhance MLIP development, and support their\nadoption in the MD field.\n", "date": "2024-11-29", "categories": []}
{"id": "2411.17755", "title": "Deciphering Acoustic Emission with Machine Learning", "abstract": "  Acoustic emission signals have been shown to accompany avalanche-like events\nin materials, such as dislocation avalanches in crystalline solids, collapse of\nvoids in porous matter or domain wall movement in ferroics. The data provided\nby acoustic emission measurements is tremendously rich, but it is rather\nchallenging to precisely connect it to the characteristics of the triggering\navalanche. In our work we propose a machine learning based method with which\none can infer microscopic details of dislocation avalanches in micropillar\ncompression tests from merely acoustic emission data. As it is demonstrated in\nthe paper, this approach is suitable for the prediction of the force-time\nresponse as it can provide outstanding prediction for the temporal location of\navalanches and can also predict the magnitude of individual deformation events.\nVarious descriptors (including frequency dependent and independent ones) are\nutilised in our machine learning approach and their importance in the\nprediction is analysed. The transferability of the method to other specimen\nsizes is also demonstrated and the possible application in more generic\nsettings is discussed.\n", "date": "2024-11-25", "categories": []}
{"id": "2411.18259", "title": "Transfer Learning for Deep Learning-based Prediction of Lattice Thermal\n  Conductivity", "abstract": "  Machine learning promises to accelerate the material discovery by enabling\nhigh-throughput prediction of desirable macro-properties from atomic-level\ndescriptors or structures. However, the limited data available about precise\nvalues of these properties have been a barrier, leading to predictive models\nwith limited precision or the ability to generalize. This is particularly true\nof lattice thermal conductivity (LTC): existing datasets of precise (ab initio,\nDFT-based) computed values are limited to a few dozen materials with little\nvariability. Based on such datasets, we study the impact of transfer learning\non both the precision and generalizability of a deep learning model\n(ParAIsite). We start from an existing model (MEGNet~\\cite{Chen2019}) and show\nthat improvements are obtained by fine-tuning a pre-trained version on\ndifferent tasks. Interestingly, we also show that a much greater improvement is\nobtained when first fine-tuning it on a large datasets of low-quality\napproximations of LTC (based on the AGL model) and then applying a second phase\nof fine-tuning with our high-quality, smaller-scale datasets. The promising\nresults obtained pave the way not only towards a greater ability to explore\nlarge databases in search of low thermal conductivity materials but also to\nmethods enabling increasingly precise predictions in areas where quality data\nare rare.\n", "date": "2024-11-27", "categories": []}
{"id": "2406.12378", "title": "Efficient mapping of phase diagrams with conditional Boltzmann\n  Generators", "abstract": "  The accurate prediction of phase diagrams is of central importance for both\nthe fundamental understanding of materials as well as for technological\napplications in material sciences. However, the computational prediction of the\nrelative stability between phases based on their free energy is a daunting\ntask, as traditional free energy estimators require a large amount of\nsimulation data to obtain uncorrelated equilibrium samples over a grid of\nthermodynamic states. In this work, we develop deep generative machine learning\nmodels based on the Boltzmann Generator approach for entire phase diagrams,\nemploying normalizing flows conditioned on the thermodynamic states, e.g.,\ntemperature and pressure, that they map to. By training a single normalizing\nflow to transform the equilibrium distribution sampled at only one reference\nthermodynamic state to a wide range of target temperatures and pressures, we\ncan efficiently generate equilibrium samples across the entire phase diagram.\nUsing a permutation-equivariant architecture allows us, thereby, to treat solid\nand liquid phases on the same footing. We demonstrate our approach by\npredicting the solid-liquid coexistence line for a Lennard-Jones system in\nexcellent agreement with state-of-the-art free energy methods while\nsignificantly reducing the number of energy evaluations needed.\n", "date": "2024-06-18", "categories": []}
{"id": "2411.13670", "title": "Graph neural network framework for energy mapping of hybrid monte-carlo\n  molecular dynamics simulations of Medium Entropy Alloys", "abstract": "  Machine learning (ML) methods have drawn significant interest in material\ndesign and discovery. Graph neural networks (GNNs), in particular, have\ndemonstrated strong potential for predicting material properties. The present\nstudy proposes a graph-based representation for modeling medium-entropy alloys\n(MEAs). Hybrid Monte-Carlo molecular dynamics (MC/MD) simulations are employed\nto achieve thermally stable structures across various annealing temperatures in\nan MEA. These simulations generate dump files and potential energy labels,\nwhich are used to construct graph representations of the atomic configurations.\nEdges are created between each atom and its 12 nearest neighbors without\nincorporating explicit edge features. These graphs then serve as input for a\nGraph Convolutional Neural Network (GCNN) based ML model to predict the\nsystem's potential energy. The GCNN architecture effectively captures the local\nenvironment and chemical ordering within the MEA structure. The GCNN-based ML\nmodel demonstrates strong performance in predicting potential energy at\ndifferent steps, showing satisfactory results on both the training data and\nunseen configurations. Our approach presents a graph-based modeling framework\nfor MEAs and high-entropy alloys (HEAs), which effectively captures the local\nchemical order (LCO) within the alloy structure. This allows us to predict key\nmaterial properties influenced by LCO in both MEAs and HEAs, providing deeper\ninsights into how atomic-scale arrangements affect the properties of these\nalloys.\n", "date": "2024-11-20", "categories": []}
{"id": "2411.14034", "title": "Assessing data-driven predictions of band gap and electrical\n  conductivity for transparent conducting materials", "abstract": "  Machine Learning (ML) has offered innovative perspectives for accelerating\nthe discovery of new functional materials, leveraging the increasing\navailability of material databases. Despite the promising advances, data-driven\nmethods face constraints imposed by the quantity and quality of available data.\nMoreover, ML is often employed in tandem with simulated datasets originating\nfrom density functional theory (DFT), and assessed through in-sample evaluation\nschemes. This scenario raises questions about the practical utility of ML in\nuncovering new and significant material classes for industrial applications.\nHere, we propose a data-driven framework aimed at accelerating the discovery of\nnew transparent conducting materials (TCMs), an important category of\nsemiconductors with a wide range of applications. To mitigate the shortage of\navailable data, we create and validate unique experimental databases,\ncomprising several examples of existing TCMs. We assess state-of-the-art (SOTA)\nML models for property prediction from the stoichiometry alone. We propose a\nbespoke evaluation scheme to provide empirical evidence on the ability of ML to\nuncover new, previously unseen materials of interest. We test our approach on a\nlist of 55 compositions containing typical elements of known TCMs. Although our\nstudy indicates that ML tends to identify new TCMs compositionally similar to\nthose in the training data, we empirically demonstrate that it can highlight\nmaterial candidates that may have been previously overlooked, offering a\nsystematic approach to identify materials that are likely to display TCMs\ncharacteristics.\n", "date": "2024-11-21", "categories": []}
{"id": "2411.12011", "title": "SynCoTrain: A Dual Classifier PU-learning Framework for Synthesizability\n  Prediction", "abstract": "  Material discovery is a cornerstone of modern science, driving advancements\nin diverse disciplines from biomedical technology to climate solutions.\nPredicting synthesizability, a critical factor in realizing novel materials,\nremains a complex challenge due to the limitations of traditional heuristics\nand thermodynamic proxies. While stability metrics such as formation energy\noffer partial insights, they fail to account for kinetic factors and\ntechnological constraints that influence synthesis outcomes. These challenges\nare further compounded by the scarcity of negative data, as failed synthesis\nattempts are often unpublished or context-specific.\n  We present SynCoTrain, a semi-supervised machine learning model designed to\npredict the synthesizability of materials. SynCoTrain employs a co-training\nframework leveraging two complementary graph convolutional neural networks:\nSchNet and ALIGNN. By iteratively exchanging predictions between classifiers,\nSynCoTrain mitigates model bias and enhances generalizability. Our approach\nuses Positive and Unlabeled (PU) Learning to address the absence of explicit\nnegative data, iteratively refining predictions through collaborative learning.\n  The model demonstrates robust performance, achieving high recall on internal\nand leave-out test sets. By focusing on oxide crystals, a well-characterized\nmaterial family with extensive experimental data, we establish SynCoTrain as a\nreliable tool for predicting synthesizability while balancing dataset\nvariability and computational efficiency. This work highlights the potential of\nco-training to advance high-throughput materials discovery and generative\nresearch, offering a scalable solution to the challenge of synthesizability\nprediction.\n", "date": "2024-11-18", "categories": []}
{"id": "2411.12612", "title": "Reward driven workflows for unsupervised explainable analysis of phases\n  and ferroic variants from atomically resolved imaging data", "abstract": "  Rapid progress in aberration corrected electron microscopy necessitates\ndevelopment of robust methods for the identification of phases, ferroic\nvariants, and other pertinent aspects of materials structure from imaging data.\nWhile unsupervised methods for clustering and classification are widely used\nfor these tasks, their performance can be sensitive to hyperparameter selection\nin the analysis workflow. In this study, we explore the effects of descriptors\nand hyperparameters on the capability of unsupervised ML methods to distill\nlocal structural information, exemplified by discovery of polarization and\nlattice distortion in Sm doped BiFeO3 (BFO) thin films. We demonstrate that a\nreward-driven approach can be used to optimize these key hyperparameters across\nthe full workflow, where rewards were designed to reflect domain wall\ncontinuity and straightness, ensuring that the analysis aligns with the\nmaterial's physical behavior. This approach allows us to discover local\ndescriptors that are best aligned with the specific physical behavior,\nproviding insight into the fundamental physics of materials. We further extend\nthe reward driven workflows to disentangle structural factors of variation via\noptimized variational autoencoder (VAE). Finally, the importance of\nwell-defined rewards was explored as a quantifiable measure of success of the\nworkflow.\n", "date": "2024-11-19", "categories": []}
{"id": "2408.15590", "title": "Bayesian optimization of atomic structures with prior probabilities from\n  universal interatomic potentials", "abstract": "  The optimization of atomic structures plays a pivotal role in understanding\nand designing materials with desired properties. However, conventional\ncomputational methods often struggle with the formidable task of navigating the\nvast potential energy surface, especially in high-dimensional spaces with\nnumerous local minima. Recent advancements in machine learning-driven surrogate\nmodels offer a promising avenue for alleviating this computational burden. In\nthis study, we propose a novel approach that combines the strengths of\nuniversal machine learning potentials with a Bayesian approach using Gaussian\nprocesses. By using the machine learning potentials as priors for the Gaussian\nprocess, the Gaussian process has to learn only the difference between the\nmachine learning potential and the target energy surface calculated for example\nby density functional theory. This turns out to improve the speed by which the\nglobal optimal structure is identified across diverse systems for a\nwell-behaved machine learning potential. The approach is tested on periodic\nbulk materials, surface structures, and a cluster.\n", "date": "2024-08-28", "categories": []}
{"id": "2411.08414", "title": "Material Property Prediction with Element Attribute Knowledge Graphs and\n  Multimodal Representation Learning", "abstract": "  Machine learning has become a crucial tool for predicting the properties of\ncrystalline materials. However, existing methods primarily represent material\ninformation by constructing multi-edge graphs of crystal structures, often\noverlooking the chemical and physical properties of elements (such as atomic\nradius, electronegativity, melting point, and ionization energy), which have a\nsignificant impact on material performance. To address this limitation, we\nfirst constructed an element property knowledge graph and utilized an embedding\nmodel to encode the element attributes within the knowledge graph. Furthermore,\nwe propose a multimodal fusion framework, ESNet, which integrates element\nproperty features with crystal structure features to generate joint multimodal\nrepresentations. This provides a more comprehensive perspective for predicting\nthe performance of crystalline materials, enabling the model to consider both\nmicrostructural composition and chemical characteristics of the materials. We\nconducted experiments on the Materials Project benchmark dataset, which showed\nleading performance in the bandgap prediction task and achieved results on a\npar with existing benchmarks in the formation energy prediction task.\n", "date": "2024-11-13", "categories": []}
{"id": "2411.08664", "title": "UniMat: Unifying Materials Embeddings through Multi-modal Learning", "abstract": "  Materials science datasets are inherently heterogeneous and are available in\ndifferent modalities such as characterization spectra, atomic structures,\nmicroscopic images, and text-based synthesis conditions. The advancements in\nmulti-modal learning, particularly in vision and language models, have opened\nnew avenues for integrating data in different forms. In this work, we evaluate\ncommon techniques in multi-modal learning (alignment and fusion) in unifying\nsome of the most important modalities in materials science: atomic structure,\nX-ray diffraction patterns (XRD), and composition. We show that structure graph\nmodality can be enhanced by aligning with XRD patterns. Additionally, we show\nthat aligning and fusing more experimentally accessible data formats, such as\nXRD patterns and compositions, can create more robust joint embeddings than\nindividual modalities across various tasks. This lays the groundwork for future\nstudies aiming to exploit the full potential of multi-modal data in materials\nscience, facilitating more informed decision-making in materials design and\ndiscovery.\n", "date": "2024-11-13", "categories": []}
{"id": "2312.14552", "title": "Machine learning for structure-guided materials and process design", "abstract": "  In recent years, there has been a growing interest in accelerated materials\ninnovation in the context of the process-structure-property chain. In this\nregard, it is essential to take into account manufacturing processes and tailor\nmaterials design approaches to support downstream process design approaches. As\na major step into this direction, we present a holistic and generic\noptimization approach that covers the entire process-structure-property chain\nin materials engineering. Our approach specifically employs machine learning to\naddress two critical identification problems: a materials design problem, which\ninvolves identifying near-optimal material microstructures that exhibit desired\nproperties, and a process design problem that is to find an optimal processing\npath to manufacture these microstructures. Both identification problems are\ntypically ill-posed, which presents a significant challenge for solution\napproaches. However, the non-unique nature of these problems offers an\nimportant advantage for processing: By having several target microstructures\nthat perform similarly well, processes can be efficiently guided towards\nmanufacturing the best reachable microstructure. The functionality of the\napproach is demonstrated at manufacturing crystallographic textures with\ndesired properties in a simulated metal forming process.\n", "date": "2023-12-22", "categories": []}
{"id": "2411.06529", "title": "Thermodynamically-Informed Iterative Neural Operators for Heterogeneous\n  Elastic Localization", "abstract": "  Engineering problems frequently require solution of governing equations with\nspatially-varying discontinuous coefficients. Even for linear elliptic\nproblems, mapping large ensembles of coefficient fields to solutions can become\na major computational bottleneck using traditional numerical solvers.\nFurthermore, machine learning methods such as neural operators struggle to fit\nthese maps due to sharp transitions and high contrast in the coefficient fields\nand a scarcity of informative training data. In this work, we focus on a\ncanonical problem in computational mechanics: prediction of local elastic\ndeformation fields over heterogeneous material structures subjected to periodic\nboundary conditions. We construct a hybrid approximation for the\ncoefficient-to-solution map using a Thermodynamically-informed Iterative Neural\nOperator (TherINO). Rather than using coefficient fields as direct inputs and\niterating over a learned latent space, we employ thermodynamic encodings --\ndrawn from the constitutive equations -- and iterate over the solution space\nitself. Through an extensive series of case studies, we elucidate the\nadvantages of these design choices in terms of efficiency, accuracy, and\nflexibility. We also analyze the model's stability and extrapolation properties\non out-of-distribution coefficient fields and demonstrate an improved\nspeed-accuracy tradeoff for predicting elastic quantities of interest.\n", "date": "2024-11-10", "categories": []}
{"id": "2411.05033", "title": "Diagonalization without Diagonalization: A Direct Optimization Approach\n  for Solid-State Density Functional Theory", "abstract": "  We present a novel approach to address the challenges of variable occupation\nnumbers in direct optimization of density functional theory (DFT). By\nparameterizing both the eigenfunctions and the occupation matrix, our method\nminimizes the free energy with respect to these parameters. As the stationary\nconditions require the occupation matrix and the Kohn-Sham Hamiltonian to be\nsimultaneously diagonalizable, this leads to the concept of\n``self-diagonalization,'' where, by assuming a diagonal occupation matrix\nwithout loss of generality, the Hamiltonian matrix naturally becomes diagonal\nat stationary points. Our method incorporates physical constraints on both the\neigenfunctions and the occupations into the parameterization, transforming the\nconstrained optimization into an fully differentiable unconstrained problem,\nwhich is solvable via gradient descent. Implemented in JAX, our method was\ntested on aluminum and silicon, confirming that it achieves efficient\nself-diagonalization, produces the correct Fermi-Dirac distribution of the\noccupation numbers and yields band structures consistent with those obtained\nwith SCF methods in Quantum Espresso.\n", "date": "2024-11-06", "categories": []}
{"id": "2410.13873", "title": "On the Robustness of Machine Learning Models in Predicting Thermodynamic\n  Properties: a Case of Searching for New Quasicrystal Approximants", "abstract": "  Despite an artificial intelligence-assisted modeling of disordered crystals\nis a widely used and well-tried method of new materials design, the issues of\nits robustness, reliability, and stability are still not resolved and even not\ndiscussed enough. To highlight it, in this work we composed a series of nested\nintermetallic approximants of quasicrystals datasets and trained various\nmachine learning models on them correspondingly. Our qualitative and, what is\nmore important, quantitative assessment of the difference in the predictions\nclearly shows that different reasonable changes in the training sample can lead\nto the completely different set of the predicted potentially new materials. We\nalso showed the advantage of pre-training and proposed a simple yet effective\ntrick of sequential training to increase stability.\n", "date": "2024-10-02", "categories": []}
{"id": "2411.04219", "title": "Equivariant Graph Network Approximations of High-Degree Polynomials for\n  Force Field Prediction", "abstract": "  Recent advancements in equivariant deep models have shown promise in\naccurately predicting atomic potentials and force fields in molecular dynamics\nsimulations. Using spherical harmonics (SH) and tensor products (TP), these\nequivariant networks gain enhanced physical understanding, like symmetries and\nmany-body interactions. Beyond encoding physical insights, SH and TP are also\ncrucial to represent equivariant polynomial functions. In this work, we analyze\nthe equivariant polynomial functions for the equivariant architecture, and\nintroduce a novel equivariant network, named PACE. The proposed PACE utilizes\nedge booster and the Atomic Cluster Expansion (ACE) technique to approximate a\ngreater number of $SE(3) \\times S_n$ equivariant polynomial functions with\nenhanced degrees. As experimented in commonly used benchmarks, PACE\ndemonstrates state-of-the-art performance in predicting atomic energy and force\nfields, with robust generalization capability across various geometric\ndistributions under molecular dynamics (MD) across different temperature\nconditions. Our code is publicly available as part of the AIRS library\nhttps://github.com/divelab/AIRS/.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.04323", "title": "Efficient Symmetry-Aware Materials Generation via Hierarchical\n  Generative Flow Networks", "abstract": "  Discovering new solid-state materials requires rapidly exploring the vast\nspace of crystal structures and locating stable regions. Generating stable\nmaterials with desired properties and compositions is extremely difficult as we\nsearch for very small isolated pockets in the exponentially many possibilities,\nconsidering elements from the periodic table and their 3D arrangements in\ncrystal lattices. Materials discovery necessitates both optimized solution\nstructures and diversity in the generated material structures. Existing methods\nstruggle to explore large material spaces and generate diverse samples with\ndesired properties and requirements. We propose the Symmetry-aware Hierarchical\nArchitecture for Flow-based Traversal (SHAFT), a novel generative model\nemploying a hierarchical exploration strategy to efficiently exploit the\nsymmetry of the materials space to generate crystal structures given desired\nproperties. In particular, our model decomposes the exponentially large\nmaterials space into a hierarchy of subspaces consisting of symmetric space\ngroups, lattice parameters, and atoms. We demonstrate that SHAFT significantly\noutperforms state-of-the-art iterative generative methods, such as Generative\nFlow Networks (GFlowNets) and Crystal Diffusion Variational AutoEncoders\n(CDVAE), in crystal structure generation tasks, achieving higher validity,\ndiversity, and stability of generated structures optimized for target\nproperties and requirements.\n", "date": "2024-11-06", "categories": []}
{"id": "2407.04648", "title": "Efficient Materials Informatics between Rockets and Electrons", "abstract": "  The true power of computational research typically can lay in either what it\naccomplishes or what it enables others to accomplish. In this work, both\navenues are simultaneously embraced across several distinct efforts existing at\nthree general scales of abstractions of what a material is - atomistic,\nphysical, and design. At each, an efficient materials informatics\ninfrastructure is being built from the ground up based on (1) the fundamental\nunderstanding of the underlying prior knowledge, including the data, (2)\ndeployment routes that take advantage of it, and (3) pathways to extend it in\nan autonomous or semi-autonomous fashion, while heavily relying on artificial\nintelligence (AI) to guide well-established DFT-based ab initio and\nCALPHAD-based thermodynamic methods.\n  The resulting multi-level discovery infrastructure is highly generalizable as\nit focuses on encoding problems to solve them easily rather than looking for an\nexisting solution. To showcase it, this dissertation discusses the design of\nmulti-alloy functionally graded materials (FGMs) incorporating ultra-high\ntemperature refractory high entropy alloys (RHEAs) towards gas turbine and jet\nengine efficiency increase reducing CO2 emissions, as well as hypersonic\nvehicles. It leverages a new graph representation of underlying mathematical\nspace using a newly developed algorithm based on combinatorics, not subject to\nmany problems troubling the community. Underneath, property models and phase\nrelations are learned from optimized samplings of the largest and highest\nquality dataset of HEA in the world, called ULTERA. At the atomistic level, a\ndata ecosystem optimized for machine learning (ML) from over 4.5 million\nrelaxed structures, called MPDD, is used to inform experimental observations\nand improve thermodynamic models by providing stability data enabled by a new\nefficient featurization framework.\n", "date": "2024-07-05", "categories": []}
{"id": "2411.03156", "title": "Unleashing the power of novel conditional generative approaches for new\n  materials discovery", "abstract": "  For a very long time, computational approaches to the design of new materials\nhave relied on an iterative process of finding a candidate material and\nmodeling its properties. AI has played a crucial role in this regard, helping\nto accelerate the discovery and optimization of crystal properties and\nstructures through advanced computational methodologies and data-driven\napproaches. To address the problem of new materials design and fasten the\nprocess of new materials search, we have applied latest generative approaches\nto the problem of crystal structure design, trying to solve the inverse\nproblem: by given properties generate a structure that satisfies them without\nutilizing supercomputer powers. In our work we propose two approaches: 1)\nconditional structure modification: optimization of the stability of an\narbitrary atomic configuration, using the energy difference between the most\nenergetically favorable structure and all its less stable polymorphs and 2)\nconditional structure generation. We used a representation for materials that\nincludes the following information: lattice, atom coordinates, atom types,\nchemical features, space group and formation energy of the structure. The loss\nfunction was optimized to take into account the periodic boundary conditions of\ncrystal structures. We have applied Diffusion models approach, Flow matching,\nusual Autoencoder (AE) and compared the results of the models and approaches.\nAs a metric for the study, physical PyMatGen matcher was employed: we compare\ntarget structure with generated one using default tolerances. So far, our\nmodifier and generator produce structures with needed properties with accuracy\n41% and 82% respectively. To prove the offered methodology efficiency,\ninference have been carried out, resulting in several potentially new\nstructures with formation energy below the AFLOW-derived convex hulls.\n", "date": "2024-11-05", "categories": []}
{"id": "2410.16165", "title": "From Tokens to Materials: Leveraging Language Models for Scientific\n  Discovery", "abstract": "  Exploring the predictive capabilities of language models in material science\nis an ongoing interest. This study investigates the application of language\nmodel embeddings to enhance material property prediction in materials science.\nBy evaluating various contextual embedding methods and pre-trained models,\nincluding Bidirectional Encoder Representations from Transformers (BERT) and\nGenerative Pre-trained Transformers (GPT), we demonstrate that domain-specific\nmodels, particularly MatBERT significantly outperform general-purpose models in\nextracting implicit knowledge from compound names and material properties. Our\nfindings reveal that information-dense embeddings from the third layer of\nMatBERT, combined with a context-averaging approach, offer the most effective\nmethod for capturing material-property relationships from the scientific\nliterature. We also identify a crucial \"tokenizer effect,\" highlighting the\nimportance of specialized text processing techniques that preserve complete\ncompound names while maintaining consistent token counts. These insights\nunderscore the value of domain-specific training and tokenization in materials\nscience applications and offer a promising pathway for accelerating the\ndiscovery and development of new materials through AI-driven approaches.\n", "date": "2024-10-21", "categories": []}
{"id": "2411.01008", "title": "AI-Guided Codesign Framework for Novel Material and Device Design\n  applied to MTJ-based True Random Number Generators", "abstract": "  Novel devices and novel computing paradigms are key for energy efficient,\nperformant future computing systems. However, designing devices for new\napplications is often time consuming and tedious. Here, we investigate the\ndesign and optimization of spin orbit torque and spin transfer torque magnetic\ntunnel junction models as the probabilistic devices for true random number\ngeneration. We leverage reinforcement learning and evolutionary optimization to\nvary key device and material properties of the various device models for\nstochastic operation. Our AI guided codesign methods generated different\ncandidate devices capable of generating stochastic samples for a desired\nprobability distribution, while also minimizing energy usage for the devices.\n", "date": "2024-11-01", "categories": []}
{"id": "2209.12605", "title": "MechProNet: Machine Learning Prediction of Mechanical Properties in\n  Metal Additive Manufacturing", "abstract": "  Predicting mechanical properties in metal additive manufacturing (MAM) is\nessential for ensuring the performance and reliability of printed parts, as\nwell as their suitability for specific applications. However, conducting\nexperiments to estimate mechanical properties in MAM processes can be laborious\nand expensive, and they are often limited to specific materials and processes.\nMachine learning (ML) methods offer a more flexible and cost-effective approach\nto predicting mechanical properties based on processing parameters and material\nproperties. In this study, we introduce a comprehensive framework for\nbenchmarking ML models for predicting mechanical properties. We compiled an\nextensive experimental dataset from over 90 MAM articles and data sheets from a\ndiverse range of sources, encompassing 140 different MAM data sheets. This\ndataset includes information on MAM processing conditions, machines, materials,\nand resulting mechanical properties such as yield strength, ultimate tensile\nstrength, elastic modulus, elongation, hardness, and surface roughness. Our\nframework incorporates physics-aware featurization specific to MAM, adjustable\nML models, and tailored evaluation metrics to construct a comprehensive\nlearning framework for predicting mechanical properties. Additionally, we\nexplore the Explainable AI method, specifically SHAP analysis, to elucidate and\ninterpret the predicted values of ML models for mechanical properties.\nFurthermore, data-driven explicit models were developed to estimate mechanical\nproperties based on processing parameters and material properties, offering\nenhanced interpretability compared to conventional ML models.\n", "date": "2022-08-21", "categories": []}
{"id": "2406.10796", "title": "Ab Initio Structure Solutions from Nanocrystalline Powder Diffraction\n  Data", "abstract": "  A major challenge in materials science is the determination of the structure\nof nanometer sized objects. Here we present a novel approach that uses a\ngenerative machine learning model based on diffusion processes that is trained\non 45,229 known structures. The model factors both the measured diffraction\npattern as well as relevant statistical priors on the unit cell of atomic\ncluster structures. Conditioned only on the chemical formula and the\ninformation-scarce finite-size broadened powder diffraction pattern, we find\nthat our model, PXRDnet, can successfully solve simulated nanocrystals as small\nas 10 angstroms across 200 materials of varying symmetry and complexity,\nincluding structures from all seven crystal systems. We show that our model can\nsuccessfully and verifiably determine structural candidates four out of five\ntimes, with average error among these candidates being only 7% (as measured by\npost-Rietveld refinement R-factor). Furthermore, PXRDnet is capable of solving\nstructures from noisy diffraction patterns gathered in real-world experiments.\nWe suggest that data driven approaches, bootstrapped from theoretical\nsimulation, will ultimately provide a path towards determining the structure of\npreviously unsolved nano-materials.\n", "date": "2024-06-16", "categories": []}
{"id": "2410.17518", "title": "Univariate Conditional Variational Autoencoder for Morphogenic Patterns\n  Design in Frontal Polymerization-Based Manufacturing", "abstract": "  Under some initial and boundary conditions, the rapid reaction-thermal\ndiffusion process taking place during frontal polymerization (FP) destabilizes\nthe planar mode of front propagation, leading to spatially varying, complex\nhierarchical patterns in thermoset polymeric materials. Although modern\nreaction-diffusion models can predict the patterns resulting from unstable FP,\nthe inverse design of patterns, which aims to retrieve process conditions that\nproduce a desired pattern, remains an open challenge due to the non-unique and\nnon-intuitive mapping between process conditions and manufactured patterns. In\nthis work, we propose a probabilistic generative model named univariate\nconditional variational autoencoder (UcVAE) for the inverse design of\nhierarchical patterns in FP-based manufacturing. Unlike the cVAE, which encodes\nboth the design space and the design target, the UcVAE encodes only the design\nspace. In the encoder of the UcVAE, the number of training parameters is\nsignificantly reduced compared to the cVAE, resulting in a shorter training\ntime while maintaining comparable performance. Given desired pattern images,\nthe trained UcVAE can generate multiple process condition solutions that\nproduce high-fidelity hierarchical patterns.\n", "date": "2024-10-23", "categories": []}
{"id": "2410.23405", "title": "FlowLLM: Flow Matching for Material Generation with Large Language\n  Models as Base Distributions", "abstract": "  Material discovery is a critical area of research with the potential to\nrevolutionize various fields, including carbon capture, renewable energy, and\nelectronics. However, the immense scale of the chemical space makes it\nchallenging to explore all possible materials experimentally. In this paper, we\nintroduce FlowLLM, a novel generative model that combines large language models\n(LLMs) and Riemannian flow matching (RFM) to design novel crystalline\nmaterials. FlowLLM first fine-tunes an LLM to learn an effective base\ndistribution of meta-stable crystals in a text representation. After converting\nto a graph representation, the RFM model takes samples from the LLM and\niteratively refines the coordinates and lattice parameters. Our approach\nsignificantly outperforms state-of-the-art methods, increasing the generation\nrate of stable materials by over three times and increasing the rate for\nstable, unique, and novel crystals by $\\sim50\\%$ - a huge improvement on a\ndifficult problem. Additionally, the crystals generated by FlowLLM are much\ncloser to their relaxed state when compared with another leading model,\nsignificantly reducing post-hoc computational cost.\n", "date": "2024-10-30", "categories": []}
{"id": "2410.22570", "title": "Orb: A Fast, Scalable Neural Network Potential", "abstract": "  We introduce Orb, a family of universal interatomic potentials for atomistic\nmodelling of materials. Orb models are 3-6 times faster than existing universal\npotentials, stable under simulation for a range of out of distribution\nmaterials and, upon release, represented a 31% reduction in error over other\nmethods on the Matbench Discovery benchmark. We explore several aspects of\nfoundation model development for materials, with a focus on diffusion\npretraining. We evaluate Orb as a model for geometry optimization, Monte Carlo\nand molecular dynamics simulations.\n", "date": "2024-10-29", "categories": []}
{"id": "2410.21317", "title": "MatExpert: Decomposing Materials Discovery by Mimicking Human Experts", "abstract": "  Material discovery is a critical research area with profound implications for\nvarious industries. In this work, we introduce MatExpert, a novel framework\nthat leverages Large Language Models (LLMs) and contrastive learning to\naccelerate the discovery and design of new solid-state materials. Inspired by\nthe workflow of human materials design experts, our approach integrates three\nkey stages: retrieval, transition, and generation. First, in the retrieval\nstage, MatExpert identifies an existing material that closely matches the\ndesired criteria. Second, in the transition stage, MatExpert outlines the\nnecessary modifications to transform this material formulation to meet specific\nrequirements outlined by the initial user query. Third, in the generation\nstate, MatExpert performs detailed computations and structural generation to\ncreate new materials based on the provided information. Our experimental\nresults demonstrate that MatExpert outperforms state-of-the-art methods in\nmaterial generation tasks, achieving superior performance across various\nmetrics including validity, distribution, and stability. As such, MatExpert\nrepresents a meaningful advancement in computational material discovery using\nlangauge-based generative models.\n", "date": "2024-10-26", "categories": []}
{"id": "2410.20402", "title": "Deep Learning-Driven Microstructure Characterization and Vickers\n  Hardness Prediction of Mg-Gd Alloys", "abstract": "  In the field of materials science, exploring the relationship between\ncomposition, microstructure, and properties has long been a critical research\nfocus. The mechanical performance of solid-solution Mg-Gd alloys is\nsignificantly influenced by Gd content, dendritic structures, and the presence\nof secondary phases. To better analyze and predict the impact of these factors,\nthis study proposes a multimodal fusion learning framework based on image\nprocessing and deep learning techniques. This framework integrates both\nelemental composition and microstructural features to accurately predict the\nVickers hardness of solid-solution Mg-Gd alloys. Initially, deep learning\nmethods were employed to extract microstructural information from a variety of\nsolid-solution Mg-Gd alloy images obtained from literature and experiments.\nThis provided precise grain size and secondary phase microstructural features\nfor performance prediction tasks. Subsequently, these quantitative analysis\nresults were combined with Gd content information to construct a performance\nprediction dataset. Finally, a regression model based on the Transformer\narchitecture was used to predict the Vickers hardness of Mg-Gd alloys. The\nexperimental results indicate that the Transformer model performs best in terms\nof prediction accuracy, achieving an R^2 value of 0.9. Additionally, SHAP\nanalysis identified critical values for four key features affecting the Vickers\nhardness of Mg-Gd alloys, providing valuable guidance for alloy design. These\nfindings not only enhance the understanding of alloy performance but also offer\ntheoretical support for future material design and optimization.\n", "date": "2024-10-27", "categories": []}
{"id": "2410.20976", "title": "Large Language Model-Guided Prediction Toward Quantum Materials\n  Synthesis", "abstract": "  The synthesis of inorganic crystalline materials is essential for modern\ntechnology, especially in quantum materials development. However, designing\nefficient synthesis workflows remains a significant challenge due to the\nprecise experimental conditions and extensive trial and error. Here, we present\na framework using large language models (LLMs) to predict synthesis pathways\nfor inorganic materials, including quantum materials. Our framework contains\nthree models: LHS2RHS, predicting products from reactants; RHS2LHS, predicting\nreactants from products; and TGT2CEQ, generating full chemical equations for\ntarget compounds. Fine-tuned on a text-mined synthesis database, our model\nraises accuracy from under 40% with pretrained models, to under 80% using\nconventional fine-tuning, and further to around 90% with our proposed\ngeneralized Tanimoto similarity, while maintaining robust to additional\nsynthesis steps. Our model further demonstrates comparable performance across\nmaterials with varying degrees of quantumness quantified using quantum weight,\nindicating that LLMs offer a powerful tool to predict balanced chemical\nequations for quantum materials discovery.\n", "date": "2024-10-28", "categories": []}
{"id": "2409.08054", "title": "Predicting and Accelerating Nanomaterials Synthesis Using Machine\n  Learning Featurization", "abstract": "  Materials synthesis optimization is constrained by serial feedback processes\nthat rely on manual tools and intuition across multiple siloed modes of\ncharacterization. We automate and generalize feature extraction of reflection\nhigh-energy electron diffraction (RHEED) data with machine learning to\nestablish quantitatively predictive relationships in small sets (\\~10) of\nexpert-labeled data, saving significant time on subsequently grown samples.\nThese predictive relationships are evaluated in a representative material\nsystem (\\ce{W_{1-x}V_xSe2} on c-plane sapphire (0001)) with two aims: 1)\npredicting grain alignment of the deposited film using pre-growth substrate\ndata, and 2) estimating vanadium dopant concentration using in-situ RHEED as a\nproxy for ex-situ methods (e.g. x-ray photoelectron spectroscopy). Both tasks\nare accomplished using the same materials-agnostic features, avoiding specific\nsystem retraining and leading to a potential 80\\% time saving over a 100-sample\nsynthesis campaign. These predictions provide guidance to avoid doomed trials,\nreduce follow-on characterization, and improve control resolution for materials\nsynthesis.\n", "date": "2024-09-12", "categories": []}
{"id": "2410.17430", "title": "Real-time experiment-theory closed-loop interaction for autonomous\n  materials science", "abstract": "  Iterative cycles of theoretical prediction and experimental validation are\nthe cornerstone of the modern scientific method. However, the proverbial\n\"closing of the loop\" in experiment-theory cycles in practice are usually ad\nhoc, often inherently difficult, or impractical to repeat on a systematic\nbasis, beset by the scale or the time constraint of computation or the\nphenomena under study. Here, we demonstrate Autonomous MAterials Search Engine\n(AMASE), where we enlist robot science to perform self-driving continuous\ncyclical interaction of experiments and computational predictions for materials\nexploration. In particular, we have applied the AMASE formalism to the rapid\nmapping of a temperature-composition phase diagram, a fundamental task for the\nsearch and discovery of new materials. Thermal processing and experimental\ndetermination of compositional phase boundaries in thin films are autonomously\ninterspersed with real-time updating of the phase diagram prediction through\nthe minimization of Gibbs free energies. AMASE was able to accurately determine\nthe eutectic phase diagram of the Sn-Bi binary thin-film system on the fly from\na self-guided campaign covering just a small fraction of the entire composition\n- temperature phase space, translating to a 6-fold reduction in the number of\nnecessary experiments. This study demonstrates for the first time the\npossibility of real-time, autonomous, and iterative interactions of experiments\nand theory carried out without any human intervention.\n", "date": "2024-10-22", "categories": []}
{"id": "2410.17631", "title": "Exploring structure diversity in atomic resolution microscopy with graph\n  neural networks", "abstract": "  The emergence of deep learning (DL) has provided great opportunities for the\nhigh-throughput analysis of atomic-resolution micrographs. However, the DL\nmodels trained by image patches in fixed size generally lack efficiency and\nflexibility when processing micrographs containing diversified atomic\nconfigurations. Herein, inspired by the similarity between the atomic\nstructures and graphs, we describe a few-shot learning framework based on an\nequivariant graph neural network (EGNN) to analyze a library of atomic\nstructures (e.g., vacancies, phases, grain boundaries, doping, etc.), showing\nsignificantly promoted robustness and three orders of magnitude reduced\ncomputing parameters compared to the image-driven DL models, which is\nespecially evident for those aggregated vacancy lines with flexible lattice\ndistortion. Besides, the intuitiveness of graphs enables quantitative and\nstraightforward extraction of the atomic-scale structural features in batches,\nthus statistically unveiling the self-assembly dynamics of vacancy lines under\nelectron beam irradiation. A versatile model toolkit is established by\nintegrating EGNN sub-models for single structure recognition to process images\ninvolving varied configurations in the form of a task chain, leading to the\ndiscovery of novel doping configurations with superior electrocatalytic\nproperties for hydrogen evolution reactions. This work provides a powerful tool\nto explore structure diversity in a fast, accurate, and intelligent manner.\n", "date": "2024-10-23", "categories": []}
{"id": "2410.13768", "title": "Rapid and Automated Alloy Design with Graph Neural Network-Powered\n  LLM-Driven Multi-Agent Systems", "abstract": "  A multi-agent AI model is used to automate the discovery of new metallic\nalloys, integrating multimodal data and external knowledge including insights\nfrom physics via atomistic simulations. Our multi-agent system features three\nkey components: (a) a suite of LLMs responsible for tasks such as reasoning and\nplanning, (b) a group of AI agents with distinct roles and expertise that\ndynamically collaborate, and (c) a newly developed graph neural network (GNN)\nmodel for rapid retrieval of key physical properties. A set of LLM-driven AI\nagents collaborate to automate the exploration of the vast design space of\nMPEAs, guided by predictions from the GNN. We focus on the NbMoTa family of\nbody-centered cubic (bcc) alloys, modeled using an ML-based interatomic\npotential, and target two key properties: the Peierls barrier and solute/screw\ndislocation interaction energy. Our GNN model accurately predicts these\natomic-scale properties, providing a faster alternative to costly brute-force\ncalculations and reducing the computational burden on multi-agent systems for\nphysics retrieval. This AI system revolutionizes materials discovery by\nreducing reliance on human expertise and overcoming the limitations of direct\nall-atom simulations. By synergizing the predictive power of GNNs with the\ndynamic collaboration of LLM-based agents, the system autonomously navigates\nvast alloy design spaces, identifying trends in atomic-scale material\nproperties and predicting macro-scale mechanical strength, as demonstrated by\nseveral computational experiments. This approach accelerates the discovery of\nadvanced alloys and holds promise for broader applications in other complex\nsystems, marking a significant step forward in automated materials design.\n", "date": "2024-10-17", "categories": []}
{"id": "2410.10897", "title": "EPi-cKANs: Elasto-Plasticity Informed Kolmogorov-Arnold Networks Using\n  Chebyshev Polynomials", "abstract": "  Multilayer perceptron (MLP) networks are predominantly used to develop\ndata-driven constitutive models for granular materials. They offer a compelling\nalternative to traditional physics-based constitutive models in predicting\nnonlinear responses of these materials, e.g., elasto-plasticity, under various\nloading conditions. To attain the necessary accuracy, MLPs often need to be\nsufficiently deep or wide, owing to the curse of dimensionality inherent in\nthese problems. To overcome this limitation, we present an elasto-plasticity\ninformed Chebyshev-based Kolmogorov-Arnold network (EPi-cKAN) in this study.\nThis architecture leverages the benefits of KANs and augmented Chebyshev\npolynomials, as well as integrates physical principles within both the network\nstructure and the loss function. The primary objective of EPi-cKAN is to\nprovide an accurate and generalizable function approximation for non-linear\nstress-strain relationships, using fewer parameters compared to standard MLPs.\nTo evaluate the efficiency, accuracy, and generalization capabilities of\nEPi-cKAN in modeling complex elasto-plastic behavior, we initially compare its\nperformance with other cKAN-based models, which include purely data-driven\nparallel and serial architectures. Furthermore, to differentiate EPi-cKAN's\ndistinct performance, we also compare it against purely data-driven and\nphysics-informed MLP-based methods. Lastly, we test EPi-cKAN's ability to\npredict blind strain-controlled paths that extend beyond the training data\ndistribution to gauge its generalization and predictive capabilities. Our\nfindings indicate that, even with limited data and fewer parameters compared to\nother approaches, EPi-cKAN provides superior accuracy in predicting stress\ncomponents and demonstrates better generalization when used to predict sand\nelasto-plastic behavior under blind triaxial axisymmetric strain-controlled\nloading paths.\n", "date": "2024-10-12", "categories": []}
{"id": "2406.03794", "title": "Infusing Self-Consistency into Density Functional Theory Hamiltonian\n  Prediction via Deep Equilibrium Models", "abstract": "  In this study, we introduce a unified neural network architecture, the Deep\nEquilibrium Density Functional Theory Hamiltonian (DEQH) model, which\nincorporates Deep Equilibrium Models (DEQs) for predicting Density Functional\nTheory (DFT) Hamiltonians. The DEQH model inherently captures the\nself-consistency nature of Hamiltonian, a critical aspect often overlooked by\ntraditional machine learning approaches for Hamiltonian prediction. By\nemploying DEQ within our model architecture, we circumvent the need for DFT\ncalculations during the training phase to introduce the Hamiltonian's\nself-consistency, thus addressing computational bottlenecks associated with\nlarge or complex systems. We propose a versatile framework that combines DEQ\nwith off-the-shelf machine learning models for predicting Hamiltonians. When\nbenchmarked on the MD17 and QH9 datasets, DEQHNet, an instantiation of the DEQH\nframework, has demonstrated a significant improvement in prediction accuracy.\nBeyond a predictor, the DEQH model is a Hamiltonian solver, in the sense that\nit uses the fixed-point solving capability of the deep equilibrium model to\niteratively solve for the Hamiltonian. Ablation studies of DEQHNet further\nelucidate the network's effectiveness, offering insights into the potential of\nDEQ-integrated networks for Hamiltonian learning. We open source our\nimplementation at https://github.com/Zun-Wang/DEQHNet.\n", "date": "2024-06-06", "categories": []}
{"id": "2410.05281", "title": "Micrometer: Micromechanics Transformer for Predicting Mechanical\n  Responses of Heterogeneous Materials", "abstract": "  Heterogeneous materials, crucial in various engineering applications, exhibit\ncomplex multiscale behavior, which challenges the effectiveness of traditional\ncomputational methods. In this work, we introduce the Micromechanics\nTransformer ({\\em Micrometer}), an artificial intelligence (AI) framework for\npredicting the mechanical response of heterogeneous materials, bridging the gap\nbetween advanced data-driven methods and complex solid mechanics problems.\nTrained on a large-scale high-resolution dataset of 2D fiber-reinforced\ncomposites, Micrometer can achieve state-of-the-art performance in predicting\nmicroscale strain fields across a wide range of microstructures, material\nproperties under any loading conditions and We demonstrate the accuracy and\ncomputational efficiency of Micrometer through applications in computational\nhomogenization and multiscale modeling, where Micrometer achieves 1\\% error in\npredicting macroscale stress fields while reducing computational time by up to\ntwo orders of magnitude compared to conventional numerical solvers. We further\nshowcase the adaptability of the proposed model through transfer learning\nexperiments on new materials with limited data, highlighting its potential to\ntackle diverse scenarios in mechanical analysis of solid materials. Our work\nrepresents a significant step towards AI-driven innovation in computational\nsolid mechanics, addressing the limitations of traditional numerical methods\nand paving the way for more efficient simulations of heterogeneous materials\nacross various industrial applications.\n", "date": "2024-09-23", "categories": []}
{"id": "2409.05928", "title": "Machine Learning Based Optimal Design of Fibrillar Adhesives", "abstract": "  Fibrillar adhesion, observed in animals like beetles, spiders, and geckos,\nrelies on nanoscopic or microscopic fibrils to enhance surface adhesion via\n'contact splitting.' This concept has inspired engineering applications across\nrobotics, transportation, and medicine. Recent studies suggest that functional\ngrading of fibril properties can improve adhesion, but this is a complex design\nchallenge that has only been explored in simplified geometries. While machine\nlearning (ML) has gained traction in adhesive design, no previous attempts have\ntargeted fibril-array scale optimization. In this study, we propose an ML-based\ntool that optimizes the distribution of fibril compliance to maximize adhesive\nstrength. Our tool, featuring two deep neural networks (DNNs), recovers\nprevious design results for simple geometries and introduces novel solutions\nfor complex configurations. The Predictor DNN estimates adhesive strength based\non random compliance distributions, while the Designer DNN optimizes compliance\nfor maximum strength using gradient-based optimization. Our method\nsignificantly reduces test error and accelerates the optimization process,\noffering a high-performance solution for designing fibrillar adhesives and\nmicro-architected materials aimed at fracture resistance by achieving equal\nload sharing (ELS).\n", "date": "2024-09-09", "categories": []}
{"id": "2410.04765", "title": "Molecular topological deep learning for polymer property prediction", "abstract": "  Accurate and efficient prediction of polymer properties is of key importance\nfor polymer design. Traditional experimental tools and density function theory\n(DFT)-based simulations for polymer property evaluation, are both expensive and\ntime-consuming. Recently, a gigantic amount of graph-based molecular models\nhave emerged and demonstrated huge potential in molecular data analysis. Even\nwith the great progresses, these models tend to ignore the high-order and\nmutliscale information within the data. In this paper, we develop molecular\ntopological deep learning (Mol-TDL) for polymer property analysis. Our Mol-TDL\nincorporates both high-order interactions and multiscale properties into\ntopological deep learning architecture. The key idea is to represent polymer\nmolecules as a series of simplicial complices at different scales and build up\nsimplical neural networks accordingly. The aggregated information from\ndifferent scales provides a more accurate prediction of polymer molecular\nproperties.\n", "date": "2024-10-07", "categories": []}
{"id": "2402.13221", "title": "CHILI: Chemically-Informed Large-scale Inorganic Nanomaterials Dataset\n  for Advancing Graph Machine Learning", "abstract": "  Advances in graph machine learning (ML) have been driven by applications in\nchemistry as graphs have remained the most expressive representations of\nmolecules. While early graph ML methods focused primarily on small organic\nmolecules, recently, the scope of graph ML has expanded to include inorganic\nmaterials. Modelling the periodicity and symmetry of inorganic crystalline\nmaterials poses unique challenges, which existing graph ML methods are unable\nto address. Moving to inorganic nanomaterials increases complexity as the scale\nof number of nodes within each graph can be broad ($10$ to $10^5$). The bulk of\nexisting graph ML focuses on characterising molecules and materials by\npredicting target properties with graphs as input. However, the most exciting\napplications of graph ML will be in their generative capabilities, which is\ncurrently not at par with other domains such as images or text.\n  We invite the graph ML community to address these open challenges by\npresenting two new chemically-informed large-scale inorganic (CHILI)\nnanomaterials datasets: A medium-scale dataset (with overall >6M nodes, >49M\nedges) of mono-metallic oxide nanomaterials generated from 12 selected crystal\ntypes (CHILI-3K) and a large-scale dataset (with overall >183M nodes, >1.2B\nedges) of nanomaterials generated from experimentally determined crystal\nstructures (CHILI-100K). We define 11 property prediction tasks and 6 structure\nprediction tasks, which are of special interest for nanomaterial research. We\nbenchmark the performance of a wide array of baseline methods and use these\nbenchmarking results to highlight areas which need future work. To the best of\nour knowledge, CHILI-3K and CHILI-100K are the first open-source nanomaterial\ndatasets of this scale -- both on the individual graph level and of the dataset\nas a whole -- and the only nanomaterials datasets with high structural and\nelemental diversity.\n", "date": "2024-02-20", "categories": []}
{"id": "2405.17881", "title": "Crystal-LSBO: Automated Design of De Novo Crystals with Latent Space\n  Bayesian Optimization", "abstract": "  Generative modeling of crystal structures is significantly challenged by the\ncomplexity of input data, which constrains the ability of these models to\nexplore and discover novel crystals. This complexity often confines de novo\ndesign methodologies to merely small perturbations of known crystals and\nhampers the effective application of advanced optimization techniques. One such\noptimization technique, Latent Space Bayesian Optimization (LSBO) has\ndemonstrated promising results in uncovering novel objects across various\ndomains, especially when combined with Variational Autoencoders (VAEs).\nRecognizing LSBO's potential and the critical need for innovative crystal\ndiscovery, we introduce Crystal-LSBO, a de novo design framework for crystals\nspecifically tailored to enhance explorability within LSBO frameworks.\nCrystal-LSBO employs multiple VAEs, each dedicated to a distinct aspect of\ncrystal structure: lattice, coordinates, and chemical elements, orchestrated by\nan integrative model that synthesizes these components into a cohesive output.\nThis setup not only streamlines the learning process but also produces\nexplorable latent spaces thanks to the decreased complexity of the learning\ntask for each model, enabling LSBO approaches to operate. Our study pioneers\nthe use of LSBO for de novo crystal design, demonstrating its efficacy through\noptimization tasks focused mainly on formation energy values. Our results\nhighlight the effectiveness of our methodology, offering a new perspective for\nde novo crystal discovery.\n", "date": "2024-05-28", "categories": []}
{"id": "2409.11782", "title": "Smart Data-Driven GRU Predictor for SnO$_2$ Thin films Characteristics", "abstract": "  In material physics, characterization techniques are foremost crucial for\nobtaining the materials data regarding the physical properties as well as\nstructural, electronics, magnetic, optic, dielectric, and spectroscopic\ncharacteristics. However, for many materials, ensuring availability and safe\naccessibility is not always easy and fully warranted. Moreover, the use of\nmodeling and simulation techniques need a lot of theoretical knowledge, in\naddition of being associated to costly computation time and a great complexity\ndeal. Thus, analyzing materials with different techniques for multiple samples\nsimultaneously, still be very challenging for engineers and researchers. It is\nworth noting that although of being very risky, X-ray diffraction is the well\nknown and widely used characterization technique which gathers data from\nstructural properties of crystalline 1d, 2d or 3d materials. We propose in this\npaper, a Smart GRU for Gated Recurrent Unit model to forcast structural\ncharacteristics or properties of thin films of tin oxide SnO$_2$(110). Indeed,\nthin films samples are elaborated and managed experimentally and the collected\ndata dictionary is then used to generate an AI -- Artificial Intelligence --\nGRU model for the thin films of tin oxide SnO$_2$(110) structural property\ncharacterization.\n", "date": "2024-09-18", "categories": []}
{"id": "2409.13908", "title": "Nonlinear Inverse Design of Mechanical Multi-Material Metamaterials\n  Enabled by Video Denoising Diffusion and Structure Identifier", "abstract": "  Metamaterials, synthetic materials with customized properties, have emerged\nas a promising field due to advancements in additive manufacturing. These\nmaterials derive unique mechanical properties from their internal lattice\nstructures, which are often composed of multiple materials that repeat\ngeometric patterns. While traditional inverse design approaches have shown\npotential, they struggle to map nonlinear material behavior to multiple\npossible structural configurations. This paper presents a novel framework\nleveraging video diffusion models, a type of generative artificial Intelligence\n(AI), for inverse multi-material design based on nonlinear stress-strain\nresponses. Our approach consists of two key components: (1) a fields generator\nusing a video diffusion model to create solution fields based on target\nnonlinear stress-strain responses, and (2) a structure identifier employing two\nUNet models to determine the corresponding multi-material 2D design. By\nincorporating multiple materials, plasticity, and large deformation, our\ninnovative design method allows for enhanced control over the highly nonlinear\nmechanical behavior of metamaterials commonly seen in real-world applications.\nIt offers a promising solution for generating next-generation metamaterials\nwith finely tuned mechanical characteristics.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.19209", "title": "Boosting SISSO Performance on Small Sample Datasets by Using Random\n  Forests Prescreening for Complex Feature Selection", "abstract": "  In materials science, data-driven methods accelerate material discovery and\noptimization while reducing costs and improving success rates. Symbolic\nregression is a key to extracting material descriptors from large datasets, in\nparticular the Sure Independence Screening and Sparsifying Operator (SISSO)\nmethod. While SISSO needs to store the entire expression space to impose heavy\nmemory demands, it limits the performance in complex problems. To address this\nissue, we propose a RF-SISSO algorithm by combining Random Forests (RF) with\nSISSO. In this algorithm, the Random Forest algorithm is used for prescreening,\ncapturing non-linear relationships and improving feature selection, which may\nenhance the quality of the input data and boost the accuracy and efficiency on\nregression and classification tasks. For a testing on the SISSO's verification\nproblem for 299 materials, RF-SISSO demonstrates its robust performance and\nhigh accuracy. RF-SISSO can maintain the testing accuracy above 0.9 across all\nfour training sample sizes and significantly enhancing regression efficiency,\nespecially in training subsets with smaller sample sizes. For the training\nsubset with 45 samples, the efficiency of RF-SISSO was 265 times higher than\nthat of original SISSO. As collecting large datasets would be both costly and\ntime-consuming in the practical experiments, it is thus believed that RF-SISSO\nmay benefit scientific researches by offering a high predicting accuracy with\nlimited data efficiently.\n", "date": "2024-09-28", "categories": []}
{"id": "2409.15600", "title": "Polyatomic Complexes: A topologically-informed learning representation\n  for atomistic systems", "abstract": "  Developing robust representations of chemical structures that enable models\nto learn topological inductive biases is challenging. In this manuscript, we\npresent a representation of atomistic systems. We begin by proving that our\nrepresentation satisfies all structural, geometric, efficiency, and\ngeneralizability constraints. Afterward, we provide a general algorithm to\nencode any atomistic system. Finally, we report performance comparable to\nstate-of-the-art methods on numerous tasks. We open-source all code and\ndatasets. The code and data are available at\nhttps://github.com/rahulkhorana/PolyatomicComplexes.\n", "date": "2024-09-23", "categories": []}
{"id": "2409.17808", "title": "Generative Modeling of Molecular Dynamics Trajectories", "abstract": "  Molecular dynamics (MD) is a powerful technique for studying microscopic\nphenomena, but its computational cost has driven significant interest in the\ndevelopment of deep learning-based surrogate models. We introduce generative\nmodeling of molecular trajectories as a paradigm for learning flexible\nmulti-task surrogate models of MD from data. By conditioning on appropriately\nchosen frames of the trajectory, we show such generative models can be adapted\nto diverse tasks such as forward simulation, transition path sampling, and\ntrajectory upsampling. By alternatively conditioning on part of the molecular\nsystem and inpainting the rest, we also demonstrate the first steps towards\ndynamics-conditioned molecular design. We validate the full set of these\ncapabilities on tetrapeptide simulations and show that our model can produce\nreasonable ensembles of protein monomers. Altogether, our work illustrates how\ngenerative modeling can unlock value from MD data towards diverse downstream\ntasks that are not straightforward to address with existing methods or even MD\nitself. Code is available at https://github.com/bjing2016/mdgen.\n", "date": "2024-09-26", "categories": []}
{"id": "2409.15391", "title": "Supply Risk-Aware Alloy Discovery and Design", "abstract": "  Materials design is a critical driver of innovation, yet overlooking the\ntechnological, economic, and environmental risks inherent in materials and\ntheir supply chains can lead to unsustainable and risk-prone solutions. To\naddress this, we present a novel risk-aware design approach that integrates\nSupply-Chain Aware Design Strategies into the materials development process.\nThis approach leverages existing language models and text analysis to develop a\nspecialized model for predicting materials feedstock supply risk indices. To\nefficiently navigate the multi-objective, multi-constraint design space, we\nemploy Batch Bayesian Optimization (BBO), enabling the identification of\nPareto-optimal high entropy alloys (HEAs) that balance performance objectives\nwith minimized supply risk. A case study using the MoNbTiVW system demonstrates\nthe efficacy of our approach in four scenarios, highlighting the significant\nimpact of incorporating supply risk into the design process. By optimizing for\nboth performance and supply risk, we ensure that the developed alloys are not\nonly high-performing but also sustainable and economically viable. This\nintegrated approach represents a critical step towards a future where materials\ndiscovery and design seamlessly consider sustainability, supply chain dynamics,\nand comprehensive life cycle analysis.\n", "date": "2024-09-22", "categories": []}
{"id": "2409.15616", "title": "Reinforcement Feature Transformation for Polymer Property Performance\n  Prediction", "abstract": "  Polymer property performance prediction aims to forecast specific features or\nattributes of polymers, which has become an efficient approach to measuring\ntheir performance. However, existing machine learning models face challenges in\neffectively learning polymer representations due to low-quality polymer\ndatasets, which consequently impact their overall performance. This study\nfocuses on improving polymer property performance prediction tasks by\nreconstructing an optimal and explainable descriptor representation space.\nNevertheless, prior research such as feature engineering and representation\nlearning can only partially solve this task since they are either\nlabor-incentive or unexplainable. This raises two issues: 1) automatic\ntransformation and 2) explainable enhancement. To tackle these issues, we\npropose our unique Traceable Group-wise Reinforcement Generation Perspective.\nSpecifically, we redefine the reconstruction of the representation space into\nan interactive process, combining nested generation and selection. Generation\ncreates meaningful descriptors, and selection eliminates redundancies to\ncontrol descriptor sizes. Our approach employs cascading reinforcement learning\nwith three Markov Decision Processes, automating descriptor and operation\nselection, and descriptor crossing. We utilize a group-wise generation strategy\nto explore and enhance reward signals for cascading agents. Ultimately, we\nconduct experiments to indicate the effectiveness of our proposed framework.\n", "date": "2024-09-23", "categories": []}
{"id": "2409.13390", "title": "Hydrogen under Pressure as a Benchmark for Machine-Learning Interatomic\n  Potentials", "abstract": "  Machine-learning interatomic potentials (MLPs) are fast, data-driven\nsurrogate models of atomistic systems' potential energy surfaces that can\naccelerate ab-initio molecular dynamics (MD) simulations by several orders of\nmagnitude. The performance of MLPs is commonly measured as the prediction error\nin energies and forces on data not used in their training. While low prediction\nerrors on a test set are necessary, they do not guarantee good performance in\nMD simulations. The latter requires physically motivated performance measures\nobtained from running accelerated simulations. However, the adoption of such\nmeasures has been limited by the effort and domain knowledge required to\ncalculate and interpret them.\n  To overcome this limitation, we present a benchmark that automatically\nquantifies the performance of MLPs in MD simulations of a liquid-liquid phase\ntransition in hydrogen under pressure, a challenging benchmark system. The\nbenchmark's h-llpt-24 dataset provides reference geometries, energies, forces,\nand stresses from density functional theory MD simulations at different\ntemperatures and mass densities. The benchmark's Python code automatically runs\nMLP-accelerated MD simulations and calculates, quantitatively compares and\nvisualizes pressures, stable molecular fractions, diffusion coefficients, and\nradial distribution functions. Employing this benchmark, we show that several\nstate-of-the-art MLPs fail to reproduce the liquid-liquid phase transition.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.12244", "title": "Sparks of Artificial General Intelligence(AGI) in Semiconductor Material\n  Science: Early Explorations into the Next Frontier of Generative AI-Assisted\n  Electron Micrograph Analysis", "abstract": "  Characterizing materials with electron micrographs poses significant\nchallenges for automated labeling due to the complex nature of nanomaterial\nstructures. To address this, we introduce a fully automated, end-to-end\npipeline that leverages recent advances in Generative AI. It is designed for\nanalyzing and understanding the microstructures of semiconductor materials with\neffectiveness comparable to that of human experts, contributing to the pursuit\nof Artificial General Intelligence (AGI) in nanomaterial identification. Our\napproach utilizes Large MultiModal Models (LMMs) such as GPT-4V, alongside\ntext-to-image models like DALLE-3. We integrate a GPT-4 guided Visual Question\nAnswering (VQA) method to analyze nanomaterial images, generate synthetic\nnanomaterial images via DALLE-3, and employ in-context learning with few-shot\nprompting in GPT-4V for accurate nanomaterial identification. Our method\nsurpasses traditional techniques by enhancing the precision of nanomaterial\nidentification and optimizing the process for high-throughput screening.\n", "date": "2024-09-17", "categories": []}
{"id": "2409.09192", "title": "Automated design of nonreciprocal thermal emitters via Bayesian\n  optimization", "abstract": "  Nonreciprocal thermal emitters that break Kirchhoff's law of thermal\nradiation promise exciting applications for thermal and energy applications.\nThe design of the bandwidth and angular range of the nonreciprocal effect,\nwhich directly affects the performance of nonreciprocal emitters, typically\nrelies on physical intuition. In this study, we present a general numerical\napproach to maximize the nonreciprocal effect. We choose doped magneto-optic\nmaterials and magnetic Weyl semimetal materials as model materials and focus on\npattern-free multilayer structures. The optimization randomly starts from a\nless effective structure and incrementally improves the broadband\nnonreciprocity through the combination of Bayesian optimization and\nreparameterization. Optimization results show that the proposed approach can\ndiscover structures that can achieve broadband nonreciprocal emission at\nwavelengths from 5 to 40 micrometers using only a fewer layers, significantly\noutperforming current state-of-the-art designs based on intuition in terms of\nboth performance and simplicity.\n", "date": "2024-09-13", "categories": []}
{"id": "2409.09583", "title": "Machine learning assisted screening of metal binary alloys for anode\n  materials", "abstract": "  In the dynamic and rapidly advancing battery field, alloy anode materials are\na focal point due to their superior electrochemical performance. Traditional\nscreening methods are inefficient and time-consuming. Our research introduces a\nmachine learning-assisted strategy to expedite the discovery and optimization\nof these materials. We compiled a vast dataset from the MP and AFLOW databases,\nencompassing tens of thousands of alloy compositions and properties. Utilizing\na CGCNN, we accurately predicted the potential and specific capacity of alloy\nanodes, validated against experimental data. This approach identified\napproximately 120 low potential and high specific capacity alloy anodes\nsuitable for various battery systems including Li, Na, K, Zn, Mg, Ca, and\nAl-based. Our method not only streamlines the screening of battery anode\nmaterials but also propels the advancement of battery material research and\ninnovation in energy storage technology.\n", "date": "2024-09-15", "categories": []}
{"id": "2409.09691", "title": "Extrapolative ML Models for Copolymers", "abstract": "  Machine learning models have been progressively used for predicting materials\nproperties. These models can be built using pre-existing data and are useful\nfor rapidly screening the physicochemical space of a material, which is\nastronomically large. However, ML models are inherently interpolative, and\ntheir efficacy for searching candidates outside a material's known range of\nproperty is unresolved. Moreover, the performance of an ML model is intricately\nconnected to its learning strategy and the volume of training data. Here, we\ndetermine the relationship between the extrapolation ability of an ML model,\nthe size and range of its training dataset, and its learning approach. We focus\non a canonical problem of predicting the properties of a copolymer as a\nfunction of the sequence of its monomers. Tree search algorithms, which learn\nthe similarity between polymer structures, are found to be inefficient for\nextrapolation. Conversely, the extrapolation capability of neural networks and\nXGBoost models, which attempt to learn the underlying functional correlation\nbetween the structure and property of polymers, show strong correlations with\nthe volume and range of training data. These findings have important\nimplications on ML-based new material development.\n", "date": "2024-09-15", "categories": []}
{"id": "2311.00787", "title": "Accelerating Electronic Stopping Power Predictions by 10 Million Times\n  with a Combination of Time-Dependent Density Functional Theory and Machine\n  Learning", "abstract": "  Knowing the rate at which particle radiation releases energy in a material,\nthe stopping power, is key to designing nuclear reactors, medical treatments,\nsemiconductor and quantum materials, and many other technologies. While the\nnuclear contribution to stopping power, i.e., elastic scattering between atoms,\nis well understood in the literature, the route for gathering data on the\nelectronic contribution has for decades remained costly and reliant on many\nsimplifying assumptions, including that materials are isotropic. We establish a\nmethod that combines time-dependent density functional theory (TDDFT) and\nmachine learning to reduce the time to assess new materials to mere hours on a\nsupercomputer and provides valuable data on how atomic details influence\nelectronic stopping. Our approach uses TDDFT to compute the electronic stopping\ncontributions to stopping power from first principles in several directions and\nthen machine learning to interpolate to other directions at a cost of 10\nmillion times fewer core-hours. We demonstrate the combined approach in a study\nof proton irradiation in aluminum and employ it to predict how the depth of\nmaximum energy deposition, the \"Bragg Peak,\" varies depending on incident angle\n-- a quantity otherwise inaccessible to modelers. The lack of any experimental\ninformation requirement makes our method applicable to most materials, and its\nspeed makes it a prime candidate for enabling quantum-to-continuum models of\nradiation damage. The prospect of reusing valuable TDDFT data for training the\nmodel make our approach appealing for applications in the age of materials data\nscience.\n", "date": "2023-11-01", "categories": []}
{"id": "2409.07610", "title": "When More Data Hurts: Optimizing Data Coverage While Mitigating\n  Diversity Induced Underfitting in an Ultra-Fast Machine-Learned Potential", "abstract": "  Machine-learned interatomic potentials (MLIPs) are becoming an essential tool\nin materials modeling. However, optimizing the generation of training data used\nto parameterize the MLIPs remains a significant challenge. This is because\nMLIPs can fail when encountering local enviroments too different from those\npresent in the training data. The difficulty of determining \\textit{a priori}\nthe environments that will be encountered during molecular dynamics (MD)\nsimulation necessitates diverse, high-quality training data. This study\ninvestigates how training data diversity affects the performance of MLIPs using\nthe Ultra-Fast Force Field (UF$^3$) to model amorphous silicon nitride. We\nemploy expert and autonomously generated data to create the training data and\nfit four force-field variants to subsets of the data. Our findings reveal a\ncritical balance in training data diversity: insufficient diversity hinders\ngeneralization, while excessive diversity can exceed the MLIP's learning\ncapacity, reducing simulation accuracy. Specifically, we found that the UF$^3$\nvariant trained on a subset of the training data, in which nitrogen-rich\nstructures were removed, offered vastly better prediction and simulation\naccuracy than any other variant. By comparing these UF$^3$ variants, we\nhighlight the nuanced requirements for creating accurate MLIPs, emphasizing the\nimportance of application-specific training data to achieve optimal performance\nin modeling complex material behaviors.\n", "date": "2024-09-11", "categories": []}
{"id": "2409.06756", "title": "Beyond designer's knowledge: Generating materials design hypotheses via\n  large language models", "abstract": "  Materials design often relies on human-generated hypotheses, a process\ninherently limited by cognitive constraints such as knowledge gaps and limited\nability to integrate and extract knowledge implications, particularly when\nmultidisciplinary expertise is required. This work demonstrates that large\nlanguage models (LLMs), coupled with prompt engineering, can effectively\ngenerate non-trivial materials hypotheses by integrating scientific principles\nfrom diverse sources without explicit design guidance by human experts. These\ninclude design ideas for high-entropy alloys with superior cryogenic properties\nand halide solid electrolytes with enhanced ionic conductivity and formability.\nThese design ideas have been experimentally validated in high-impact\npublications in 2023 not available in the LLM training data, demonstrating the\nLLM's ability to generate highly valuable and realizable innovative ideas not\nestablished in the literature. Our approach primarily leverages materials\nsystem charts encoding processing-structure-property relationships, enabling\nmore effective data integration by condensing key information from numerous\npapers, and evaluation and categorization of numerous hypotheses for human\ncognition, both through the LLM. This LLM-driven approach opens the door to new\navenues of artificial intelligence-driven materials discovery by accelerating\ndesign, democratizing innovation, and expanding capabilities beyond the\ndesigner's direct knowledge.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.06762", "title": "Generative Hierarchical Materials Search", "abstract": "  Generative models trained at scale can now produce text, video, and more\nrecently, scientific data such as crystal structures. In applications of\ngenerative approaches to materials science, and in particular to crystal\nstructures, the guidance from the domain expert in the form of high-level\ninstructions can be essential for an automated system to output candidate\ncrystals that are viable for downstream research. In this work, we formulate\nend-to-end language-to-structure generation as a multi-objective optimization\nproblem, and propose Generative Hierarchical Materials Search (GenMS) for\ncontrollable generation of crystal structures. GenMS consists of (1) a language\nmodel that takes high-level natural language as input and generates\nintermediate textual information about a crystal (e.g., chemical formulae), and\n(2) a diffusion model that takes intermediate information as input and\ngenerates low-level continuous value crystal structures. GenMS additionally\nuses a graph neural network to predict properties (e.g., formation energy) from\nthe generated crystal structures. During inference, GenMS leverages all three\ncomponents to conduct a forward tree search over the space of possible\nstructures. Experiments show that GenMS outperforms other alternatives of\ndirectly using language models to generate structures both in satisfying user\nrequest and in generating low-energy structures. We confirm that GenMS is able\nto generate common crystal structures such as double perovskites, or spinels,\nsolely from natural language input, and hence can form the foundation for more\ncomplex structure generation in near future.\n", "date": "2024-09-10", "categories": []}
{"id": "2405.09052", "title": "Dielectric Tensor Prediction for Inorganic Materials Using Latent\n  Information from Preferred Potential", "abstract": "  Dielectrics are crucial for technologies like flash memory, CPUs,\nphotovoltaics, and capacitors, but public data on these materials are scarce,\nrestricting research and development. Existing machine learning models have\nfocused on predicting scalar polycrystalline dielectric constants, neglecting\nthe directional nature of dielectric tensors essential for material design.\nThis study leverages multi-rank equivariant structural embeddings from a\nuniversal neural network potential to enhance predictions of dielectric\ntensors. We develop an equivariant readout decoder to predict total,\nelectronic, and ionic dielectric tensors while preserving O(3) equivariance,\nand benchmark its performance against state-of-the-art algorithms. Virtual\nscreening of thermodynamically stable materials from Materials Project for two\ndiscovery tasks, high-dielectric and highly anisotropic materials, identifies\npromising candidates including Cs2Ti(WO4)3 (band gap $E_g=2.93 \\mathrm{eV}$,\ndielectric constant $\\varepsilon=180.90$) and CsZrCuSe3 (anisotropic ratio\n$\\alpha_r = 121.89$). The results demonstrate our model's accuracy in\npredicting dielectric tensors and its potential for discovering novel\ndielectric materials.\n", "date": "2024-05-15", "categories": []}
{"id": "2409.06498", "title": "Learning local and semi-local density functionals from exact\n  exchange-correlation potentials and energies", "abstract": "  Finding accurate exchange-correlation (XC) functionals remains the defining\nchallenge in density functional theory (DFT). Despite 40 years of active\ndevelopment, the desired chemical accuracy is still elusive with existing\nfunctionals. We present a data-driven pathway to learn the XC functionals by\nutilizing the exact density, XC energy, and XC potential. While the exact\ndensities are obtained from accurate configuration interaction (CI), the exact\nXC energies and XC potentials are obtained via inverse DFT calculations on the\nCI densities. We demonstrate how simple neural network (NN) based local density\napproximation (LDA) and generalized gradient approximation (GGA), trained on\njust five atoms and two molecules, provide remarkable improvement in total\nenergies, densities, atomization energies, and barrier heights for hundreds of\nmolecules outside the training set. Particularly, the NN-based GGA functional\nattains similar accuracy as the higher rung SCAN meta-GGA, highlighting the\npromise of using the XC potential in modeling XC functionals. We expect this\napproach to pave the way for systematic learning of increasingly accurate and\nsophisticated XC functionals.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.02648", "title": "Creating a Microstructure Latent Space with Rich Material Information\n  for Multiphase Alloy Design", "abstract": "  The intricate microstructure serves as the cornerstone for the\ncomposition/processing-structure-property (CPSP) connection in multiphase\nalloys. Traditional alloy design methods often overlook microstructural\ndetails, which diminishes the reliability and effectiveness of the outcomes.\nThis study introduces an improved alloy design algorithm that integrates\nauthentic microstructural information to establish precise CPSP relationships.\nThe approach utilizes a deep-learning framework based on a variational\nautoencoder to map real microstructural data to a latent space, enabling the\nprediction of composition, processing steps, and material properties from the\nlatent space vector. By integrating this deep learning model with a specific\nsampling strategy in the latent space, a novel, microstructure-centered\nalgorithm for multiphase alloy design is developed. This algorithm is\ndemonstrated through the design of a unified dual-phase steel, and the results\nare assessed at three performance levels. Moreover, an exploration into the\nlatent vector space of the model highlights its seamless interpolation ability\nand its rich material information content. Notably, the current configuration\nof the latent space is particularly advantageous for alloy design, offering an\nexhaustive representation of microstructure, composition, processing, and\nproperty variations essential for multiphase alloys.\n", "date": "2024-09-04", "categories": []}
{"id": "2408.17255", "title": "Self-supervised learning for crystal property prediction via denoising", "abstract": "  Accurate prediction of the properties of crystalline materials is crucial for\ntargeted discovery, and this prediction is increasingly done with data-driven\nmodels. However, for many properties of interest, the number of materials for\nwhich a specific property has been determined is much smaller than the number\nof known materials. To overcome this disparity, we propose a novel\nself-supervised learning (SSL) strategy for material property prediction. Our\napproach, crystal denoising self-supervised learning (CDSSL), pretrains\npredictive models (e.g., graph networks) with a pretext task based on\nrecovering valid material structures when given perturbed versions of these\nstructures. We demonstrate that CDSSL models out-perform models trained without\nSSL, across material types, properties, and dataset sizes.\n", "date": "2024-08-30", "categories": []}
{"id": "2408.15601", "title": "Grand canonical generative diffusion model for crystalline phases and\n  grain boundaries", "abstract": "  The diffusion model has emerged as a powerful tool for generating atomic\nstructures for materials science. This work calls attention to the deficiency\nof current particle-based diffusion models, which represent atoms as a point\ncloud, in generating even the simplest ordered crystalline structures. The\nproblem is attributed to particles being trapped in local minima during the\nscore-driven simulated annealing of the diffusion process, similar to the\nphysical process of force-driven simulated annealing. We develop a solution,\nthe grand canonical diffusion model, which adopts an alternative voxel-based\nrepresentation with continuous rather than fixed number of particles. The\nmethod is applied towards generation of several common crystalline phases as\nwell as the technologically important and challenging problem of grain boundary\nstructures.\n", "date": "2024-08-28", "categories": []}
{"id": "2408.14488", "title": "Multi-Task Multi-Fidelity Learning of Properties for Energetic Materials", "abstract": "  Data science and artificial intelligence are playing an increasingly\nimportant role in the physical sciences. Unfortunately, in the field of\nenergetic materials data scarcity limits the accuracy and even applicability of\nML tools. To address data limitations, we compiled multi-modal data: both\nexperimental and computational results for several properties. We find that\nmulti-task neural networks can learn from multi-modal data and outperform\nsingle-task models trained for specific properties. As expected, the\nimprovement is more significant for data-scarce properties. These models are\ntrained using descriptors built from simple molecular information and can be\nreadily applied for large-scale materials screening to explore multiple\nproperties simultaneously. This approach is widely applicable to fields outside\nenergetic materials.\n", "date": "2024-08-21", "categories": []}
{"id": "2408.15097", "title": "Data-Driven Nonlinear Deformation Design of 3D-Printable Shells", "abstract": "  Designing and fabricating structures with specific mechanical properties\nrequires understanding the intricate relationship between design parameters and\nperformance. Understanding the design-performance relationship becomes\nincreasingly complicated for nonlinear deformations. Though successful at\nmodeling elastic deformations, simulation-based techniques struggle to model\nlarge elastoplastic deformations exhibiting plasticity and densification. We\npropose a neural network trained on experimental data to learn the\ndesign-performance relationship between 3D-printable shells and their\ncompressive force-displacement behavior. Trained on thousands of physical\nexperiments, our network aids in both forward and inverse design to generate\nshells exhibiting desired elastoplastic and hyperelastic deformations. We\nvalidate a subset of generated designs through fabrication and testing.\nFurthermore, we demonstrate the network's inverse design efficacy in generating\ncustom shells for several applications.\n", "date": "2024-08-27", "categories": []}
{"id": "2404.12899", "title": "Bayesian Co-navigation: Dynamic Designing of the Materials Digital Twins\n  via Active Learning", "abstract": "  Scientific advancement is universally based on the dynamic interplay between\ntheoretical insights, modelling, and experimental discoveries. However, this\nfeedback loop is often slow, including delayed community interactions and the\ngradual integration of experimental data into theoretical frameworks. This\nchallenge is particularly exacerbated in domains dealing with high-dimensional\nobject spaces, such as molecules and complex microstructures. Hence, the\nintegration of theory within automated and autonomous experimental setups, or\ntheory in the loop automated experiment, is emerging as a crucial objective for\naccelerating scientific research. The critical aspect is not only to use theory\nbut also on-the-fly theory updates during the experiment. Here, we introduce a\nmethod for integrating theory into the loop through Bayesian co-navigation of\ntheoretical model space and experimentation. Our approach leverages the\nconcurrent development of surrogate models for both simulation and experimental\ndomains at the rates determined by latencies and costs of experiments and\ncomputation, alongside the adjustment of control parameters within theoretical\nmodels to minimize epistemic uncertainty over the experimental object spaces.\nThis methodology facilitates the creation of digital twins of material\nstructures, encompassing both the surrogate model of behavior that includes the\ncorrelative part and the theoretical model itself. While demonstrated here\nwithin the context of functional responses in ferroelectric materials, our\napproach holds promise for broader applications, the exploration of optical\nproperties in nanoclusters, microstructure-dependent properties in complex\nmaterials, and properties of molecular systems. The analysis code that supports\nthe funding is publicly available at\nhttps://github.com/Slautin/2024_Co-navigation/tree/main\n", "date": "2024-04-19", "categories": []}
{"id": "2408.13532", "title": "FFT-based surrogate modeling of auxetic metamaterials with real-time\n  prediction of effective elastic properties and swift inverse design", "abstract": "  Auxetic structures, known for their negative Poisson's ratio, exhibit\neffective elastic properties heavily influenced by their underlying structural\ngeometry and base material properties. While periodic homogenization of auxetic\nunit cells can be used to investigate these properties, it is computationally\nexpensive and limits design space exploration and inverse analysis. In this\npaper, surrogate models are developed for the real-time prediction of the\neffective elastic properties of auxetic unit cells with orthogonal voids of\ndifferent shapes. The unit cells feature orthogonal voids in four distinct\nshapes, including rectangular, diamond, oval, and peanut-shaped voids, each\ncharacterized by specific void diameters. The generated surrogate models accept\ngeometric parameters and the elastic properties of the base material as inputs\nto predict the effective elastic constants in real-time. This rapid evaluation\nenables a practical inverse analysis framework for obtaining the optimal design\nparameters that yield the desired effective response. The fast Fourier\ntransform (FFT)-based homogenization approach is adopted to efficiently\ngenerate data for developing the surrogate models, bypassing concerns about\nperiodic mesh generation and boundary conditions typically associated with the\nfinite element method (FEM). The performance of the generated surrogate models\nis rigorously examined through a train/test split methodology, a parametric\nstudy, and an inverse problem. Finally, a graphical user interface (GUI) is\ndeveloped, offering real-time prediction of the effective tangent stiffness and\nperforming inverse analysis to determine optimal geometric parameters.\n", "date": "2024-08-24", "categories": []}
{"id": "2311.06228", "title": "Learning material synthesis-process-structure-property relationship by\n  data fusion: Bayesian Coregionalization N-Dimensional Piecewise Function\n  Learning", "abstract": "  Autonomous materials research labs require the ability to combine and learn\nfrom diverse data streams. This is especially true for learning material\nsynthesis-process-structure-property relationships, key to accelerating\nmaterials optimization and discovery as well as accelerating mechanistic\nunderstanding. We present the Synthesis-process-structure-property relAtionship\ncoreGionalized lEarner (SAGE) algorithm. A fully Bayesian algorithm that uses\nmultimodal coregionalization to merge knowledge across data sources to learn\nsynthesis-process-structure-property relationships. SAGE outputs a\nprobabilistic posterior for the relationships including the most likely\nrelationships given the data.\n", "date": "2023-11-10", "categories": []}
{"id": "2408.09297", "title": "Out-of-distribution materials property prediction using adversarial\n  learning based fine-tuning", "abstract": "  The accurate prediction of material properties is crucial in a wide range of\nscientific and engineering disciplines. Machine learning (ML) has advanced the\nstate of the art in this field, enabling scientists to discover novel materials\nand design materials with specific desired properties. However, one major\nchallenge that persists in material property prediction is the generalization\nof models to out-of-distribution (OOD) samples,i.e., samples that differ\nsignificantly from those encountered during training. In this paper, we explore\nthe application of advancements in OOD learning approaches to enhance the\nrobustness and reliability of material property prediction models. We propose\nand apply the Crystal Adversarial Learning (CAL) algorithm for OOD materials\nproperty prediction,which generates synthetic data during training to bias the\ntraining towards those samples with high prediction uncertainty. We further\npropose an adversarial learning based targeting finetuning approach to make the\nmodel adapted to a particular OOD dataset, as an alternative to traditional\nfine-tuning. Our experiments demonstrate the success of our CAL algorithm with\nits high effectiveness in ML with limited samples which commonly occurs in\nmaterials science. Our work represents a promising direction toward better OOD\nlearning and materials property prediction.\n", "date": "2024-08-17", "categories": []}
{"id": "2408.08428", "title": "Phononic materials with effectively scale-separated hierarchical\n  features using interpretable machine learning", "abstract": "  Manipulating the dispersive characteristics of vibrational waves is\nbeneficial for many applications, e.g., high-precision instruments. architected\nhierarchical phononic materials have sparked promise tunability of\nelastodynamic waves and vibrations over multiple frequency ranges. In this\narticle, hierarchical unit-cells are obtained, where features at each length\nscale result in a band gap within a targeted frequency range. Our novel\napproach, the ``hierarchical unit-cell template method,'' is an interpretable\nmachine-learning approach that uncovers global unit-cell shape/topology\npatterns corresponding to predefined band-gap objectives. A scale-separation\neffect is observed where the coarse-scale band-gap objective is mostly\nunaffected by the fine-scale features despite the closeness of their length\nscales, thus enabling an efficient hierarchical algorithm. Moreover, the\nhierarchical patterns revealed are not predefined or self-similar hierarchies\nas common in current hierarchical phononic materials. Thus, our approach offers\na flexible and efficient method for the exploration of new regions in the\nhierarchical design space, extracting minimal effective patterns for inverse\ndesign in applications targeting multiple frequency ranges.\n", "date": "2024-08-15", "categories": []}
{"id": "2408.08526", "title": "Inverse design with conditional cascaded diffusion models", "abstract": "  Adjoint-based design optimizations are usually computationally expensive and\nthose costs scale with resolution. To address this, researchers have proposed\nmachine learning approaches for inverse design that can predict\nhigher-resolution solutions from lower cost/resolution ones. Due to the recent\nsuccess of diffusion models over traditional generative models, we extend the\nuse of diffusion models for multi-resolution tasks by proposing the conditional\ncascaded diffusion model (cCDM). Compared to GANs, cCDM is more stable to\ntrain, and each diffusion model within the cCDM can be trained independently,\nthus each model's parameters can be tuned separately to maximize the\nperformance of the pipeline. Our study compares cCDM against a cGAN model with\ntransfer learning.\n  Our results demonstrate that the cCDM excels in capturing finer details,\npreserving volume fraction constraints, and minimizing compliance errors in\nmulti-resolution tasks when a sufficient amount of high-resolution training\ndata (more than 102 designs) is available. Furthermore, we explore the impact\nof training data size on the performance of both models. While both models show\ndecreased performance with reduced high-resolution training data, the cCDM\nloses its superiority to the cGAN model with transfer learning when training\ndata is limited (less than 102), and we show the break-even point for this\ntransition. Also, we highlight that while the diffusion model may achieve\nbetter pixel-wise performance in both low-resolution and high-resolution\nscenarios, this does not necessarily guarantee that the model produces optimal\ncompliance error or constraint satisfaction.\n", "date": "2024-08-16", "categories": []}
{"id": "2408.05237", "title": "Biomimetic Machine Learning approach for prediction of mechanical\n  properties of Additive Friction Stir Deposited Aluminum alloys based walled\n  structures", "abstract": "  This study presents a novel approach to predicting mechanical properties of\nAdditive Friction Stir Deposited (AFSD) aluminum alloy walled structures using\nbiomimetic machine learning. The research combines numerical modeling of the\nAFSD process with genetic algorithm-optimized machine learning models to\npredict von Mises stress and logarithmic strain. Finite element analysis was\nemployed to simulate the AFSD process for five aluminum alloys: AA2024, AA5083,\nAA5086, AA7075, and AA6061, capturing complex thermal and mechanical\ninteractions. A dataset of 200 samples was generated from these simulations.\nSubsequently, Decision Tree (DT) and Random Forest (RF) regression models,\noptimized using genetic algorithms, were developed to predict key mechanical\nproperties. The GA-RF model demonstrated superior performance in predicting\nboth von Mises stress (R square = 0.9676) and logarithmic strain (R square =\n0.7201). This innovative approach provides a powerful tool for understanding\nand optimizing the AFSD process across multiple aluminum alloys, offering\ninsights into material behavior under various process parameters.\n", "date": "2024-08-05", "categories": []}
{"id": "2408.06300", "title": "Inverse designing metamaterials with programmable nonlinear functional\n  responses in graph space", "abstract": "  Material responses to static and dynamic stimuli, represented as nonlinear\ncurves, are design targets for engineering functionalities like structural\nsupport, impact protection, and acoustic and photonic bandgaps.\nThree-dimensional metamaterials offer significant tunability due to their\ninternal structure, yet existing methods struggle to capture their complex\nbehavior-to-structure relationships. We present GraphMetaMat, a graph-based\nframework capable of designing three-dimensional metamaterials with\nprogrammable responses and arbitrary manufacturing constraints. Integrating\ngraph networks, physics biases, reinforcement learning, and tree search,\nGraphMetaMat can target stress-strain curves spanning four orders of magnitude\nand complex behaviors, as well as viscoelastic transmission responses with\nvarying attenuation gaps. GraphMetaMat can create cushioning materials for\nprotective equipment and vibration-damping panels for electric vehicles,\noutperforming commercial materials, and enabling the automatic design of\nmaterials with on-demand functionalities.\n", "date": "2024-08-12", "categories": []}
{"id": "2408.04042", "title": "Scaling Law of Sim2Real Transfer Learning in Expanding Computational\n  Materials Databases for Real-World Predictions", "abstract": "  To address the challenge of limited experimental materials data, extensive\nphysical property databases are being developed based on high-throughput\ncomputational experiments, such as molecular dynamics simulations. Previous\nstudies have shown that fine-tuning a predictor pretrained on a computational\ndatabase to a real system can result in models with outstanding generalization\ncapabilities compared to learning from scratch. This study demonstrates the\nscaling law of simulation-to-real (Sim2Real) transfer learning for several\nmachine learning tasks in materials science. Case studies of three prediction\ntasks for polymers and inorganic materials reveal that the prediction error on\nreal systems decreases according to a power-law as the size of the\ncomputational data increases. Observing the scaling behavior offers various\ninsights for database development, such as determining the sample size\nnecessary to achieve a desired performance, identifying equivalent sample sizes\nfor physical and computational experiments, and guiding the design of data\nproduction protocols for downstream real-world tasks.\n", "date": "2024-08-07", "categories": []}
{"id": "2402.19462", "title": "Accelerating materials discovery for polymer solar cells: Data-driven\n  insights enabled by natural language processing", "abstract": "  We present a simulation of various active learning strategies for the\ndiscovery of polymer solar cell donor/acceptor pairs using data extracted from\nthe literature spanning $\\sim$20 years by a natural language processing\npipeline. While data-driven methods have been well established to discover\nnovel materials faster than Edisonian trial-and-error approaches, their\nbenefits have not been quantified for material discovery problems that can take\ndecades. Our approach demonstrates a potential reduction in discovery time by\napproximately 75 %, equivalent to a 15 year acceleration in material\ninnovation. Our pipeline enables us to extract data from greater than 3300\npapers which is $\\sim$5 times larger and therefore more diverse than similar\ndata sets reported by others. We also trained machine learning models to\npredict the power conversion efficiency and used our model to identify\npromising donor-acceptor combinations that are as yet unreported. We thus\ndemonstrate a pipeline that goes from published literature to extracted\nmaterial property data which in turn is used to obtain data-driven insights.\nOur insights include active learning strategies that can be used to train\nstrong predictive models of material properties or be robust to the initial\nmaterial system used. This work provides a valuable framework for data-driven\nresearch in materials science.\n", "date": "2024-02-29", "categories": []}
{"id": "2402.11103", "title": "Toward Learning Latent-Variable Representations of Microstructures by\n  Optimizing in Spatial Statistics Space", "abstract": "  In Materials Science, material development involves evaluating and optimizing\nthe internal structures of the material, generically referred to as\nmicrostructures. Microstructures structure is stochastic, analogously to image\ntextures. A particular microstructure can be well characterized by its spatial\nstatistics, analogously to image texture being characterized by the response to\na Fourier-like filter bank. Material design would benefit from low-dimensional\nrepresentation of microstructures Paulson et al. (2017).\n  In this work, we train a Variational Autoencoders (VAE) to produce\nreconstructions of textures that preserve the spatial statistics of the\noriginal texture, while not necessarily reconstructing the same image in data\nspace. We accomplish this by adding a differentiable term to the cost function\nin order to minimize the distance between the original and the reconstruction\nin spatial statistics space.\n  Our experiments indicate that it is possible to train a VAE that minimizes\nthe distance in spatial statistics space between the original and the\nreconstruction of synthetic images. In future work, we will apply the same\ntechniques to microstructures, with the goal of obtaining low-dimensional\nrepresentations of material microstructures.\n", "date": "2024-02-16", "categories": []}
{"id": "2406.02591", "title": "Unveiling the Potential of AI for Nanomaterial Morphology Prediction", "abstract": "  Creation of nanomaterials with specific morphology remains a complex\nexperimental process, even though there is a growing demand for these materials\nin various industry sectors. This study explores the potential of AI to predict\nthe morphology of nanoparticles within the data availability constraints. For\nthat, we first generated a new multi-modal dataset that is double the size of\nanalogous studies. Then, we systematically evaluated performance of classical\nmachine learning and large language models in prediction of nanomaterial shapes\nand sizes. Finally, we prototyped a text-to-image system, discussed the\nobtained empirical results, as well as the limitations and promises of existing\napproaches.\n", "date": "2024-05-31", "categories": []}
{"id": "2402.07472", "title": "Cartesian atomic cluster expansion for machine learning interatomic\n  potentials", "abstract": "  Machine learning interatomic potentials are revolutionizing large-scale,\naccurate atomistic modelling in material science and chemistry. Many potentials\nuse atomic cluster expansion or equivariant message passing frameworks. Such\nframeworks typically use spherical harmonics as angular basis functions, and\nthen use Clebsch-Gordan contraction to maintain rotational symmetry, which may\nintroduce redundancies in representations and computational overhead. We\npropose an alternative: a Cartesian-coordinates-based atomic density expansion.\nThis approach provides a complete set of polynormially indepedent features of\natomic environments while maintaining interaction body orders. Additionally, we\nintegrate low-dimensional embeddings of various chemical elements and\ninter-atomic message passing. The resulting potential, named Cartesian Atomic\nCluster Expansion (CACE), exhibits good accuracy, stability, and\ngeneralizability. We validate its performance in diverse systems, including\nbulk water, small molecules, and 25-element high-entropy alloys.\n", "date": "2024-02-12", "categories": []}
{"id": "2407.20251", "title": "An Uncertainty-aware Deep Learning Framework-based Robust Design\n  Optimization of Metamaterial Units", "abstract": "  Mechanical metamaterials represent an innovative class of artificial\nstructures, distinguished by their extraordinary mechanical characteristics,\nwhich are beyond the scope of traditional natural materials. The use of deep\ngenerative models has become increasingly popular in the design of metamaterial\nunits. The effectiveness of using deep generative models lies in their capacity\nto compress complex input data into a simplified, lower-dimensional latent\nspace, while also enabling the creation of novel optimal designs through\nsampling within this space. However, the design process does not take into\naccount the effect of model uncertainty due to data sparsity or the effect of\ninput data uncertainty due to inherent randomness in the data. This might lead\nto the generation of undesirable structures with high sensitivity to the\nuncertainties in the system. To address this issue, a novel uncertainty-aware\ndeep learning framework-based robust design approach is proposed for the design\nof metamaterial units with optimal target properties. The proposed approach\nutilizes the probabilistic nature of the deep learning framework and quantifies\nboth aleatoric and epistemic uncertainties associated with surrogate-based\ndesign optimization. We demonstrate that the proposed design approach is\ncapable of designing high-performance metamaterial units with high reliability.\nTo showcase the effectiveness of the proposed design approach, a\nsingle-objective design optimization problem and a multi-objective design\noptimization problem are presented. The optimal robust designs obtained are\nvalidated by comparing them to the designs obtained from the topology\noptimization method as well as the designs obtained from a deterministic deep\nlearning framework-based design optimization where none of the uncertainties in\nthe system are explicitly considered.\n", "date": "2024-07-19", "categories": []}
{"id": "2405.01491", "title": "FeNNol: an Efficient and Flexible Library for Building\n  Force-field-enhanced Neural Network Potentials", "abstract": "  Neural network interatomic potentials (NNPs) have recently proven to be\npowerful tools to accurately model complex molecular systems while bypassing\nthe high numerical cost of ab-initio molecular dynamics simulations. In recent\nyears, numerous advances in model architectures as well as the development of\nhybrid models combining machine-learning (ML) with more traditional,\nphysically-motivated, force-field interactions have considerably increased the\ndesign space of ML potentials. In this paper, we present FeNNol, a new library\nfor building, training and running force-field-enhanced neural network\npotentials. It provides a flexible and modular system for building hybrid\nmodels, allowing to easily combine state-of-the-art embeddings with\nML-parameterized physical interaction terms without the need for explicit\nprogramming. Furthermore, FeNNol leverages the automatic differentiation and\njust-in-time compilation features of the Jax Python library to enable fast\nevaluation of NNPs, shrinking the performance gap between ML potentials and\nstandard force-fields. This is demonstrated with the popular ANI-2x model\nreaching simulation speeds nearly on par with the AMOEBA polarizable\nforce-field on commodity GPUs (GPU=Graphics processing unit). We hope that\nFeNNol will facilitate the development and application of new hybrid NNP\narchitectures for a wide range of molecular simulation problems.\n", "date": "2024-05-02", "categories": []}
{"id": "2406.04727", "title": "MMPolymer: A Multimodal Multitask Pretraining Framework for Polymer\n  Property Prediction", "abstract": "  Polymers are high-molecular-weight compounds constructed by the covalent\nbonding of numerous identical or similar monomers so that their 3D structures\nare complex yet exhibit unignorable regularity. Typically, the properties of a\npolymer, such as plasticity, conductivity, bio-compatibility, and so on, are\nhighly correlated with its 3D structure. However, existing polymer property\nprediction methods heavily rely on the information learned from polymer SMILES\nsequences (P-SMILES strings) while ignoring crucial 3D structural information,\nresulting in sub-optimal performance. In this work, we propose MMPolymer, a\nnovel multimodal multitask pretraining framework incorporating polymer 1D\nsequential and 3D structural information to encourage downstream polymer\nproperty prediction tasks. Besides, considering the scarcity of polymer 3D\ndata, we further introduce the \"Star Substitution\" strategy to extract 3D\nstructural information effectively. During pretraining, in addition to\npredicting masked tokens and recovering clear 3D coordinates, MMPolymer\nachieves the cross-modal alignment of latent representations. Then we further\nfine-tune the pretrained MMPolymer for downstream polymer property prediction\ntasks in the supervised learning paradigm. Experiments show that MMPolymer\nachieves state-of-the-art performance in downstream property prediction tasks.\nMoreover, given the pretrained MMPolymer, utilizing merely a single modality in\nthe fine-tuning phase can also outperform existing methods, showcasing the\nexceptional capability of MMPolymer in polymer feature extraction and\nutilization.\n", "date": "2024-06-07", "categories": []}
{"id": "2407.18847", "title": "Enhancing material property prediction with ensemble deep graph\n  convolutional networks", "abstract": "  Machine learning (ML) models have emerged as powerful tools for accelerating\nmaterials discovery and design by enabling accurate predictions of properties\nfrom compositional and structural data. These capabilities are vital for\ndeveloping advanced technologies across fields such as energy, electronics, and\nbiomedicine, potentially reducing the time and resources needed for new\nmaterial exploration and promoting rapid innovation cycles. Recent efforts have\nfocused on employing advanced ML algorithms, including deep learning - based\ngraph neural network, for property prediction. Additionally, ensemble models\nhave proven to enhance the generalizability and robustness of ML and DL.\nHowever, the use of such ensemble strategies in deep graph networks for\nmaterial property prediction remains underexplored. Our research provides an\nin-depth evaluation of ensemble strategies in deep learning - based graph\nneural network, specifically targeting material property prediction tasks. By\ntesting the Crystal Graph Convolutional Neural Network (CGCNN) and its\nmultitask version, MT-CGCNN, we demonstrated that ensemble techniques,\nespecially prediction averaging, substantially improve precision beyond\ntraditional metrics for key properties like formation energy per atom ($\\Delta\nE^{f}$), band gap ($E_{g}$) and density ($\\rho$) in 33,990 stable inorganic\nmaterials. These findings support the broader application of ensemble methods\nto enhance predictive accuracy in the field.\n", "date": "2024-07-26", "categories": []}
{"id": "2407.15665", "title": "A spatiotemporal deep learning framework for prediction of crack\n  dynamics in heterogeneous solids: efficient mapping of concrete\n  microstructures to its fracture properties", "abstract": "  A spatiotemporal deep learning framework is proposed that is capable of 2D\nfull-field prediction of fracture in concrete mesostructures. This framework\nnot only predicts fractures but also captures the entire history of the\nfracture process, from the crack initiation in the interfacial transition zone\nto the subsequent propagation of the cracks in the mortar matrix. In addition,\na convolutional neural network is developed which can predict the averaged\nstress-strain curve of the mesostructures. The UNet modeling framework, which\ncomprises an encoder-decoder section with skip connections, is used as the deep\nlearning surrogate model. Training and test data are generated from\nhigh-fidelity fracture simulations of randomly generated concrete\nmesostructures. These mesostructures include geometric variabilities such as\ndifferent aggregate particle geometrical features, spatial distribution, and\nthe total volume fraction of aggregates. The fracture simulations are carried\nout in Abaqus, utilizing the cohesive phase-field fracture modeling technique\nas the fracture modeling approach. In this work, to reduce the number of\ntraining datasets, the spatial distribution of three sets of material\nproperties for three-phase concrete mesostructures, along with the spatial\nphase-field damage index, are fed to the UNet to predict the corresponding\nstress and spatial damage index at the subsequent step. It is shown that after\nthe training process using this methodology, the UNet model is capable of\naccurately predicting damage on the unseen test dataset by using 470 datasets.\nMoreover, another novel aspect of this work is the conversion of irregular\nfinite element data into regular grids using a developed pipeline. This\napproach allows for the implementation of less complex UNet architecture and\nfacilitates the integration of phase-field fracture equations into surrogate\nmodels for future developments.\n", "date": "2024-07-22", "categories": []}
{"id": "2310.19306", "title": "A Planning-and-Exploring Approach to Extreme-Mechanics Force Fields", "abstract": "  Extreme mechanical processes such as strong lattice distortion and bond\nbreakage during fracture are ubiquitous in nature and engineering, which often\nlead to catastrophic failure of structures. However, understanding the\nnucleation and growth of cracks is challenged by their multiscale\ncharacteristics spanning from atomic-level structures at the crack tip to the\nstructural features where the load is applied. Molecular simulations offer an\nimportant tool to resolve the progressive microstructural changes at crack\nfronts and are widely used to explore processes therein, such as mechanical\nenergy dissipation, crack path selection, and dynamic instabilities (e.g.,\nkinking, branching). Empirical force fields developed based on local\ndescriptors based on atomic positions and the bond orders do not yield\nsatisfying predictions of fracture, even for the nonlinear, anisotropic\nstress-strain relations and the energy densities of edges. High-fidelity force\nfields thus should include the tensorial nature of strain and the energetics of\nrare events during fracture, which, unfortunately, have not been taken into\naccount in both the state-of-the-art empirical and machine-learning force\nfields. Based on data generated by first-principles calculations, we develop a\nneural network-based force field for fracture, NN-F$^3$, by combining\npre-sampling of the space of strain states and active-learning techniques to\nexplore the transition states at critical bonding distances. The capability of\nNN-F$^3$ is demonstrated by studying the rupture of h-BN and twisted bilayer\ngraphene as model problems. The simulation results confirm recent experimental\nfindings and highlight the necessity to include the knowledge of electronic\nstructures from first-principles calculations in predicting extreme mechanical\nprocesses.\n", "date": "2023-10-30", "categories": []}
{"id": "2407.15175", "title": "${\\it Asparagus}$: A Toolkit for Autonomous, User-Guided Construction of\n  Machine-Learned Potential Energy Surfaces", "abstract": "  With the establishment of machine learning (ML) techniques in the scientific\ncommunity, the construction of ML potential energy surfaces (ML-PES) has become\na standard process in physics and chemistry. So far, improvements in the\nconstruction of ML-PES models have been conducted independently, creating an\ninitial hurdle for new users to overcome and complicating the reproducibility\nof results. Aiming to reduce the bar for the extensive use of ML-PES, we\nintroduce ${\\it Asparagus}$, a software package encompassing the different\nparts into one coherent implementation that allows an autonomous, user-guided\nconstruction of ML-PES models. ${\\it Asparagus}$ combines capabilities of\ninitial data sampling with interfaces to ${\\it ab\n  initio}$ calculation programs, ML model training, as well as model evaluation\nand its application within other codes such as ASE or CHARMM. The\nfunctionalities of the code are illustrated in different examples, including\nthe dynamics of small molecules, the representation of reactive potentials in\norganometallic compounds, and atom diffusion on periodic surface structures.\nThe modular framework of ${\\it Asparagus}$ is designed to allow simple\nimplementations of further ML-related methods and models to provide constant\nuser-friendly access to state-of-the-art ML techniques.\n", "date": "2024-07-21", "categories": []}
{"id": "2407.05333", "title": "Generating multi-scale NMC particles with radial grain architectures\n  using spatial stochastics and GANs", "abstract": "  Understanding structure-property relationships of Li-ion battery cathodes is\ncrucial for optimizing rate-performance and cycle-life resilience. However,\ncorrelating the morphology of cathode particles, such as in NMC811, and their\ninner grain architecture with electrode performance is challenging,\nparticularly, due to the significant length-scale difference between grain and\nparticle sizes. Experimentally, it is currently not feasible to image such a\nhigh number of particles with full granular detail to achieve representivity. A\nsecond challenge is that sufficiently high-resolution 3D imaging techniques\nremain expensive and are sparsely available at research institutions. To\naddress these challenges, a stereological generative adversarial network\n(GAN)-based model fitting approach is presented that can generate\nrepresentative 3D information from 2D data, enabling characterization of\nmaterials in 3D using cost-effective 2D data. Once calibrated, this multi-scale\nmodel is able to rapidly generate virtual cathode particles that are\nstatistically similar to experimental data, and thus is suitable for virtual\ncharacterization and materials testing through numerical simulations. A large\ndataset of simulated particles with inner grain architecture has been made\npublicly available.\n", "date": "2024-07-07", "categories": []}
{"id": "2407.10022", "title": "AtomAgents: Alloy design and discovery through physics-aware multi-modal\n  multi-agent artificial intelligence", "abstract": "  The design of alloys is a multi-scale problem that requires a holistic\napproach that involves retrieving relevant knowledge, applying advanced\ncomputational methods, conducting experimental validations, and analyzing the\nresults, a process that is typically reserved for human experts. Machine\nlearning (ML) can help accelerate this process, for instance, through the use\nof deep surrogate models that connect structural features to material\nproperties, or vice versa. However, existing data-driven models often target\nspecific material objectives, offering limited flexibility to integrate\nout-of-domain knowledge and cannot adapt to new, unforeseen challenges. Here,\nwe overcome these limitations by leveraging the distinct capabilities of\nmultiple AI agents that collaborate autonomously within a dynamic environment\nto solve complex materials design tasks. The proposed physics-aware generative\nAI platform, AtomAgents, synergizes the intelligence of large language models\n(LLM) the dynamic collaboration among AI agents with expertise in various\ndomains, including knowledge retrieval, multi-modal data integration,\nphysics-based simulations, and comprehensive results analysis across modalities\nthat includes numerical data and images of physical simulation results. The\nconcerted effort of the multi-agent system allows for addressing complex\nmaterials design problems, as demonstrated by examples that include\nautonomously designing metallic alloys with enhanced properties compared to\ntheir pure counterparts. Our results enable accurate prediction of key\ncharacteristics across alloys and highlight the crucial role of solid solution\nalloying to steer the development of advanced metallic alloys. Our framework\nenhances the efficiency of complex multi-objective design tasks and opens new\navenues in fields such as biomedical materials engineering, renewable energy,\nand environmental sustainability.\n", "date": "2024-07-13", "categories": []}
{"id": "2407.08074", "title": "Smooth Like Butter: Evaluating Multi-Lattice Transitions in\n  Property-Augmented Latent Spaces", "abstract": "  Additive manufacturing has revolutionized structural optimization by\nenhancing component strength and reducing material requirements. One approach\nused to achieve these improvements is the application of multi-lattice\nstructures, where the macro-scale performance relies on the detailed design of\nmesostructural lattice elements. Many current approaches to designing such\nstructures use data-driven design to generate multi-lattice transition regions,\nmaking use of machine learning models that are informed solely by the geometry\nof the mesostructures. However, it remains unclear if the integration of\nmechanical properties into the dataset used to train such machine learning\nmodels would be beneficial beyond using geometric data alone. To address this\nissue, this work implements and evaluates a hybrid geometry/property\nVariational Autoencoder (VAE) for generating multi-lattice transition regions.\nIn our study, we found that hybrid VAEs demonstrate enhanced performance in\nmaintaining stiffness continuity through transition regions, indicating their\nsuitability for design tasks requiring smooth mechanical properties.\n", "date": "2024-07-10", "categories": []}
{"id": "2407.08270", "title": "SciQu: Accelerating Materials Properties Prediction with Automated\n  Literature Mining for Self-Driving Laboratories", "abstract": "  Assessing different material properties to predict specific attributes, such\nas band gap, resistivity, young modulus, work function, and refractive index,\nis a fundamental requirement for materials science-based applications. However,\nthe process is time-consuming and often requires extensive literature reviews\nand numerous experiments. Our study addresses these challenges by leveraging\nmachine learning to analyze material properties with greater precision and\nefficiency. By automating the data extraction process and using the extracted\ninformation to train machine learning models, our developed model, SciQu,\noptimizes material properties. As a proof of concept, we predicted the\nrefractive index of materials using data extracted from numerous research\narticles with SciQu, considering input descriptors such as space group, volume,\nand bandgap with Root Mean Square Error (RMSE) 0.068 and R2 0.94. Thus, SciQu\nnot only predicts the properties of materials but also plays a key role in\nself-driving laboratories by optimizing the synthesis parameters to achieve\nprecise shape, size, and phase of the materials subjected to the input\nparameters.\n", "date": "2024-07-11", "categories": []}
{"id": "2312.03404", "title": "A cyclical route linking fundamental mechanism and AI algorithm: An\n  example from tuning Poisson's ratio in amorphous networks", "abstract": "  \"AI for science\" is widely recognized as a future trend in the development of\nscientific research. Currently, although machine learning algorithms have\nplayed a crucial role in scientific research with numerous successful cases,\nrelatively few instances exist where AI assists researchers in uncovering the\nunderlying physical mechanisms behind a certain phenomenon and subsequently\nusing that mechanism to improve machine learning algorithms' efficiency. This\narticle uses the investigation into the relationship between extreme Poisson's\nratio values and the structure of amorphous networks as a case study to\nillustrate how machine learning methods can assist in revealing underlying\nphysical mechanisms. Upon recognizing that the Poisson's ratio relies on the\nlow-frequency vibrational modes of dynamical matrix, we can then employ a\nconvolutional neural network, trained on the dynamical matrix instead of\ntraditional image recognition, to predict the Poisson's ratio of amorphous\nnetworks with a much higher efficiency. Through this example, we aim to\nshowcase the role that artificial intelligence can play in revealing\nfundamental physical mechanisms, which subsequently improves the machine\nlearning algorithms significantly.\n", "date": "2023-12-06", "categories": []}
{"id": "2407.04877", "title": "Leveraging Data Mining, Active Learning, and Domain Adaptation in a\n  Multi-Stage, Machine Learning-Driven Approach for the Efficient Discovery of\n  Advanced Acidic Oxygen Evolution Electrocatalysts", "abstract": "  Developing advanced catalysts for acidic oxygen evolution reaction (OER) is\ncrucial for sustainable hydrogen production. This study introduces a novel,\nmulti-stage machine learning (ML) approach to streamline the discovery and\noptimization of complex multi-metallic catalysts. Our method integrates data\nmining, active learning, and domain adaptation throughout the materials\ndiscovery process. Unlike traditional trial-and-error methods, this approach\nsystematically narrows the exploration space using domain knowledge with\nminimized reliance on subjective intuition. Then the active learning module\nefficiently refines element composition and synthesis conditions through\niterative experimental feedback. The process culminated in the discovery of a\npromising Ru-Mn-Ca-Pr oxide catalyst. Our workflow also enhances theoretical\nsimulations with domain adaptation strategy, providing deeper mechanistic\ninsights aligned with experimental findings. By leveraging diverse data sources\nand multiple ML strategies, we establish an efficient pathway for\nelectrocatalyst discovery and optimization. This comprehensive, data-driven\napproach represents a paradigm shift and potentially new benchmark in\nelectrocatalysts research.\n", "date": "2024-07-05", "categories": []}
{"id": "2407.02770", "title": "Large language models, physics-based modeling, experimental\n  measurements: the trinity of data-scarce learning of polymer properties", "abstract": "  Large language models (LLMs) bear promise as a fast and accurate material\nmodeling paradigm for evaluation, analysis, and design. Their vast number of\ntrainable parameters necessitates a wealth of data to achieve accuracy and\nmitigate overfitting. However, experimental measurements are often limited and\ncostly to obtain in sufficient quantities for finetuning. To this end, we\npresent a physics-based training pipeline that tackles the pathology of data\nscarcity. The core enabler is a physics-based modeling framework that generates\na multitude of synthetic data to align the LLM to a physically consistent\ninitial state before finetuning. Our framework features a two-phase training\nstrategy: (1) utilizing the large-in-amount while less accurate synthetic data\nfor supervised pretraining, and (2) finetuning the phase-1 model with limited\nexperimental data. We empirically demonstrate that supervised pretraining is\nvital to obtaining accurate finetuned LLMs, via the lens of learning polymer\nflammability metrics where cone calorimeter data is sparse.\n", "date": "2024-07-03", "categories": []}
{"id": "2407.00671", "title": "Establishing Deep InfoMax as an effective self-supervised learning\n  methodology in materials informatics", "abstract": "  The scarcity of property labels remains a key challenge in materials\ninformatics, whereas materials data without property labels are abundant in\ncomparison. By pretraining supervised property prediction models on\nself-supervised tasks that depend only on the \"intrinsic information\" available\nin any Crystallographic Information File (CIF), there is potential to leverage\nthe large amount of crystal data without property labels to improve property\nprediction results on small datasets. We apply Deep InfoMax as a\nself-supervised machine learning framework for materials informatics that\nexplicitly maximises the mutual information between a point set (or graph)\nrepresentation of a crystal and a vector representation suitable for downstream\nlearning. This allows the pretraining of supervised models on large materials\ndatasets without the need for property labels and without requiring the model\nto reconstruct the crystal from a representation vector. We investigate the\nbenefits of Deep InfoMax pretraining implemented on the Site-Net architecture\nto improve the performance of downstream property prediction models with small\namounts (<10^3) of data, a situation relevant to experimentally measured\nmaterials property databases. Using a property label masking methodology, where\nwe perform self-supervised learning on larger supervised datasets and then\ntrain supervised models on a small subset of the labels, we isolate Deep\nInfoMax pretraining from the effects of distributional shift. We demonstrate\nperformance improvements in the contexts of representation learning and\ntransfer learning on the tasks of band gap and formation energy prediction.\nHaving established the effectiveness of Deep InfoMax pretraining in a\ncontrolled environment, our findings provide a foundation for extending the\napproach to address practical challenges in materials informatics.\n", "date": "2024-06-30", "categories": []}
{"id": "2406.19792", "title": "Improving Performance Prediction of Electrolyte Formulations with\n  Transformer-based Molecular Representation Model", "abstract": "  Development of efficient and high-performing electrolytes is crucial for\nadvancing energy storage technologies, particularly in batteries. Predicting\nthe performance of battery electrolytes rely on complex interactions between\nthe individual constituents. Consequently, a strategy that adeptly captures\nthese relationships and forms a robust representation of the formulation is\nessential for integrating with machine learning models to predict properties\naccurately. In this paper, we introduce a novel approach leveraging a\ntransformer-based molecular representation model to effectively and efficiently\ncapture the representation of electrolyte formulations. The performance of the\nproposed approach is evaluated on two battery property prediction tasks and the\nresults show superior performance compared to the state-of-the-art methods.\n", "date": "2024-06-28", "categories": []}
{"id": "2406.15515", "title": "Machine Learning Models for Accurately Predicting Properties of CsPbCl3\n  Perovskite Quantum Dots", "abstract": "  Perovskite Quantum Dots (PQDs) have a promising future for several\napplications due to their unique properties. This study investigates the\neffectiveness of Machine Learning (ML) in predicting the size, absorbance (1S\nabs) and photoluminescence (PL) properties of $\\mathrm{CsPbCl}_3$ PQDs using\nsynthesizing features as the input dataset. the study employed ML models of\nSupport Vector Regression (SVR), Nearest Neighbour Distance (NND), Random\nForest (RF), Gradient Boosting Machine (GBM), Decision Tree (DT) and Deep\nLearning (DL). Although all models performed highly accurate results, SVR and\nNND demonstrated the best accurate property prediction by achieving excellent\nperformance on the test and training datasets, with high $\\mathrm{R}^2$ and low\nRoot Mean Squared Error (RMSE) and low Mean Absolute Error (MAE) metric values.\nGiven that ML is becoming more superior, its ability to understand the QDs\nfield could prove invaluable to shape the future of nanomaterials designing.\n", "date": "2024-06-20", "categories": []}
{"id": "2406.15873", "title": "NeuralSCF: Neural network self-consistent fields for density functional\n  theory", "abstract": "  Kohn-Sham density functional theory (KS-DFT) has found widespread application\nin accurate electronic structure calculations. However, it can be\ncomputationally demanding especially for large-scale simulations, motivating\nrecent efforts toward its machine-learning (ML) acceleration. We propose a\nneural network self-consistent fields (NeuralSCF) framework that establishes\nthe Kohn-Sham density map as a deep learning objective, which encodes the\nmechanics of the Kohn-Sham equations. Modeling this map with an\nSE(3)-equivariant graph transformer, NeuralSCF emulates the Kohn-Sham\nself-consistent iterations to obtain electron densities, from which other\nproperties can be derived. NeuralSCF achieves state-of-the-art accuracy in\nelectron density prediction and derived properties, featuring exceptional\nzero-shot generalization to a remarkable range of out-of-distribution systems.\nNeuralSCF reveals that learning from KS-DFT's intrinsic mechanics significantly\nenhances the model's accuracy and transferability, offering a promising\nstepping stone for accelerating electronic structure calculations through\nmechanics learning.\n", "date": "2024-06-22", "categories": []}
{"id": "2406.16166", "title": "Composite Material Design for Optimized Fracture Toughness Using Machine\n  Learning", "abstract": "  This paper investigates the optimization of 2D and 3D composite structures\nusing machine learning (ML) techniques, focusing on fracture toughness and\ncrack propagation in the Double Cantilever Beam (DCB) test. By exploring the\nintricate relationship between microstructural arrangements and macroscopic\nproperties of composites, the study demonstrates the potential of ML as a\npowerful tool to expedite the design optimization process, offering notable\nadvantages over traditional finite element analysis. The research encompasses\nfour distinct cases, examining crack propagation and fracture toughness in both\n2D and 3D composite models. Through the application of ML algorithms, the study\nshowcases the capability for rapid and accurate exploration of vast design\nspaces in composite materials. The findings highlight the efficiency of ML in\npredicting mechanical behaviors with limited training data, paving the way for\nbroader applications in composite design and optimization. This work\ncontributes to advancing the understanding of ML's role in enhancing the\nefficiency of composite material design processes.\n", "date": "2024-06-23", "categories": []}
{"id": "2406.14838", "title": "Bayesian neural networks for predicting uncertainty in full-field\n  material response", "abstract": "  Stress and material deformation field predictions are among the most\nimportant tasks in computational mechanics. These predictions are typically\nmade by solving the governing equations of continuum mechanics using finite\nelement analysis, which can become computationally prohibitive considering\ncomplex microstructures and material behaviors. Machine learning (ML) methods\noffer potentially cost effective surrogates for these applications. However,\nexisting ML surrogates are either limited to low-dimensional problems and/or do\nnot provide uncertainty estimates in the predictions. This work proposes an ML\nsurrogate framework for stress field prediction and uncertainty quantification\nfor diverse materials microstructures. A modified Bayesian U-net architecture\nis employed to provide a data-driven image-to-image mapping from initial\nmicrostructure to stress field with prediction (epistemic) uncertainty\nestimates. The Bayesian posterior distributions for the U-net parameters are\nestimated using three state-of-the-art inference algorithms: the posterior\nsampling-based Hamiltonian Monte Carlo method and two variational approaches,\nthe Monte-Carlo Dropout method and the Bayes by Backprop algorithm. A\nsystematic comparison of the predictive accuracy and uncertainty estimates for\nthese methods is performed for a fiber reinforced composite material and\npolycrystalline microstructure application. It is shown that the proposed\nmethods yield predictions of high accuracy compared to the FEA solution, while\nuncertainty estimates depend on the inference approach. Generally, the\nHamiltonian Monte Carlo and Bayes by Backprop methods provide consistent\nuncertainty estimates. Uncertainty estimates from Monte Carlo Dropout, on the\nother hand, are more difficult to interpret and depend strongly on the method's\ndesign.\n", "date": "2024-06-21", "categories": []}
{"id": "2406.15004", "title": "Dislocation cartography: Representations and unsupervised classification\n  of dislocation networks with unique fingerprints", "abstract": "  Detecting structure in data is the first step to arrive at meaningful\nrepresentations for systems. This is particularly challenging for dislocation\nnetworks evolving as a consequence of plastic deformation of crystalline\nsystems. Our study employs Isomap, a manifold learning technique, to unveil the\nintrinsic structure of high-dimensional density field data of dislocation\nstructures from different compression axis. The resulting maps provide a\nsystematic framework for quantitatively comparing dislocation structures,\noffering unique fingerprints based on density fields. Our novel, unbiased\napproach contributes to the quantitative classification of dislocation\nstructures which can be systematically extended.\n", "date": "2024-06-21", "categories": []}
{"id": "2406.12888", "title": "A Space Group Symmetry Informed Network for O(3) Equivariant Crystal\n  Tensor Prediction", "abstract": "  We consider the prediction of general tensor properties of crystalline\nmaterials, including dielectric, piezoelectric, and elastic tensors. A key\nchallenge here is how to make the predictions satisfy the unique tensor\nequivariance to O(3) group and invariance to crystal space groups. To this end,\nwe propose a General Materials Tensor Network (GMTNet), which is carefully\ndesigned to satisfy the required symmetries. To evaluate our method, we curate\na dataset and establish evaluation metrics that are tailored to the intricacies\nof crystal tensor predictions. Experimental results show that our GMTNet not\nonly achieves promising performance on crystal tensors of various orders but\nalso generates predictions fully consistent with the intrinsic crystal\nsymmetries. Our code is publicly available as part of the AIRS library\n(https://github.com/divelab/AIRS).\n", "date": "2024-06-03", "categories": []}
{"id": "2406.13163", "title": "LLMatDesign: Autonomous Materials Discovery with Large Language Models", "abstract": "  Discovering new materials can have significant scientific and technological\nimplications but remains a challenging problem today due to the enormity of the\nchemical space. Recent advances in machine learning have enabled data-driven\nmethods to rapidly screen or generate promising materials, but these methods\nstill depend heavily on very large quantities of training data and often lack\nthe flexibility and chemical understanding often desired in materials\ndiscovery. We introduce LLMatDesign, a novel language-based framework for\ninterpretable materials design powered by large language models (LLMs).\nLLMatDesign utilizes LLM agents to translate human instructions, apply\nmodifications to materials, and evaluate outcomes using provided tools. By\nincorporating self-reflection on its previous decisions, LLMatDesign adapts\nrapidly to new tasks and conditions in a zero-shot manner. A systematic\nevaluation of LLMatDesign on several materials design tasks, in silico,\nvalidates LLMatDesign's effectiveness in developing new materials with\nuser-defined target properties in the small data regime. Our framework\ndemonstrates the remarkable potential of autonomous LLM-guided materials\ndiscovery in the computational setting and towards self-driving laboratories in\nthe future.\n", "date": "2024-06-19", "categories": []}
{"id": "2402.09251", "title": "Universal Machine Learning Kohn-Sham Hamiltonian for Materials", "abstract": "  While density functional theory (DFT) serves as a prevalent computational\napproach in electronic structure calculations, its computational demands and\nscalability limitations persist. Recently, leveraging neural networks to\nparameterize the Kohn-Sham DFT Hamiltonian has emerged as a promising avenue\nfor accelerating electronic structure computations. Despite advancements,\nchallenges such as the necessity for computing extensive DFT training data to\nexplore each new system and the complexity of establishing accurate ML models\nfor multi-elemental materials still exist. Addressing these hurdles, this study\nintroduces a universal electronic Hamiltonian model trained on Hamiltonian\nmatrices obtained from first-principles DFT calculations of nearly all crystal\nstructures on the Materials Project. We demonstrate its generality in\npredicting electronic structures across the whole periodic table, including\ncomplex multi-elemental systems, solid-state electrolytes, Moir\\'e twisted\nbilayer heterostructure, and metal-organic frameworks (MOFs). Moreover, we\nutilize the universal model to conduct high-throughput calculations of\nelectronic structures for crystals in GeNOME datasets, identifying 3,940\ncrystals with direct band gaps and 5,109 crystals with flat bands. By offering\na reliable efficient framework for computing electronic properties, this\nuniversal Hamiltonian model lays the groundwork for advancements in diverse\nfields, such as easily providing a huge data set of electronic structures and\nalso making the materials design across the whole periodic table possible.\n", "date": "2024-02-14", "categories": []}
{"id": "2406.09263", "title": "Generative Inverse Design of Crystal Structures via Diffusion Models\n  with Transformers", "abstract": "  Recent advances in deep learning have enabled the generation of realistic\ndata by training generative models on large datasets of text, images, and\naudio. While these models have demonstrated exceptional performance in\ngenerating novel and plausible data, it remains an open question whether they\ncan effectively accelerate scientific discovery through the data generation and\ndrive significant advancements across various scientific fields. In particular,\nthe discovery of new inorganic materials with promising properties poses a\ncritical challenge, both scientifically and for industrial applications.\nHowever, unlike textual or image data, materials, or more specifically crystal\nstructures, consist of multiple types of variables - including lattice vectors,\natom positions, and atomic species. This complexity in data give rise to a\nvariety of approaches for representing and generating such data. Consequently,\nthe design choices of generative models for crystal structures remain an open\nquestion. In this study, we explore a new type of diffusion model for the\ngenerative inverse design of crystal structures, with a backbone based on a\nTransformer architecture. We demonstrate our models are superior to previous\nmethods in their versatility for generating crystal structures with desired\nproperties. Furthermore, our empirical results suggest that the optimal\nconditioning methods vary depending on the dataset.\n", "date": "2024-06-13", "categories": []}
{"id": "2312.16473", "title": "MolSets: Molecular Graph Deep Sets Learning for Mixture Property\n  Modeling", "abstract": "  Recent advances in machine learning (ML) have expedited materials discovery\nand design. One significant challenge faced in ML for materials is the\nexpansive combinatorial space of potential materials formed by diverse\nconstituents and their flexible configurations. This complexity is particularly\nevident in molecular mixtures, a frequently explored space for materials such\nas battery electrolytes. Owing to the complex structures of molecules and the\nsequence-independent nature of mixtures, conventional ML methods have\ndifficulties in modeling such systems. Here we present MolSets, a specialized\nML model for molecular mixtures. Representing individual molecules as graphs\nand their mixture as a set, MolSets leverages a graph neural network and the\ndeep sets architecture to extract information at the molecule level and\naggregate it at the mixture level, thus addressing local complexity while\nretaining global flexibility. We demonstrate the efficacy of MolSets in\npredicting the conductivity of lithium battery electrolytes and highlight its\nbenefits in virtual screening of the combinatorial chemical space.\n", "date": "2023-12-27", "categories": []}
{"id": "2401.13570", "title": "Guided Diffusion for Fast Inverse Design of Density-based Mechanical\n  Metamaterials", "abstract": "  Mechanical metamaterial is a synthetic material that can possess\nextraordinary physical characteristics, such as abnormal elasticity, stiffness,\nand stability, by carefully designing its internal structure. To make\nmetamaterials contain delicate local structures with unique mechanical\nproperties, it is a potential method to represent them through high-resolution\nvoxels. However, it brings a substantial computational burden. To this end,\nthis paper proposes a fast inverse design method, whose core is an advanced\ndeep generative AI algorithm, to generate voxel-based mechanical metamaterials.\nSpecifically, we use the self-conditioned diffusion model, capable of\ngenerating a microstructure with a resolution of $128^3$ to approach the\nspecified homogenized tensor matrix in just 3 seconds. Accordingly, this rapid\nreverse design tool facilitates the exploration of extreme metamaterials, the\nsequence interpolation in metamaterials, and the generation of diverse\nmicrostructures for multi-scale design. This flexible and adaptive generative\ntool is of great value in structural engineering or other mechanical systems\nand can stimulate more subsequent research.\n", "date": "2024-01-24", "categories": []}
{"id": "2406.05142", "title": "Machine Learning-Driven Optimization of TPMS Architected Materials Using\n  Simulated Annealing", "abstract": "  The research paper presents a novel approach to optimizing the tensile stress\nof Triply Periodic Minimal Surface (TPMS) structures through machine learning\nand Simulated Annealing (SA). The study evaluates the performance of Random\nForest, Decision Tree, and XGBoost models in predicting tensile stress, using a\ndataset generated from finite element analysis of TPMS models. The objective\nfunction minimized the negative R-squared value on the validation set to\nenhance model accuracy. The SA-XGBoost model outperformed the others, achieving\nan R-squared value of 0.96. In contrast, the SA-Random Forest model achieved an\nR squared value of 0.89 while the SA-Decision Tree model exhibited greater\nfluctuations in validation scores. This demonstrates that the SA-XGBoost model\nis most effective in capturing the complex relationships within the data. The\nintegration of SA helps in optimizing the hyperparameters of these machine\nlearning models, thereby enhancing their predictive capabilities.\n", "date": "2024-05-28", "categories": []}
{"id": "2402.16119", "title": "DeepForge: Leveraging AI for Microstructural Control in Metal Forming\n  via Model Predictive Control", "abstract": "  This study presents a novel method for microstructure control in closed die\nhot forging that combines Model Predictive Control (MPC) with a developed\nmachine learning model called DeepForge. DeepForge uses an architecture that\ncombines 1D convolutional neural networks and gated recurrent units. It uses\nsurface temperature measurements of a workpiece as input to predict\nmicrostructure changes during forging. The paper also details DeepForge's\narchitecture and the finite element simulation model used to generate the data\nset, using a three-stroke forging process. The results demonstrate DeepForge's\nability to predict microstructure with a mean absolute error of 0.4$\\pm$0.3%.\nIn addition, the study explores the use of MPC to adjust inter-stroke wait\ntimes, effectively counteracting temperature disturbances to achieve a target\ngrain size of less than 35 microns within a specific 2D region of the\nworkpiece. These results are then verified experimentally, demonstrating a\nsignificant step towards improved control and quality in forging processes\nwhere temperature can be used as an additional degree of freedom in the\nprocess.\n", "date": "2024-02-25", "categories": []}
{"id": "2406.04713", "title": "FlowMM: Generating Materials with Riemannian Flow Matching", "abstract": "  Crystalline materials are a fundamental component in next-generation\ntechnologies, yet modeling their distribution presents unique computational\nchallenges. Of the plausible arrangements of atoms in a periodic lattice only a\nvanishingly small percentage are thermodynamically stable, which is a key\nindicator of the materials that can be experimentally realized. Two fundamental\ntasks in this area are to (a) predict the stable crystal structure of a known\ncomposition of elements and (b) propose novel compositions along with their\nstable structures. We present FlowMM, a pair of generative models that achieve\nstate-of-the-art performance on both tasks while being more efficient and more\nflexible than competing methods. We generalize Riemannian Flow Matching to suit\nthe symmetries inherent to crystals: translation, rotation, permutation, and\nperiodic boundary conditions. Our framework enables the freedom to choose the\nflow base distributions, drastically simplifying the problem of learning\ncrystal structures compared with diffusion models. In addition to standard\nbenchmarks, we validate FlowMM's generated structures with quantum chemistry\ncalculations, demonstrating that it is about 3x more efficient, in terms of\nintegration steps, at finding stable materials compared to previous open\nmethods.\n", "date": "2024-06-07", "categories": []}
{"id": "2311.06380", "title": "Theory and implementation of inelastic Constitutive Artificial Neural\n  Networks", "abstract": "  Nature has always been our inspiration in the research, design and\ndevelopment of materials and has driven us to gain a deep understanding of the\nmechanisms that characterize anisotropy and inelastic behavior. All this\nknowledge has been accumulated in the principles of thermodynamics. Deduced\nfrom these principles, the multiplicative decomposition combined with pseudo\npotentials are powerful and universal concepts. Simultaneously, the tremendous\nincrease in computational performance enabled us to investigate and rethink our\nhistory-dependent material models to make the most of our predictions. Today,\nwe have reached a point where materials and their models are becoming\nincreasingly sophisticated. This raises the question: How do we find the best\nmodel that includes all inelastic effects to explain our complex data?\nConstitutive Artificial Neural Networks (CANN) may answer this question. Here,\nwe extend the CANNs to inelastic materials (iCANN). Rigorous considerations of\nobjectivity, rigid motion of the reference configuration, multiplicative\ndecomposition and its inherent non-uniqueness, restrictions of energy and\npseudo potential, and consistent evolution guide us towards the architecture of\nthe iCANN satisfying thermodynamics per design. We combine feed-forward\nnetworks of the free energy and pseudo potential with a recurrent neural\nnetwork approach to take time dependencies into account. We demonstrate that\nthe iCANN is capable of autonomously discovering models for artificially\ngenerated data, the response of polymers for cyclic loading and the relaxation\nbehavior of muscle data. As the design of the network is not limited to\nvisco-elasticity, our vision is that the iCANN will reveal to us new ways to\nfind the various inelastic phenomena hidden in the data and to understand their\ninteraction. Our source code, data, and examples are available at\ndoi.org/10.5281/zenodo.10066805\n", "date": "2023-11-10", "categories": []}
{"id": "2311.09235", "title": "Scalable Diffusion for Materials Generation", "abstract": "  Generative models trained on internet-scale data are capable of generating\nnovel and realistic texts, images, and videos. A natural next question is\nwhether these models can advance science, for example by generating novel\nstable materials. Traditionally, models with explicit structures (e.g., graphs)\nhave been used in modeling structural relationships in scientific data (e.g.,\natoms and bonds in crystals), but generating structures can be difficult to\nscale to large and complex systems. Another challenge in generating materials\nis the mismatch between standard generative modeling metrics and downstream\napplications. For instance, common metrics such as the reconstruction error do\nnot correlate well with the downstream goal of discovering stable materials. In\nthis work, we tackle the scalability challenge by developing a unified crystal\nrepresentation that can represent any crystal structure (UniMat), followed by\ntraining a diffusion probabilistic model on these UniMat representations. Our\nempirical results suggest that despite the lack of explicit structure modeling,\nUniMat can generate high fidelity crystal structures from larger and more\ncomplex chemical systems, outperforming previous graph-based approaches under\nvarious generative modeling metrics. To better connect the generation quality\nof materials to downstream applications, such as discovering novel stable\nmaterials, we propose additional metrics for evaluating generative models of\nmaterials, including per-composition formation energy and stability with\nrespect to convex hulls through decomposition energy from Density Function\nTheory (DFT). Lastly, we show that conditional generation with UniMat can scale\nto previously established crystal datasets with up to millions of crystals\nstructures, outperforming random structure search (the current leading method\nfor structure discovery) in discovering new stable materials.\n", "date": "2023-10-18", "categories": []}
{"id": "2406.01471", "title": "Inverse design of photonic surfaces on Inconel via multi-fidelity\n  machine learning ensemble framework and high throughput femtosecond laser\n  processing", "abstract": "  We demonstrate a multi-fidelity (MF) machine learning ensemble framework for\nthe inverse design of photonic surfaces, trained on a dataset of 11,759 samples\nthat we fabricate using high throughput femtosecond laser processing. The MF\nensemble combines an initial low fidelity model for generating design\nsolutions, with a high fidelity model that refines these solutions through\nlocal optimization. The combined MF ensemble can generate multiple disparate\nsets of laser-processing parameters that can each produce the same target input\nspectral emissivity with high accuracy (root mean squared errors < 2%). SHapley\nAdditive exPlanations analysis shows transparent model interpretability of the\ncomplex relationship between laser parameters and spectral emissivity. Finally,\nthe MF ensemble is experimentally validated by fabricating and evaluating\nphotonic surface designs that it generates for improved efficiency energy\nharvesting devices. Our approach provides a powerful tool for advancing the\ninverse design of photonic surfaces in energy harvesting applications.\n", "date": "2024-06-03", "categories": []}
{"id": "2304.07877", "title": "A Neural Network Transformer Model for Composite Microstructure\n  Homogenization", "abstract": "  Heterogeneity and uncertainty in a composite microstructure lead to either\ncomputational bottlenecks if modeled rigorously or to solution inaccuracies in\nthe stress field and failure predictions if approximated. Although methods\nsuitable for analyzing arbitrary and non-linear microstructures exist, their\ncomputational cost makes them impractical to use in large-scale structural\nanalysis. Surrogate models or Reduced Order Models (ROMs) commonly enhance\nefficiencies but are typically calibrated with a single microstructure.\nHomogenization methods, such as the Mori-Tanaka method, offer rapid\nhomogenization for a wide range of constituent properties. However, simplifying\nassumptions, like stress and strain averaging in phases, render the\nconsideration of both deterministic and stochastic variations in microstructure\ninfeasible. This paper illustrates a transformer neural network architecture\nthat captures the knowledge of various microstructures and constituents,\nenabling it to function as a computationally efficient homogenization surrogate\nmodel. Given an image or an abstraction of an arbitrary composite\nmicrostructure of linearly elastic fibers in an elastoplastic matrix, the\ntransformer network predicts the history-dependent, non-linear, and homogenized\nstress-strain response. Two methods for encoding microstructure features were\ntested: calculating two-point statistics using Principal Component Analysis\n(PCA) for dimensionality reduction and employing an autoencoder with a\nConvolutional Neural Network (CNN). Both methods accurately predict the\nhomogenized material response. The developed transformer neural network offers\nan efficient means for microstructure-to-property translation, generalizable\nand extendable to a variety of microstructures. The paper describes the network\narchitecture, training and testing data generation, and performance under\ncycling and random loadings.\n", "date": "2023-04-16", "categories": []}
{"id": "2402.11101", "title": "Physics-based material parameters extraction from perovskite experiments\n  via Bayesian optimization", "abstract": "  The ability to extract material parameters of perovskite from quantitative\nexperimental analysis is essential for rational design of photovoltaic and\noptoelectronic applications. However, the difficulty of this analysis increases\nsignificantly with the complexity of the theoretical model and the number of\nmaterial parameters for perovskite. Here we use Bayesian optimization to\ndevelop an analysis platform that can extract up to 8 fundamental material\nparameters of an organometallic perovskite semiconductor from a transient\nphotoluminescence experiment, based on a complex full physics model that\nincludes drift-diffusion of carriers and dynamic defect occupation. An example\nstudy of thermal degradation reveals that the carrier mobility and\ntrap-assisted recombination coefficient are reduced noticeably, while the\ndefect energy level remains nearly unchanged. The reduced carrier mobility can\ndominate the overall effect on thermal degradation of perovskite solar cells by\nreducing the fill factor, despite the opposite effect of the reduced\ntrap-assisted recombination coefficient on increasing the fill factor. In\nfuture, this platform can be conveniently applied to other experiments or to\ncombinations of experiments, accelerating materials discovery and optimization\nof semiconductor materials for photovoltaics and other applications.\n", "date": "2024-02-16", "categories": []}
{"id": "2405.19276", "title": "A Recipe for Charge Density Prediction", "abstract": "  In density functional theory, charge density is the core attribute of atomic\nsystems from which all chemical properties can be derived. Machine learning\nmethods are promising in significantly accelerating charge density prediction,\nyet existing approaches either lack accuracy or scalability. We propose a\nrecipe that can achieve both. In particular, we identify three key ingredients:\n(1) representing the charge density with atomic and virtual orbitals (spherical\nfields centered at atom/virtual coordinates); (2) using expressive and\nlearnable orbital basis sets (basis function for the spherical fields); and (3)\nusing high-capacity equivariant neural network architecture. Our method\nachieves state-of-the-art accuracy while being more than an order of magnitude\nfaster than existing methods. Furthermore, our method enables flexible\nefficiency-accuracy trade-offs by adjusting the model/basis sizes.\n", "date": "2024-05-29", "categories": []}
{"id": "2402.00853", "title": "LTAU-FF: Loss Trajectory Analysis for Uncertainty in Atomistic Force\n  Fields", "abstract": "  Model ensembles are effective tools for estimating prediction uncertainty in\ndeep learning atomistic force fields. However, their widespread adoption is\nhindered by high computational costs and overconfident error estimates. In this\nwork, we address these challenges by leveraging distributions of per-sample\nerrors obtained during training and employing a distance-based similarity\nsearch in the model latent space. Our method, which we call LTAU, efficiently\nestimates the full probability distribution function (PDF) of errors for any\ntest point using the logged training errors, achieving speeds that are 2--3\norders of magnitudes faster than typical ensemble methods and allowing it to be\nused for tasks where training or evaluating multiple models would be\ninfeasible. We apply LTAU towards estimating parametric uncertainty in\natomistic force fields (LTAU-FF), demonstrating that its improved ensemble\ndiversity produces well-calibrated confidence intervals and predicts errors\nthat correlate strongly with the true errors for data near the training domain.\nFurthermore, we show that the errors predicted by LTAU-FF can be used in\npractical applications for detecting out-of-domain data, tuning model\nperformance, and predicting failure during simulations. We believe that LTAU\nwill be a valuable tool for uncertainty quantification (UQ) in atomistic force\nfields and is a promising method that should be further explored in other\ndomains of machine learning.\n", "date": "2024-02-01", "categories": []}
{"id": "2402.17660", "title": "TorchMD-Net 2.0: Fast Neural Network Potentials for Molecular\n  Simulations", "abstract": "  Achieving a balance between computational speed, prediction accuracy, and\nuniversal applicability in molecular simulations has been a persistent\nchallenge. This paper presents substantial advancements in the TorchMD-Net\nsoftware, a pivotal step forward in the shift from conventional force fields to\nneural network-based potentials. The evolution of TorchMD-Net into a more\ncomprehensive and versatile framework is highlighted, incorporating\ncutting-edge architectures such as TensorNet. This transformation is achieved\nthrough a modular design approach, encouraging customized applications within\nthe scientific community. The most notable enhancement is a significant\nimprovement in computational efficiency, achieving a very remarkable\nacceleration in the computation of energy and forces for TensorNet models, with\nperformance gains ranging from 2-fold to 10-fold over previous iterations.\nOther enhancements include highly optimized neighbor search algorithms that\nsupport periodic boundary conditions and the smooth integration with existing\nmolecular dynamics frameworks. Additionally, the updated version introduces the\ncapability to integrate physical priors, further enriching its application\nspectrum and utility in research. The software is available at\nhttps://github.com/torchmd/torchmd-net.\n", "date": "2024-02-27", "categories": []}
{"id": "2405.13759", "title": "Enhancing Multiscale Simulations with Constitutive Relations-Aware Deep\n  Operator Networks", "abstract": "  Multiscale problems are widely observed across diverse domains in physics and\nengineering. Translating these problems into numerical simulations and solving\nthem using numerical schemes, e.g. the finite element method, is costly due to\nthe demand of solving initial boundary-value problems at multiple scales. On\nthe other hand, multiscale finite element computations are commended for their\nability to integrate micro-structural properties into macroscopic computational\nanalyses using homogenization techniques. Recently, neural operator-based\nsurrogate models have shown trustworthy performance for solving a wide range of\npartial differential equations. In this work, we propose a hybrid method in\nwhich we utilize deep operator networks for surrogate modeling of the\nmicroscale physics. This allows us to embed the constitutive relations of the\nmicroscale into the model architecture and to predict microscale strains and\nstresses based on the prescribed macroscale strain inputs. Furthermore,\nnumerical homogenization is carried out to obtain the macroscale quantities of\ninterest. We apply the proposed approach to quasi-static problems of solid\nmechanics. The results demonstrate that our constitutive relations-aware\nDeepONet can yield accurate solutions even when being confronted with a\nrestricted dataset during model development.\n", "date": "2024-05-22", "categories": []}
{"id": "2311.11343", "title": "A Generative Model for Accelerated Inverse Modelling Using a Novel\n  Embedding for Continuous Variables", "abstract": "  In materials science, the challenge of rapid prototyping materials with\ndesired properties often involves extensive experimentation to find suitable\nmicrostructures. Additionally, finding microstructures for given properties is\ntypically an ill-posed problem where multiple solutions may exist. Using\ngenerative machine learning models can be a viable solution which also reduces\nthe computational cost. This comes with new challenges because, e.g., a\ncontinuous property variable as conditioning input to the model is required. We\ninvestigate the shortcomings of an existing method and compare this to a novel\nembedding strategy for generative models that is based on the binary\nrepresentation of floating point numbers. This eliminates the need for\nnormalization, preserves information, and creates a versatile embedding space\nfor conditioning the generative model. This technique can be applied to\ncondition a network on any number, to provide fine control over generated\nmicrostructure images, thereby contributing to accelerated materials design.\n", "date": "2023-11-19", "categories": []}
{"id": "2405.11404", "title": "How big is Big Data?", "abstract": "  Big data has ushered in a new wave of predictive power using machine learning\nmodels. In this work, we assess what {\\it big} means in the context of typical\nmaterials-science machine-learning problems. This concerns not only data\nvolume, but also data quality and veracity as much as infrastructure issues.\nWith selected examples, we ask (i) how models generalize to similar datasets,\n(ii) how high-quality datasets can be gathered from heterogenous sources, (iii)\nhow the feature set and complexity of a model can affect expressivity, and (iv)\nwhat infrastructure requirements are needed to create larger datasets and train\nmodels on them. In sum, we find that big data present unique challenges along\nvery different aspects that should serve to motivate further work.\n", "date": "2024-05-18", "categories": []}
{"id": "2405.09057", "title": "Response Matching for generating materials and molecules", "abstract": "  Machine learning has recently emerged as a powerful tool for generating new\nmolecular and material structures. The success of state-of-the-art models stems\nfrom their ability to incorporate physical symmetries, such as translation,\nrotation, and periodicity. Here, we present a novel generative method called\nResponse Matching (RM), which leverages the fact that each stable material or\nmolecule exists at the minimum of its potential energy surface. Consequently,\nany perturbation induces a response in energy and stress, driving the structure\nback to equilibrium. Matching to such response is closely related to score\nmatching in diffusion models. By employing the combination of a machine\nlearning interatomic potential and random structure search as the denoising\nmodel, RM exploits the locality of atomic interactions, and inherently respects\npermutation, translation, rotation, and periodic invariances. RM is the first\nmodel to handle both molecules and bulk materials under the same framework. We\ndemonstrate the efficiency and generalization of RM across three systems: a\nsmall organic molecular dataset, stable crystals from the Materials Project,\nand one-shot learning on a single diamond configuration.\n", "date": "2024-05-15", "categories": []}
{"id": "2212.14813", "title": "Unsupervised learning for structure detection in plastically deformed\n  crystals", "abstract": "  Detecting structures at the particle scale within plastically deformed\ncrystalline materials allows a better understanding of the occurring phenomena.\nWhile previous approaches mostly relied on applying hand-chosen criteria on\ndifferent local parameters, these approaches could only detect already known\nstructures.We introduce an unsupervised learning algorithm to automatically\ndetect structures within a crystal under plastic deformation. This approach is\nbased on a study developed for structural detection on colloidal materials.\nThis algorithm has the advantage of being computationally fast and easy to\nimplement. We show that by using local parameters based on bond-angle\ndistributions, we are able to detect more structures and with a higher degree\nof precision than traditional hand-made criteria.\n", "date": "2022-12-22", "categories": []}
{"id": "2311.02495", "title": "Uncertainty Quantification in Multivariable Regression for Material\n  Property Prediction with Bayesian Neural Networks", "abstract": "  With the increased use of data-driven approaches and machine learning-based\nmethods in material science, the importance of reliable uncertainty\nquantification (UQ) of the predicted variables for informed decision-making\ncannot be overstated. UQ in material property prediction poses unique\nchallenges, including the multi-scale and multi-physics nature of advanced\nmaterials, intricate interactions between numerous factors, limited\navailability of large curated datasets for model training, etc. Recently,\nBayesian Neural Networks (BNNs) have emerged as a promising approach for UQ,\noffering a probabilistic framework for capturing uncertainties within neural\nnetworks. In this work, we introduce an approach for UQ within physics-informed\nBNNs, which integrates knowledge from governing laws in material modeling to\nguide the models toward physically consistent predictions. To evaluate the\neffectiveness of this approach, we present case studies for predicting the\ncreep rupture life of steel alloys. Experimental validation with three datasets\nof collected measurements from creep tests demonstrates the ability of BNNs to\nproduce accurate point and uncertainty estimates that are competitive or exceed\nthe performance of the conventional method of Gaussian Process Regression.\nSimilarly, we evaluated the suitability of BNNs for UQ in an active learning\napplication and reported competitive performance. The most promising framework\nfor creep life prediction is BNNs based on Markov Chain Monte Carlo\napproximation of the posterior distribution of network parameters, as it\nprovided more reliable results in comparison to BNNs based on variational\ninference approximation or related NNs with probabilistic outputs. The codes\nare available at:\nhttps://github.com/avakanski/Creep-uncertainty-quantification.\n", "date": "2023-11-04", "categories": []}
{"id": "2405.08137", "title": "LATTE: an atomic environment descriptor based on Cartesian tensor\n  contractions", "abstract": "  We propose a new descriptor for local atomic environments, to be used in\ncombination with machine learning models for the construction of interatomic\npotentials. The Local Atomic Tensors Trainable Expansion (LATTE) allows for the\nefficient construction of a variable number of many-body terms with learnable\nparameters, resulting in a descriptor that is efficient, expressive, and can be\nscaled to suit different accuracy and computational cost requirements. We\ncompare this new descriptor to existing ones on several systems, showing it to\nbe competitive with very fast potentials at one end of the spectrum, and\nextensible to an accuracy close to the state of the art.\n", "date": "2024-05-13", "categories": []}
{"id": "2405.07105", "title": "Overcoming systematic softening in universal machine learning\n  interatomic potentials by fine-tuning", "abstract": "  Machine learning interatomic potentials (MLIPs) have introduced a new\nparadigm for atomic simulations. Recent advancements have seen the emergence of\nuniversal MLIPs (uMLIPs) that are pre-trained on diverse materials datasets,\nproviding opportunities for both ready-to-use universal force fields and robust\nfoundations for downstream machine learning refinements. However, their\nperformance in extrapolating to out-of-distribution complex atomic environments\nremains unclear. In this study, we highlight a consistent potential energy\nsurface (PES) softening effect in three uMLIPs: M3GNet, CHGNet, and MACE-MP-0,\nwhich is characterized by energy and force under-prediction in a series of\natomic-modeling benchmarks including surfaces, defects, solid-solution\nenergetics, phonon vibration modes, ion migration barriers, and general\nhigh-energy states.\n  We find that the PES softening behavior originates from a systematic\nunderprediction error of the PES curvature, which derives from the biased\nsampling of near-equilibrium atomic arrangements in uMLIP pre-training\ndatasets. We demonstrate that the PES softening issue can be effectively\nrectified by fine-tuning with a single additional data point. Our findings\nsuggest that a considerable fraction of uMLIP errors are highly systematic, and\ncan therefore be efficiently corrected. This result rationalizes the\ndata-efficient fine-tuning performance boost commonly observed with\nfoundational MLIPs. We argue for the importance of a comprehensive materials\ndataset with improved PES sampling for next-generation foundational MLIPs.\n", "date": "2024-05-11", "categories": []}
{"id": "2405.07599", "title": "Transferable Neural Wavefunctions for Solids", "abstract": "  Deep-Learning-based Variational Monte Carlo (DL-VMC) has recently emerged as\na highly accurate approach for finding approximate solutions to the\nmany-electron Schr\\\"odinger equation. Despite its favorable scaling with the\nnumber of electrons, $\\mathcal{O}(n_\\text{el}^{4})$, the practical value of\nDL-VMC is limited by the high cost of optimizing the neural network weights for\nevery system studied. To mitigate this problem, recent research has proposed\noptimizing a single neural network across multiple systems, reducing the cost\nper system. Here we extend this approach to solids, where similar but distinct\ncalculations using different geometries, boundary conditions, and supercell\nsizes are often required. We show how to optimize a single ansatz across all of\nthese variations, reducing the required number of optimization steps by an\norder of magnitude. Furthermore, we exploit the transfer capabilities of a\npre-trained network. We successfully transfer a network, pre-trained on 2x2x2\nsupercells of LiH, to 3x3x3 supercells. This reduces the number of optimization\nsteps required to simulate the large system by a factor of 50 compared to\nprevious work.\n", "date": "2024-05-13", "categories": []}
{"id": "2212.11246", "title": "Material Property Prediction using Graphs based on Generically Complete\n  Isometry Invariants", "abstract": "  The structure-property hypothesis says that the properties of all materials\nare determined by an underlying crystal structure. The main obstacle was the\nambiguity of conventional crystal representations based on incomplete or\ndiscontinuous descriptors that allow false negatives or false positives. This\nambiguity was resolved by the ultra-fast Pointwise Distance Distribution (PDD),\nwhich distinguished all periodic structures in the world's largest collection\nof real materials (Cambridge Structural Database). The state-of-the-art results\nin property predictions were previously achieved by graph neural networks based\non various graph representations of periodic crystals, including the Crystal\nGraph with vertices at all atoms in a crystal unit cell. This work adapts the\nPointwise Distance Distribution for a simpler graph whose vertex set is not\nlarger than the asymmetric unit of a crystal structure. The new Distribution\nGraph reduces mean-absolute-error by 0.6\\%-12\\% while having 44\\%-88\\% of the\nnumber of vertices when compared to the crystal graph when applied on the\nMaterials Project and Jarvis-DFT datasets using CGCNN and ALIGNN. Methods for\nhyper-parameters selection for the graph are backed by the theoretical results\nof the Pointwise Distance Distribution and are then experimentally justified.\n", "date": "2022-12-19", "categories": []}
{"id": "2303.11833", "title": "Materials Discovery with Extreme Properties via Reinforcement\n  Learning-Guided Combinatorial Chemistry", "abstract": "  The goal of most materials discovery is to discover materials that are\nsuperior to those currently known. Fundamentally, this is close to\nextrapolation, which is a weak point for most machine learning models that\nlearn the probability distribution of data. Herein, we develop reinforcement\nlearning-guided combinatorial chemistry, which is a rule-based molecular\ndesigner driven by trained policy for selecting subsequent molecular fragments\nto get a target molecule. Since our model has the potential to generate all\npossible molecular structures that can be obtained from combinations of\nmolecular fragments, unknown molecules with superior properties can be\ndiscovered. We theoretically and empirically demonstrate that our model is more\nsuitable for discovering better compounds than probability\ndistribution-learning models. In an experiment aimed at discovering molecules\nthat hit seven extreme target properties, our model discovered 1,315 of all\ntarget-hitting molecules and 7,629 of five target-hitting molecules out of\n100,000 trials, whereas the probability distribution-learning models failed.\nMoreover, it has been confirmed that every molecule generated under the binding\nrules of molecular fragments is 100% chemically valid. To illustrate the\nperformance in actual problems, we also demonstrate that our models work well\non two practical applications: discovering protein docking molecules and HIV\ninhibitors.\n", "date": "2023-03-21", "categories": []}
{"id": "2405.03962", "title": "AdsorbDiff: Adsorbate Placement via Conditional Denoising Diffusion", "abstract": "  Determining the optimal configuration of adsorbates on a slab (adslab) is\npivotal in the exploration of novel catalysts across diverse applications.\nTraditionally, the quest for the lowest energy adslab configuration involves\nplacing the adsorbate onto the slab followed by an optimization process. Prior\nmethodologies have relied on heuristics, problem-specific intuitions, or\nbrute-force approaches to guide adsorbate placement. In this work, we propose a\nnovel framework for adsorbate placement using denoising diffusion. The model is\ndesigned to predict the optimal adsorbate site and orientation corresponding to\nthe lowest energy configuration. Further, we have an end-to-end evaluation\nframework where diffusion-predicted adslab configuration is optimized with a\npretrained machine learning force field and finally evaluated with Density\nFunctional Theory (DFT). Our findings demonstrate an acceleration of up to 5x\nor 3.5x improvement in accuracy compared to the previous best approach. Given\nthe novelty of this framework and application, we provide insights into the\nimpact of pre-training, model architectures, and conduct extensive experiments\nto underscore the significance of this approach.\n", "date": "2024-05-07", "categories": []}
{"id": "2310.16802", "title": "From Molecules to Materials: Pre-training Large Generalizable Models for\n  Atomic Property Prediction", "abstract": "  Foundation models have been transformational in machine learning fields such\nas natural language processing and computer vision. Similar success in atomic\nproperty prediction has been limited due to the challenges of training\neffective models across multiple chemical domains. To address this, we\nintroduce Joint Multi-domain Pre-training (JMP), a supervised pre-training\nstrategy that simultaneously trains on multiple datasets from different\nchemical domains, treating each dataset as a unique pre-training task within a\nmulti-task framework. Our combined training dataset consists of $\\sim$120M\nsystems from OC20, OC22, ANI-1x, and Transition-1x. We evaluate performance and\ngeneralization by fine-tuning over a diverse set of downstream tasks and\ndatasets including: QM9, rMD17, MatBench, QMOF, SPICE, and MD22. JMP\ndemonstrates an average improvement of 59% over training from scratch, and\nmatches or sets state-of-the-art on 34 out of 40 tasks. Our work highlights the\npotential of pre-training strategies that utilize diverse data to advance\nproperty prediction across chemical domains, especially for low-data tasks.\nPlease visit https://nima.sh/jmp for further information.\n", "date": "2023-10-25", "categories": []}
{"id": "2308.04103", "title": "Explainable machine learning to enable high-throughput electrical\n  conductivity optimization and discovery of doped conjugated polymers", "abstract": "  The combination of high-throughput experimentation techniques and machine\nlearning (ML) has recently ushered in a new era of accelerated material\ndiscovery, enabling the identification of materials with cutting-edge\nproperties. However, the measurement of certain physical quantities remains\nchallenging to automate. Specifically, meticulous process control,\nexperimentation and laborious measurements are required to achieve optimal\nelectrical conductivity in doped polymer materials. We propose a ML approach,\nwhich relies on readily measured absorbance spectra, to accelerate the workflow\nassociated with measuring electrical conductivity. The classification model\naccurately classifies samples with a conductivity > 25 to 100 S/cm, achieving a\nmaximum of 100 % accuracy rate. For the subset of highly conductive samples, we\nemployed a regression model to predict their conductivities, yielding an\nimpressive test R2 value of 0.984. We tested the models with samples of the two\nhighest conductivities (498 and 506 S/cm) and showed that they were able to\ncorrectly classify and predict the two extrapolative conductivities at\nsatisfactory levels of errors. The proposed ML-assisted workflow results in an\nimprovement in the efficiency of the conductivity measurements by 89 % of the\nmaximum achievable using our experimental techniques. Furthermore, our approach\naddressed the common challenge of the lack of explainability in ML models by\nexploiting bespoke mathematical properties of the descriptors and ML model,\nallowing us to gain corroborated insights into the spectral influences on\nconductivity. Through this study, we offer an accelerated pathway for\noptimizing the properties of doped polymer materials while showcasing the\nvaluable insights that can be derived from purposeful utilization of ML in\nexperimental science.\n", "date": "2023-08-08", "categories": []}
{"id": "2404.17583", "title": "Physically recurrent neural network for rate and path-dependent\n  heterogeneous materials in a finite strain framework", "abstract": "  In this work, a hybrid physics-based data-driven surrogate model for the\nmicroscale analysis of heterogeneous material is investigated. The proposed\nmodel benefits from the physics-based knowledge contained in the constitutive\nmodels used in the full-order micromodel by embedding them in a neural network.\nFollowing previous developments, this paper extends the applicability of the\nphysically recurrent neural network (PRNN) by introducing an architecture\nsuitable for rate-dependent materials in a finite strain framework. In this\nmodel, the homogenized deformation gradient of the micromodel is encoded into a\nset of deformation gradients serving as input to the embedded constitutive\nmodels. These constitutive models compute stresses, which are combined in a\ndecoder to predict the homogenized stress, such that the internal variables of\nthe history-dependent constitutive models naturally provide physics-based\nmemory for the network. To demonstrate the capabilities of the surrogate model,\nwe consider a unidirectional composite micromodel with transversely isotropic\nelastic fibers and elasto-viscoplastic matrix material. The extrapolation\nproperties of the surrogate model trained to replace such micromodel are tested\non loading scenarios unseen during training, ranging from different\nstrain-rates to cyclic loading and relaxation. Speed-ups of three orders of\nmagnitude with respect to the runtime of the original micromodel are obtained.\n", "date": "2024-04-05", "categories": []}
{"id": "2404.15207", "title": "Simulation-Free Determination of Microstructure Representative Volume\n  Element Size via Fisher Scores", "abstract": "  A representative volume element (RVE) is a reasonably small unit of\nmicrostructure that can be simulated to obtain the same effective properties as\nthe entire microstructure sample. Finite element (FE) simulation of RVEs, as\nopposed to much larger samples, saves computational expense, especially in\nmultiscale modeling. Therefore, it is desirable to have a framework that\ndetermines RVE size prior to FE simulations. Existing methods select the RVE\nsize based on when the FE-simulated properties of samples of increasing size\nconverge with insignificant statistical variations, with the drawback that many\nsamples must be simulated. We propose a simulation-free alternative that\ndetermines RVE size based only on a micrograph. The approach utilizes a machine\nlearning model trained to implicitly characterize the stochastic nature of the\ninput micrograph. The underlying rationale is to view RVE size as the smallest\nmoving window size for which the stochastic nature of the microstructure within\nthe window is stationary as the window moves across a large micrograph. For\nthis purpose, we adapt a recently developed Fisher score-based framework for\nmicrostructure nonstationarity monitoring. Because the resulting RVE size is\nbased solely on the micrograph and does not involve any FE simulation of\nspecific properties, it constitutes an RVE for any property of interest that\nsolely depends on the microstructure characteristics. Through numerical\nexperiments of simple and complex microstructures, we validate our approach and\nshow that our selected RVE sizes are consistent with when the chosen\nFE-simulated properties converge.\n", "date": "2024-04-07", "categories": []}
{"id": "2210.07930", "title": "Machine learning frontier orbital energies of nanodiamonds", "abstract": "  Nanodiamonds have a wide range of applications including catalysis, sensing,\ntribology and biomedicine. To leverage nanodiamond design via machine learning,\nwe introduce the new dataset ND5k, consisting of 5,089 diamondoid and\nnanodiamond structures and their frontier orbital energies. ND5k structures are\noptimized via tight-binding density functional theory (DFTB) and their frontier\norbital energies are computed using density functional theory (DFT) with the\nPBE0 hybrid functional. We also compare recent machine learning models for\npredicting frontier orbital energies for similar structures as they have been\ntrained on (interpolation on ND5k), and we test their abilities to extrapolate\npredictions to larger structures. For both the interpolation and extrapolation\ntask, we find best performance using the equivariant graph neural network\nPaiNN. The second best results are achieved with a message passing neural\nnetwork using a tailored set of atomic descriptors proposed here.\n", "date": "2022-09-30", "categories": []}
{"id": "2404.13215", "title": "Machine Learning-Guided Design of Non-Reciprocal and Asymmetric Elastic\n  Chiral Metamaterials", "abstract": "  There has been significant recent interest in the mechanics community to\ndesign structures that can either violate reciprocity, or exhibit elastic\nasymmetry or odd elasticity. While these properties are highly desirable to\nenable mechanical metamaterials to exhibit novel wave propagation phenomena, it\nremains an open question as to how to design passive structures that exhibit\nboth significant non-reciprocity and elastic asymmetry. In this paper, we first\ndefine several design spaces for chiral metamaterials leveraging specific\ndesign parameters, including the ligament contact angles, the ligament shape,\nand circle radius. Having defined the design spaces, we then leverage machine\nlearning approaches, and specifically Bayesian optimization, to determine\noptimally performing designs within each design space satisfying maximal\nnon-reciprocity or stiffness asymmetry. Finally, we perform multi-objective\noptimization by determining the Pareto optimum and find chiral metamaterials\nthat simultaneously exhibit high non-reciprocity and stiffness asymmetry. Our\nanalysis of the underlying mechanisms reveals that chiral metamaterials that\ncan display multiple different contact states under loading in different\ndirections are able to simultaneously exhibit both high non-reciprocity and\nstiffness asymmetry. Overall, this work demonstrates the effectiveness of\nemploying ML to bring insights to a novel domain with limited prior\ninformation, and more generally will pave the way for metamaterials with unique\nproperties and functionality in directing and guiding mechanical wave energy.\n", "date": "2024-04-19", "categories": []}
{"id": "2404.09470", "title": "LatticeML: A data-driven application for predicting the effective Young\n  Modulus of high temperature graph based architected materials", "abstract": "  Architected materials with their unique topology and geometry offer the\npotential to modify physical and mechanical properties. Machine learning can\naccelerate the design and optimization of these materials by identifying\noptimal designs and forecasting performance. This work presents LatticeML, a\ndata-driven application for predicting the effective Young's Modulus of\nhigh-temperature graph-based architected materials. The study considers eleven\ngraph-based lattice structures with two high-temperature alloys, Ti-6Al-4V and\nInconel 625. Finite element simulations were used to compute the effective\nYoung's Modulus of the 2x2x2 unit cell configurations. A machine learning\nframework was developed to predict Young's Modulus, involving data collection,\npreprocessing, implementation of regression models, and deployment of the\nbest-performing model. Five supervised learning algorithms were evaluated, with\nthe XGBoost Regressor achieving the highest accuracy (MSE = 2.7993, MAE =\n1.1521, R-squared = 0.9875). The application uses the Streamlit framework to\ncreate an interactive web interface, allowing users to input material and\ngeometric parameters and obtain predicted Young's Modulus values.\n", "date": "2024-04-15", "categories": []}
{"id": "2404.08657", "title": "Advancing Extrapolative Predictions of Material Properties through\n  Learning to Learn", "abstract": "  Recent advancements in machine learning have showcased its potential to\nsignificantly accelerate the discovery of new materials. Central to this\nprogress is the development of rapidly computable property predictors, enabling\nthe identification of novel materials with desired properties from vast\nmaterial spaces. However, the limited availability of data resources poses a\nsignificant challenge in data-driven materials research, particularly hindering\nthe exploration of innovative materials beyond the boundaries of existing data.\nWhile machine learning predictors are inherently interpolative, establishing a\ngeneral methodology to create an extrapolative predictor remains a fundamental\nchallenge, limiting the search for innovative materials beyond existing data\nboundaries. In this study, we leverage an attention-based architecture of\nneural networks and meta-learning algorithms to acquire extrapolative\ngeneralization capability. The meta-learners, experienced repeatedly with\narbitrarily generated extrapolative tasks, can acquire outstanding\ngeneralization capability in unexplored material spaces. Through the tasks of\npredicting the physical properties of polymeric materials and hybrid\norganic--inorganic perovskites, we highlight the potential of such\nextrapolatively trained models, particularly with their ability to rapidly\nadapt to unseen material domains in transfer learning scenarios.\n", "date": "2024-03-25", "categories": []}
{"id": "2402.03992", "title": "Space Group Constrained Crystal Generation", "abstract": "  Crystals are the foundation of numerous scientific and industrial\napplications. While various learning-based approaches have been proposed for\ncrystal generation, existing methods seldom consider the space group constraint\nwhich is crucial in describing the geometry of crystals and closely relevant to\nmany desirable properties. However, considering space group constraint is\nchallenging owing to its diverse and nontrivial forms. In this paper, we reduce\nthe space group constraint into an equivalent formulation that is more\ntractable to be handcrafted into the generation process. In particular, we\ntranslate the space group constraint into two parts: the basis constraint of\nthe invariant logarithmic space of the lattice matrix and the Wyckoff position\nconstraint of the fractional coordinates. Upon the derived constraints, we then\npropose DiffCSP++, a novel diffusion model that has enhanced a previous work\nDiffCSP by further taking space group constraint into account. Experiments on\nseveral popular datasets verify the benefit of the involvement of the space\ngroup constraint, and show that our DiffCSP++ achieves promising performance on\ncrystal structure prediction, ab initio crystal generation and controllable\ngeneration with customized space groups.\n", "date": "2024-02-06", "categories": []}
{"id": "2404.04810", "title": "AlphaCrystal-II: Distance matrix based crystal structure prediction\n  using deep learning", "abstract": "  Computational prediction of stable crystal structures has a profound impact\non the large-scale discovery of novel functional materials. However, predicting\nthe crystal structure solely from a material's composition or formula is a\npromising yet challenging task, as traditional ab initio crystal structure\nprediction (CSP) methods rely on time-consuming global searches and\nfirst-principles free energy calculations. Inspired by the recent success of\ndeep learning approaches in protein structure prediction, which utilize\npairwise amino acid interactions to describe 3D structures, we present\nAlphaCrystal-II, a novel knowledge-based solution that exploits the abundant\ninter-atomic interaction patterns found in existing known crystal structures.\nAlphaCrystal-II predicts the atomic distance matrix of a target crystal\nmaterial and employs this matrix to reconstruct its 3D crystal structure. By\nleveraging the wealth of inter-atomic relationships of known crystal\nstructures, our approach demonstrates remarkable effectiveness and reliability\nin structure prediction through comprehensive experiments. This work highlights\nthe potential of data-driven methods in accelerating the discovery and design\nof new materials with tailored properties.\n", "date": "2024-04-07", "categories": []}
{"id": "2404.04825", "title": "Gradient-based Design of Computational Granular Crystals", "abstract": "  There is growing interest in engineering unconventional computing devices\nthat leverage the intrinsic dynamics of physical substrates to perform fast and\nenergy-efficient computations. Granular metamaterials are one such substrate\nthat has emerged as a promising platform for building wave-based information\nprocessing devices with the potential to integrate sensing, actuation, and\ncomputation. Their high-dimensional and nonlinear dynamics result in nontrivial\nand sometimes counter-intuitive wave responses that can be shaped by the\nmaterial properties, geometry, and configuration of individual grains. Such\nhighly tunable rich dynamics can be utilized for mechanical computing in\nspecial-purpose applications. However, there are currently no general\nframeworks for the inverse design of large-scale granular materials. Here, we\nbuild upon the similarity between the spatiotemporal dynamics of wave\npropagation in material and the computational dynamics of Recurrent Neural\nNetworks to develop a gradient-based optimization framework for harmonically\ndriven granular crystals. We showcase how our framework can be utilized to\ndesign basic logic gates where mechanical vibrations carry the information at\npredetermined frequencies. We compare our design methodology with classic\ngradient-free methods and find that our approach discovers higher-performing\nconfigurations with less computational effort. Our findings show that a\ngradient-based optimization method can greatly expand the design space of\nmetamaterials and provide the opportunity to systematically traverse the\nparameter space to find materials with the desired functionalities.\n", "date": "2024-04-07", "categories": []}
{"id": "2403.19783", "title": "AlloyBERT: Alloy Property Prediction with Large Language Models", "abstract": "  The pursuit of novel alloys tailored to specific requirements poses\nsignificant challenges for researchers in the field. This underscores the\nimportance of developing predictive techniques for essential physical\nproperties of alloys based on their chemical composition and processing\nparameters. This study introduces AlloyBERT, a transformer encoder-based model\ndesigned to predict properties such as elastic modulus and yield strength of\nalloys using textual inputs. Leveraging the pre-trained RoBERTa encoder model\nas its foundation, AlloyBERT employs self-attention mechanisms to establish\nmeaningful relationships between words, enabling it to interpret human-readable\ninput and predict target alloy properties. By combining a tokenizer trained on\nour textual data and a RoBERTa encoder pre-trained and fine-tuned for this\nspecific task, we achieved a mean squared error (MSE) of 0.00015 on the Multi\nPrincipal Elemental Alloys (MPEA) data set and 0.00611 on the Refractory Alloy\nYield Strength (RAYS) dataset. This surpasses the performance of shallow\nmodels, which achieved a best-case MSE of 0.00025 and 0.0076 on the MPEA and\nRAYS datasets respectively. Our results highlight the potential of language\nmodels in material science and establish a foundational framework for\ntext-based prediction of alloy properties that does not rely on complex\nunderlying representations, calculations, or simulations.\n", "date": "2024-03-28", "categories": []}
{"id": "2403.12659", "title": "Graph Neural Networks for Carbon Dioxide Adsorption Prediction in\n  Aluminium-Exchanged Zeolites", "abstract": "  The ability to efficiently predict adsorption properties of zeolites can be\nof large benefit in accelerating the design process of novel materials. The\nexisting configuration space for these materials is wide, while existing\nmolecular simulation methods are computationally expensive. In this work, we\npropose a model which is 4 to 5 orders of magnitude faster at adsorption\nproperties compared to molecular simulations. To validate the model, we\ngenerated datasets containing various aluminium configurations for the MOR,\nMFI, RHO and ITW zeolites along with their heat of adsorptions and Henry\ncoefficients for CO$_2$, obtained from Monte Carlo simulations. The predictions\nobtained from the Machine Learning model are in agreement with the values\nobtained from the Monte Carlo simulations, confirming that the model can be\nused for property prediction. Furthermore, we show that the model can be used\nfor identifying adsorption sites. Finally, we evaluate the capability of our\nmodel for generating novel zeolite configurations by using it in combination\nwith a genetic algorithm.\n", "date": "2024-03-19", "categories": []}
{"id": "2403.18310", "title": "A thermodynamically consistent physics-informed deep learning material\n  model for short fiber/polymer nanocomposites", "abstract": "  This work proposes a physics-informed deep learning (PIDL)-based constitutive\nmodel for investigating the viscoelastic-viscoplastic behavior of short\nfiber-reinforced nanoparticle-filled epoxies under various ambient conditions.\nThe deep-learning model is trained to enforce thermodynamic principles, leading\nto a thermodynamically consistent constitutive model. To accomplish this, a\nlong short-term memory network is combined with a feed-forward neural network\nto predict internal variables required for characterizing the internal\ndissipation of the nanocomposite materials. In addition, another feed-forward\nneural network is used to indicate the free-energy function, which enables\ndefining the thermodynamic state of the entire system. The PIDL model is\ninitially developed for the three-dimensional case by generating synthetic data\nfrom a classical constitutive model. The model is then trained by extracting\nthe data directly from cyclic loading-unloading experimental tests. Numerical\nexamples show that the PIDL model can accurately predict the mechanical\nbehavior of epoxy-based nanocomposites for different volume fractions of fibers\nand nanoparticles under various hygrothermal conditions.\n", "date": "2024-03-27", "categories": []}
{"id": "2309.15127", "title": "Grad DFT: a software library for machine learning enhanced density\n  functional theory", "abstract": "  Density functional theory (DFT) stands as a cornerstone method in\ncomputational quantum chemistry and materials science due to its remarkable\nversatility and scalability. Yet, it suffers from limitations in accuracy,\nparticularly when dealing with strongly correlated systems. To address these\nshortcomings, recent work has begun to explore how machine learning can expand\nthe capabilities of DFT; an endeavor with many open questions and technical\nchallenges. In this work, we present Grad DFT: a fully differentiable JAX-based\nDFT library, enabling quick prototyping and experimentation with machine\nlearning-enhanced exchange-correlation energy functionals. Grad DFT employs a\npioneering parametrization of exchange-correlation functionals constructed\nusing a weighted sum of energy densities, where the weights are determined\nusing neural networks. Moreover, Grad DFT encompasses a comprehensive suite of\nauxiliary functions, notably featuring a just-in-time compilable and fully\ndifferentiable self-consistent iterative procedure. To support training and\nbenchmarking efforts, we additionally compile a curated dataset of experimental\ndissociation energies of dimers, half of which contain transition metal atoms\ncharacterized by strong electronic correlations. The software library is tested\nagainst experimental results to study the generalization capabilities of a\nneural functional across potential energy surfaces and atomic species, as well\nas the effect of training data noise on the resulting model accuracy.\n", "date": "2023-09-23", "categories": []}
{"id": "2401.16914", "title": "Energy-conserving equivariant GNN for elasticity of lattice architected\n  metamaterials", "abstract": "  Lattices are architected metamaterials whose properties strongly depend on\ntheir geometrical design. The analogy between lattices and graphs enables the\nuse of graph neural networks (GNNs) as a faster surrogate model compared to\ntraditional methods such as finite element modelling. In this work, we generate\na big dataset of structure-property relationships for strut-based lattices. The\ndataset is made available to the community which can fuel the development of\nmethods anchored in physical principles for the fitting of fourth-order\ntensors. In addition, we present a higher-order GNN model trained on this\ndataset. The key features of the model are (i) SE(3) equivariance, and (ii)\nconsistency with the thermodynamic law of conservation of energy. We compare\nthe model to non-equivariant models based on a number of error metrics and\ndemonstrate its benefits in terms of predictive performance and reduced\ntraining requirements. Finally, we demonstrate an example application of the\nmodel to an architected material design task. The methods which we developed\nare applicable to fourth-order tensors beyond elasticity such as piezo-optical\ntensor etc.\n", "date": "2024-01-30", "categories": []}
{"id": "2403.13243", "title": "A Comparative Study of Machine Learning Models Predicting Energetics of\n  Interacting Defects", "abstract": "  Interacting defect systems are ubiquitous in materials under realistic\nscenarios, yet gaining an atomic-level understanding of these systems from a\ncomputational perspective is challenging - it often demands substantial\nresources due to the necessity of employing supercell calculations. While\nmachine learning techniques have shown potential in accelerating materials\nsimulations, their application to systems involving interacting defects remains\nrelatively rare. In this work, we present a comparative study of three\ndifferent methods to predict the free energy change of systems with interacting\ndefects. We leveraging a limited dataset from Density Functional Theory(DFT)\ncalculations to assess the performance models using materials descriptors,\ngraph neural networks and cluster expansion. Our findings indicate that the\ncluster expansion model can achieve precise energetics predictions even with\nthis limited dataset. Furthermore, with synthetic data generate from cluster\nexpansion model at near-DFT levels, we obtained enlarged dataset to assess the\ndemands on data for training accurate prediction models using graph neural\nnetworks for systems featuring interacting defects. A brief discussion of the\ncomputational cost for each method is provided at the end. This research\nprovide a preliminary evaluation of applying machine learning techniques in\nimperfect surface systems.\n", "date": "2024-03-20", "categories": []}
{"id": "2403.10682", "title": "Evaluation of GlassNet for physics-informed machine learning of glass\n  stability and glass-forming ability", "abstract": "  Glasses form the basis of many modern applications and also hold great\npotential for future medical and environmental applications. However, their\nstructural complexity and large composition space make design and optimization\nchallenging for certain applications. Of particular importance for glass\nprocessing is an estimate of a given composition's glass-forming ability (GFA).\nHowever, there remain many open questions regarding the physical mechanisms of\nglass formation, especially in oxide glasses. It is apparent that a proxy for\nGFA would be highly useful in glass processing and design, but identifying such\na surrogate property has proven itself to be difficult. Here, we explore the\napplication of an open-source pre-trained NN model, GlassNet, that can predict\nthe characteristic temperatures necessary to compute glass stability (GS) and\nassess the feasibility of using these physics-informed ML (PIML)-predicted GS\nparameters to estimate GFA. In doing so, we track the uncertainties at each\nstep of the computation - from the original ML prediction errors, to the\ncompounding of errors during GS estimation, and finally to the final estimation\nof GFA. While GlassNet exhibits reasonable accuracy on all individual\nproperties, we observe a large compounding of error in the combination of these\nindividual predictions for the prediction of GS, finding that random forest\nmodels offer similar accuracy to GlassNet. We also breakdown the ML performance\non different glass families and find that the error in GS prediction is\ncorrelated with the error in crystallization peak temperature prediction.\nLastly, we utilize this finding to assess the relationship between\ntop-performing GS parameters and GFA for two ternary glass systems: sodium\nborosilicate and sodium iron phosphate glasses. We conclude that to obtain true\nML predictive capability of GFA, significantly more data needs to be collected.\n", "date": "2024-03-15", "categories": []}
{"id": "2312.11708", "title": "Accelerating the prediction of inorganic surfaces with machine learning\n  interatomic potentials", "abstract": "  The surface properties of solid-state materials often dictate their\nfunctionality, especially for applications where nanoscale effects become\nimportant. The relevant surface(s) and their properties are determined, in\nlarge part, by the materials synthesis or operating conditions. These\nconditions dictate thermodynamic driving forces and kinetic rates responsible\nfor yielding the observed surface structure and morphology. Computational\nsurface science methods have long been applied to connect thermochemical\nconditions to surface phase stability, particularly in the heterogeneous\ncatalysis and thin film growth communities. This review provides a brief\nintroduction to first-principles approaches to compute surface phase diagrams\nbefore introducing emerging data-driven approaches. The remainder of the review\nfocuses on the application of machine learning, predominantly in the form of\nlearned interatomic potentials, to study complex surfaces. As machine learning\nalgorithms and large datasets on which to train them become more commonplace in\nmaterials science, computational methods are poised to become even more\npredictive and powerful for modeling the complexities of inorganic surfaces at\nthe nanoscale.\n", "date": "2023-12-18", "categories": []}
{"id": "2306.08695", "title": "A generative artificial intelligence framework based on a molecular\n  diffusion model for the design of metal-organic frameworks for carbon capture", "abstract": "  Metal-organic frameworks (MOFs) exhibit great promise for CO2 capture.\nHowever, finding the best performing materials poses computational and\nexperimental grand challenges in view of the vast chemical space of potential\nbuilding blocks. Here, we introduce GHP-MOFassemble, a generative artificial\nintelligence (AI), high performance framework for the rational and accelerated\ndesign of MOFs with high CO2 adsorption capacity and synthesizable linkers.\nGHP-MOFassemble generates novel linkers, assembled with one of three\npre-selected metal nodes (Cu paddlewheel, Zn paddlewheel, Zn tetramer) into\nMOFs in a primitive cubic topology. GHP-MOFassemble screens and validates\nAI-generated MOFs for uniqueness, synthesizability, structural validity, uses\nmolecular dynamics simulations to study their stability and chemical\nconsistency, and crystal graph neural networks and Grand Canonical Monte Carlo\nsimulations to quantify their CO2 adsorption capacities. We present the top six\nAI-generated MOFs with CO2 capacities greater than 2 $m mol/g$, i.e., higher\nthan 96.9% of structures in the hypothetical MOF dataset.\n", "date": "2023-06-14", "categories": []}
{"id": "2211.12020", "title": "PhAST: Physics-Aware, Scalable, and Task-specific GNNs for Accelerated\n  Catalyst Design", "abstract": "  Mitigating the climate crisis requires a rapid transition towards\nlower-carbon energy. Catalyst materials play a crucial role in the\nelectrochemical reactions involved in numerous industrial processes key to this\ntransition, such as renewable energy storage and electrofuel synthesis. To\nreduce the energy spent on such activities, we must quickly discover more\nefficient catalysts to drive electrochemical reactions. Machine learning (ML)\nholds the potential to efficiently model materials properties from large\namounts of data, accelerating electrocatalyst design. The Open Catalyst Project\nOC20 dataset was constructed to that end. However, ML models trained on OC20\nare still neither scalable nor accurate enough for practical applications. In\nthis paper, we propose task-specific innovations applicable to most\narchitectures, enhancing both computational efficiency and accuracy. This\nincludes improvements in (1) the graph creation step, (2) atom representations,\n(3) the energy prediction head, and (4) the force prediction head. We describe\nthese contributions, referred to as PhAST, and evaluate them thoroughly on\nmultiple architectures. Overall, PhAST improves energy MAE by 4 to 42$\\%$ while\ndividing compute time by 3 to 8$\\times$ depending on the targeted task/model.\nPhAST also enables CPU training, leading to 40$\\times$ speedups in highly\nparallelized settings. Python package: \\url{https://phast.readthedocs.io}.\n", "date": "2022-11-22", "categories": []}
{"id": "2403.06955", "title": "Accurate Crystal Structure Prediction of New 2D Hybrid Organic Inorganic\n  Perovskites", "abstract": "  Low dimensional hybrid organic-inorganic perovskites (HOIPs) represent a\npromising class of electronically active materials for both light absorption\nand emission. The design space of HOIPs is extremely large, since a diverse\nspace of organic cations can be combined with different inorganic frameworks.\nThis immense design space allows for tunable electronic and mechanical\nproperties, but also necessitates the development of new tools for in silico\nhigh throughput analysis of candidate structures. In this work, we present an\naccurate, efficient, transferable and widely applicable machine learning\ninteratomic potential (MLIP) for predicting the structure of new 2D HOIPs.\nUsing the MACE architecture, an MLIP is trained on 86 diverse experimentally\nreported HOIP structures. The model is tested on 73 unseen perovskite\ncompositions, and achieves chemical accuracy with respect to the reference\nelectronic structure method. Our model is then combined with a simple random\nstructure search algorithm to predict the structure of hypothetical HOIPs given\nonly the proposed composition. Success is demonstrated by correctly and\nreliably recovering the crystal structure of a set of experimentally known 2D\nperovskites. Such a random structure search is impossible with ab initio\nmethods due to the associated computational cost, but is relatively inexpensive\nwith the MACE potential. Finally, the procedure is used to predict the\nstructure formed by a new organic cation with no previously known corresponding\nperovskite. Laboratory synthesis of the new hybrid perovskite confirms the\naccuracy of our prediction. This capability, applied at scale, enables\nefficient screening of thousands of combinations of organic cations and\ninorganic layers.\n", "date": "2024-03-11", "categories": []}
{"id": "2403.05119", "title": "Estimation of Electronic Band Gap Energy From Material Properties Using\n  Machine Learning", "abstract": "  Machine learning techniques are utilized to estimate the electronic band gap\nenergy and forecast the band gap category of materials based on experimentally\nquantifiable properties. The determination of band gap energy is critical for\ndiscerning various material properties, such as its metallic nature, and\npotential applications in electronic and optoelectronic devices. While\nnumerical methods exist for computing band gap energy, they often entail high\ncomputational costs and have limitations in accuracy and scalability. A machine\nlearning-driven model capable of swiftly predicting material band gap energy\nusing easily obtainable experimental properties would offer a superior\nalternative to conventional density functional theory (DFT) methods. Our model\ndoes not require any preliminary DFT-based calculation or knowledge of the\nstructure of the material. We present a scheme for improving the performance of\nsimple regression and classification models by partitioning the dataset into\nmultiple clusters. A new evaluation scheme for comparing the performance of\nML-based models in material sciences involving both regression and\nclassification tasks is introduced based on traditional evaluation metrics. It\nis shown that on this new evaluation metric, our method of clustering the\ndataset results in better performance.\n", "date": "2024-03-08", "categories": []}
{"id": "2309.04475", "title": "Crystal Structure Prediction by Joint Equivariant Diffusion", "abstract": "  Crystal Structure Prediction (CSP) is crucial in various scientific\ndisciplines. While CSP can be addressed by employing currently-prevailing\ngenerative models (e.g. diffusion models), this task encounters unique\nchallenges owing to the symmetric geometry of crystal structures -- the\ninvariance of translation, rotation, and periodicity. To incorporate the above\nsymmetries, this paper proposes DiffCSP, a novel diffusion model to learn the\nstructure distribution from stable crystals. To be specific, DiffCSP jointly\ngenerates the lattice and atom coordinates for each crystal by employing a\nperiodic-E(3)-equivariant denoising model, to better model the crystal\ngeometry. Notably, different from related equivariant generative approaches,\nDiffCSP leverages fractional coordinates other than Cartesian coordinates to\nrepresent crystals, remarkably promoting the diffusion and the generation\nprocess of atom positions. Extensive experiments verify that our DiffCSP\nsignificantly outperforms existing CSP methods, with a much lower computation\ncost in contrast to DFT-based methods. Moreover, the superiority of DiffCSP is\nalso observed when it is extended for ab initio crystal generation.\n", "date": "2023-07-30", "categories": []}
{"id": "2310.01678", "title": "Score dynamics: scaling molecular dynamics with picoseconds timestep via\n  conditional diffusion model", "abstract": "  We propose score dynamics (SD), a general framework for learning accelerated\nevolution operators with large timesteps from molecular-dynamics simulations.\nSD is centered around scores, or derivatives of the transition log-probability\nwith respect to the dynamical degrees of freedom. The latter play the same role\nas force fields in MD but are used in denoising diffusion probability models to\ngenerate discrete transitions of the dynamical variables in an SD timestep,\nwhich can be orders of magnitude larger than a typical MD timestep. In this\nwork, we construct graph neural network based score dynamics models of\nrealistic molecular systems that are evolved with 10~ps timesteps. We\ndemonstrate the efficacy of score dynamics with case studies of alanine\ndipeptide and short alkanes in aqueous solution. Both equilibrium predictions\nderived from the stationary distributions of the conditional probability and\nkinetic predictions for the transition rates and transition paths are in good\nagreement with MD. Our current SD implementation is about two orders of\nmagnitude faster than the MD counterpart for the systems studied in this work.\nOpen challenges and possible future remedies to improve score dynamics are also\ndiscussed.\n", "date": "2023-10-02", "categories": []}
{"id": "2310.15124", "title": "Mixed-Variable Global Sensitivity Analysis For Knowledge Discovery And\n  Efficient Combinatorial Materials Design", "abstract": "  Global Sensitivity Analysis (GSA) is the study of the influence of any given\ninputs on the outputs of a model. In the context of engineering design, GSA has\nbeen widely used to understand both individual and collective contributions of\ndesign variables on the design objectives. So far, global sensitivity studies\nhave often been limited to design spaces with only quantitative (numerical)\ndesign variables. However, many engineering systems also contain, if not only,\nqualitative (categorical) design variables in addition to quantitative design\nvariables. In this paper, we integrate Latent Variable Gaussian Process (LVGP)\nwith Sobol' analysis to develop the first metamodel-based mixed-variable GSA\nmethod. Through numerical case studies, we validate and demonstrate the\neffectiveness of our proposed method for mixed-variable problems. Furthermore,\nwhile the proposed GSA method is general enough to benefit various engineering\ndesign applications, we integrate it with multi-objective Bayesian optimization\n(BO) to create a sensitivity-aware design framework in accelerating the Pareto\nfront design exploration for metal-organic framework (MOF) materials with\nmany-level combinatorial design spaces. Although MOFs are constructed only from\nqualitative variables that are notoriously difficult to design, our method can\nutilize sensitivity analysis to navigate the optimization in the many-level\nlarge combinatorial design space, greatly expediting the exploration of novel\nMOF candidates.\n", "date": "2023-10-23", "categories": []}
{"id": "2401.05638", "title": "MatSAM: Efficient Extraction of Microstructures of Materials via Visual\n  Large Model", "abstract": "  Efficient and accurate extraction of microstructures in micrographs of\nmaterials is essential in process optimization and the exploration of\nstructure-property relationships. Deep learning-based image segmentation\ntechniques that rely on manual annotation are laborious and time-consuming and\nhardly meet the demand for model transferability and generalization on various\nsource images. Segment Anything Model (SAM), a large visual model with powerful\ndeep feature representation and zero-shot generalization capabilities, has\nprovided new solutions for image segmentation. In this paper, we propose\nMatSAM, a general and efficient microstructure extraction solution based on\nSAM. A simple yet effective point-based prompt generation strategy is designed,\ngrounded on the distribution and shape of microstructures. Specifically, in an\nunsupervised and training-free way, it adaptively generates prompt points for\ndifferent microscopy images, fuses the centroid points of the coarsely\nextracted region of interest (ROI) and native grid points, and integrates\ncorresponding post-processing operations for quantitative characterization of\nmicrostructures of materials. For common microstructures including grain\nboundary and multiple phases, MatSAM achieves superior zero-shot segmentation\nperformance to conventional rule-based methods and is even preferable to\nsupervised learning methods evaluated on 16 microscopy datasets whose\nmicrographs are imaged by the optical microscope (OM) and scanning electron\nmicroscope (SEM). Especially, on 4 public datasets, MatSAM shows unexpected\ncompetitive segmentation performance against their specialist models. We\nbelieve that, without the need for human labeling, MatSAM can significantly\nreduce the cost of quantitative characterization and statistical analysis of\nextensive microstructures of materials, and thus accelerate the design of new\nmaterials.\n", "date": "2024-01-11", "categories": []}
{"id": "2403.00259", "title": "Deciphering diffuse scattering with machine learning and the equivariant\n  foundation model: The case of molten FeO", "abstract": "  Bridging the gap between diffuse x-ray or neutron scattering measurements and\npredicted structures derived from atom-atom pair potentials in disordered\nmaterials, has been a longstanding challenge in condensed matter physics. This\nperspective gives a brief overview of the traditional approaches employed over\nthe past several decades. Namely, the use of approximate interatomic pair\npotentials that relate 3-dimensional structural models to the measured\nstructure factor and its associated pair distribution function. The use of\nmachine learned interatomic potentials has grown in the past few years, and has\nbeen particularly successful in the cases of ionic and oxide systems. Recent\nadvances in large scale sampling, along with a direct integration of scattering\nmeasurements into the model development, has provided improved agreement\nbetween experiments and large-scale models calculated with quantum mechanical\naccuracy. However, details of local polyhedral bonding and connectivity in\nmeta-stable disordered systems still require improvement. Here we leverage\nMACE-MP-0; a newly introduced equivariant foundation model and validate the\nresults against high-quality experimental scattering data for the case of\nmolten iron(II) oxide (FeO). These preliminary results suggest that the\nemerging foundation model has the potential to surpass the traditional\nlimitations of classical interatomic potentials.\n", "date": "2024-03-01", "categories": []}
{"id": "2402.17806", "title": "Material Microstructure Design Using VAE-Regression with Multimodal\n  Prior", "abstract": "  We propose a variational autoencoder (VAE)-based model for building forward\nand inverse structure-property linkages, a problem of paramount importance in\ncomputational materials science. Our model systematically combines VAE with\nregression, linking the two models through a two-level prior conditioned on the\nregression variables. The regression loss is optimized jointly with the\nreconstruction loss of the variational autoencoder, learning microstructure\nfeatures relevant for property prediction and reconstruction. The resultant\nmodel can be used for both forward and inverse prediction i.e., for predicting\nthe properties of a given microstructure as well as for predicting the\nmicrostructure required to obtain given properties. Since the inverse problem\nis ill-posed (one-to-many), we derive the objective function using a\nmulti-modal Gaussian mixture prior enabling the model to infer multiple\nmicrostructures for a target set of properties. We show that for forward\nprediction, our model is as accurate as state-of-the-art forward-only models.\nAdditionally, our method enables direct inverse inference. We show that the\nmicrostructures inferred using our model achieve desired properties reasonably\naccurately, avoiding the need for expensive optimization loops.\n", "date": "2024-02-27", "categories": []}
{"id": "2402.15815", "title": "A Generative Machine Learning Model for Material Microstructure 3D\n  Reconstruction and Performance Evaluation", "abstract": "  The reconstruction of 3D microstructures from 2D slices is considered to hold\nsignificant value in predicting the spatial structure and physical properties\nof materials.The dimensional extension from 2D to 3D is viewed as a highly\nchallenging inverse problem from the current technological\nperspective.Recently,methods based on generative adversarial networks have\ngarnered widespread attention.However,they are still hampered by numerous\nlimitations,including oversimplified models,a requirement for a substantial\nnumber of training samples,and difficulties in achieving model convergence\nduring training.In light of this,a novel generative model that integrates the\nmultiscale properties of U-net with and the generative capabilities of GAN has\nbeen proposed.Based on this,the innovative construction of a multi-scale\nchannel aggregation module,a multi-scale hierarchical feature aggregation\nmodule and a convolutional block attention mechanism can better capture the\nproperties of the material microstructure and extract the image information.The\nmodel's accuracy is further improved by combining the image regularization loss\nwith the Wasserstein distance loss.In addition,this study utilizes the\nanisotropy index to accurately distinguish the nature of the image,which can\nclearly determine the isotropy and anisotropy of the image.It is also the first\ntime that the generation quality of material samples from different domains is\nevaluated and the performance of the model itself is compared.The experimental\nresults demonstrate that the present model not only shows a very high\nsimilarity between the generated 3D structures and real samples but is also\nhighly consistent with real data in terms of statistical data analysis.\n", "date": "2024-02-24", "categories": []}
{"id": "2311.11990", "title": "Machine-Learned Atomic Cluster Expansion Potentials for Fast and\n  Quantum-Accurate Thermal Simulations of Wurtzite AlN", "abstract": "  Using the atomic cluster expansion (ACE) framework, we develop a machine\nlearning interatomic potential for fast and accurately modelling the phonon\ntransport properties of wurtzite aluminum nitride. The predictive power of the\nACE potential against density functional theory (DFT) is demonstrated across a\nbroad range of properties of w-AlN, including ground-state lattice parameters,\nspecific heat capacity, coefficients of thermal expansion, bulk modulus, and\nharmonic phonon dispersions. Validation of lattice thermal conductivity is\nfurther carried out by comparing the ACE-predicted values to the DFT\ncalculations and experiments, exhibiting the overall capability of our ACE\npotential in sufficiently describing anharmonic phonon interactions. As a\npractical application, we perform a lattice dynamics analysis using the\npotential to unravel the effects of biaxial strains on thermal conductivity and\nphonon properties of w-AlN, which is identified as a significant tuning factor\nfor near-junction thermal design of w-AlN-based electronics.\n", "date": "2023-11-20", "categories": []}
{"id": "2309.15126", "title": "From Peptides to Nanostructures: A Euclidean Transformer for Fast and\n  Stable Machine Learned Force Fields", "abstract": "  Recent years have seen vast progress in the development of machine learned\nforce fields (MLFFs) based on ab-initio reference calculations. Despite\nachieving low test errors, the reliability of MLFFs in molecular dynamics (MD)\nsimulations is facing growing scrutiny due to concerns about instability over\nextended simulation timescales. Our findings suggest a potential connection\nbetween robustness to cumulative inaccuracies and the use of equivariant\nrepresentations in MLFFs, but the computational cost associated with these\nrepresentations can limit this advantage in practice. To address this, we\npropose a transformer architecture called SO3krates that combines sparse\nequivariant representations (Euclidean variables) with a self-attention\nmechanism that separates invariant and equivariant information, eliminating the\nneed for expensive tensor products. SO3krates achieves a unique combination of\naccuracy, stability, and speed that enables insightful analysis of quantum\nproperties of matter on extended time and system size scales. To showcase this\ncapability, we generate stable MD trajectories for flexible peptides and\nsupra-molecular structures with hundreds of atoms. Furthermore, we investigate\nthe PES topology for medium-sized chainlike molecules (e.g., small peptides) by\nexploring thousands of minima. Remarkably, SO3krates demonstrates the ability\nto strike a balance between the conflicting demands of stability and the\nemergence of new minimum-energy conformations beyond the training data, which\nis crucial for realistic exploration tasks in the field of biochemistry.\n", "date": "2023-09-21", "categories": []}
{"id": "2302.14770", "title": "Completeness of Atomic Structure Representations", "abstract": "  In this paper, we address the challenge of obtaining a comprehensive and\nsymmetric representation of point particle groups, such as atoms in a molecule,\nwhich is crucial in physics and theoretical chemistry. The problem has become\neven more important with the widespread adoption of machine-learning techniques\nin science, as it underpins the capacity of models to accurately reproduce\nphysical relationships while being consistent with fundamental symmetries and\nconservation laws. However, some of the descriptors that are commonly used to\nrepresent point clouds -- most notably those based on discretized correlations\nof the neighbor density, that underpin most of the existing ML models of matter\nat the atomic scale -- are unable to distinguish between special arrangements\nof particles in three dimensions. This makes it impossible to machine learn\ntheir properties. Atom-density correlations are provably complete in the limit\nin which they simultaneously describe the mutual relationship between all\natoms, which is impractical. We present a novel approach to construct\ndescriptors of \\emph{finite} correlations based on the relative arrangement of\nparticle triplets, which can be employed to create symmetry-adapted models with\nuniversal approximation capabilities, which have the resolution of the neighbor\ndiscretization as the sole convergence parameter. Our strategy is demonstrated\non a class of atomic arrangements that are specifically built to defy a broad\nclass of conventional symmetric descriptors, showcasing its potential for\naddressing their limitations.\n", "date": "2023-02-28", "categories": []}
{"id": "2402.04030", "title": "Reducing the Cost of Quantum Chemical Data By Backpropagating Through\n  Density Functional Theory", "abstract": "  Density Functional Theory (DFT) accurately predicts the quantum chemical\nproperties of molecules, but scales as $O(N_{\\text{electrons}}^3)$. Sch\\\"utt et\nal. (2019) successfully approximate DFT 1000x faster with Neural Networks (NN).\nArguably, the biggest problem one faces when scaling to larger molecules is the\ncost of DFT labels. For example, it took years to create the PCQ dataset\n(Nakata & Shimazaki, 2017) on which subsequent NNs are trained within a week.\nDFT labels molecules by minimizing energy $E(\\cdot )$ as a \"loss function.\" We\nbypass dataset creation by directly training NNs with $E(\\cdot )$ as a loss\nfunction. For comparison, Sch\\\"utt et al. (2019) spent 626 hours creating a\ndataset on which they trained their NN for 160h, for a total of 786h; our\nmethod achieves comparable performance within 31h.\n", "date": "2024-02-06", "categories": []}
{"id": "2305.01401", "title": "Stress and heat flux via automatic differentiation", "abstract": "  Machine-learning potentials provide computationally efficient and accurate\napproximations of the Born-Oppenheimer potential energy surface. This potential\ndetermines many materials properties and simulation techniques usually require\nits gradients, in particular forces and stress for molecular dynamics, and heat\nflux for thermal transport properties. Recently developed potentials feature\nhigh body order and can include equivariant semi-local interactions through\nmessage-passing mechanisms. Due to their complex functional forms, they rely on\nautomatic differentiation (AD), overcoming the need for manual implementations\nor finite-difference schemes to evaluate gradients. This study demonstrates a\nunified AD approach to obtain forces, stress, and heat flux for such\npotentials, and provides a model-independent implementation. The method is\ntested on the Lennard-Jones potential, and then applied to predict cohesive\nproperties and thermal conductivity of tin selenide using an equivariant\nmessage-passing neural network potential.\n", "date": "2023-05-02", "categories": []}
{"id": "2312.03687", "title": "MatterGen: a generative model for inorganic materials design", "abstract": "  The design of functional materials with desired properties is essential in\ndriving technological advances in areas like energy storage, catalysis, and\ncarbon capture. Generative models provide a new paradigm for materials design\nby directly generating entirely novel materials given desired property\nconstraints. Despite recent progress, current generative models have low\nsuccess rate in proposing stable crystals, or can only satisfy a very limited\nset of property constraints. Here, we present MatterGen, a model that generates\nstable, diverse inorganic materials across the periodic table and can further\nbe fine-tuned to steer the generation towards a broad range of property\nconstraints. To enable this, we introduce a new diffusion-based generative\nprocess that produces crystalline structures by gradually refining atom types,\ncoordinates, and the periodic lattice. We further introduce adapter modules to\nenable fine-tuning towards any given property constraints with a labeled\ndataset. Compared to prior generative models, structures produced by MatterGen\nare more than twice as likely to be novel and stable, and more than 15 times\ncloser to the local energy minimum. After fine-tuning, MatterGen successfully\ngenerates stable, novel materials with desired chemistry, symmetry, as well as\nmechanical, electronic and magnetic properties. Finally, we demonstrate\nmulti-property materials design capabilities by proposing structures that have\nboth high magnetic density and a chemical composition with low supply-chain\nrisk. We believe that the quality of generated materials and the breadth of\nMatterGen's capabilities represent a major advancement towards creating a\nuniversal generative model for materials design.\n", "date": "2023-12-06", "categories": []}
{"id": "2401.08032", "title": "Structure-based out-of-distribution (OOD) materials property prediction:\n  a benchmark study", "abstract": "  In real-world material research, machine learning (ML) models are usually\nexpected to predict and discover novel exceptional materials that deviate from\nthe known materials. It is thus a pressing question to provide an objective\nevaluation of ML model performances in property prediction of\nout-of-distribution (OOD) materials that are different from the training set\ndistribution. Traditional performance evaluation of materials property\nprediction models through random splitting of the dataset frequently results in\nartificially high performance assessments due to the inherent redundancy of\ntypical material datasets. Here we present a comprehensive benchmark study of\nstructure-based graph neural networks (GNNs) for extrapolative OOD materials\nproperty prediction. We formulate five different categories of OOD ML problems\nfor three benchmark datasets from the MatBench study. Our extensive experiments\nshow that current state-of-the-art GNN algorithms significantly underperform\nfor the OOD property prediction tasks on average compared to their baselines in\nthe MatBench study, demonstrating a crucial generalization gap in realistic\nmaterial prediction tasks. We further examine the latent physical spaces of\nthese GNN models and identify the sources of CGCNN, ALIGNN, and DeeperGATGNN's\nsignificantly more robust OOD performance than those of the current best models\nin the MatBench study (coGN and coNGN), and provide insights to improve their\nperformance.\n", "date": "2024-01-16", "categories": []}
{"id": "2401.05568", "title": "Phase discovery with active learning: Application to structural phase\n  transitions in equiatomic NiTi", "abstract": "  Nickel titanium (NiTi) is a protypical shape-memory alloy used in a range of\nbiomedical and engineering devices, but direct molecular dynamics simulations\nof the martensitic B19' -> B2 phase transition driving its shape-memory\nbehavior are rare and have relied on classical force fields with limited\naccuracy. Here, we train four machine-learned force fields for equiatomic NiTi\nbased on the LDA, PBE, PBEsol, and SCAN DFT functionals. The models are trained\non the fly during NPT molecular dynamics, with DFT calculations and model\nupdates performed automatically whenever the uncertainty of a local energy\nprediction exceeds a chosen threshold. The models achieve accuracies of 1-2\nmeV/atom during training and are shown to closely track DFT predictions of B2\nand B19' elastic constants and phonon frequencies. Surprisingly, in large-scale\nmolecular dynamics simulations, only the SCAN model predicts a reversible B19'\n-> B2 phase transition, with the LDA, PBE, and PBEsol models predicting a\nreversible transition to a previously uncharacterized low-volume phase, which\nwe hypothesize to be a new stable high-pressure phase. We examine the structure\nof the new phase and estimate its stability on the temperature-pressure phase\ndiagram. This work establishes an automated active learning protocol for\nstudying displacive transformations, reveals important differences between DFT\nfunctionals that can only be detected in large-scale simulations, provides an\naccurate force field for NiTi, and identifies a new phase.\n", "date": "2024-01-10", "categories": []}
{"id": "2401.05848", "title": "Pushing the Pareto front of band gap and permittivity: ML-guided search\n  for dielectric materials", "abstract": "  Materials with high-dielectric constant easily polarize under external\nelectric fields, allowing them to perform essential functions in many modern\nelectronic devices. Their practical utility is determined by two conflicting\nproperties: high dielectric constants tend to occur in materials with narrow\nband gaps, limiting the operating voltage before dielectric breakdown. We\npresent a high-throughput workflow that combines element substitution, ML\npre-screening, ab initio simulation and human expert intuition to efficiently\nexplore the vast space of unknown materials for potential dielectrics, leading\nto the synthesis and characterization of two novel dielectric materials,\nCsTaTeO6 and Bi2Zr2O7. Our key idea is to deploy ML in a multi-objective\noptimization setting with concave Pareto front. While usually considered more\nchallenging than single-objective optimization, we argue and show preliminary\nevidence that the $1/x$-correlation between band gap and permittivity in fact\nmakes the task more amenable to ML methods by allowing separate models for band\ngap and permittivity to each operate in regions of good training support while\nstill predicting materials of exceptional merit. To our knowledge, this is the\nfirst instance of successful ML-guided multi-objective materials optimization\nachieving experimental synthesis and characterization. CsTaTeO6 is a structure\ngenerated via element substitution not present in our reference data sources,\nthus exemplifying successful de-novo materials design. Meanwhile, we report the\nfirst high-purity synthesis and dielectric characterization of Bi2Zr2O7 with a\nband gap of 2.27 eV and a permittivity of 20.5, meeting all target metrics of\nour multi-objective search.\n", "date": "2024-01-11", "categories": []}
{"id": "2401.05206", "title": "Tailoring Frictional Properties of Surfaces Using Diffusion Models", "abstract": "  This Letter introduces an approach for precisely designing surface friction\nproperties using a conditional generative machine learning model, specifically\na diffusion denoising probabilistic model (DDPM). We created a dataset of\nsynthetic surfaces with frictional properties determined by molecular dynamics\nsimulations, which trained the DDPM to predict surface structures from desired\nfrictional outcomes. Unlike traditional trial-and-error and numerical\noptimization methods, our approach directly yields surface designs meeting\nspecified frictional criteria with high accuracy and efficiency. This\nadvancement in material surface engineering demonstrates the potential of\nmachine learning in reducing the iterative nature of surface design processes.\nOur findings not only provide a new pathway for precise surface property\ntailoring but also suggest broader applications in material science where\nsurface characteristics are critical.\n", "date": "2024-01-05", "categories": []}
{"id": "2302.01078", "title": "Computational Discovery of Microstructured Composites with Optimal\n  Stiffness-Toughness Trade-Offs", "abstract": "  The conflict between stiffness and toughness is a fundamental problem in\nengineering materials design. However, the systematic discovery of\nmicrostructured composites with optimal stiffness-toughness trade-offs has\nnever been demonstrated, hindered by the discrepancies between simulation and\nreality and the lack of data-efficient exploration of the entire Pareto front.\nWe introduce a generalizable pipeline that integrates physical experiments,\nnumerical simulations, and artificial neural networks to address both\nchallenges. Without any prescribed expert knowledge of material design, our\napproach implements a nested-loop proposal-validation workflow to bridge the\nsimulation-to-reality gap and discover microstructured composites that are\nstiff and tough with high sample efficiency. Further analysis of Pareto-optimal\ndesigns allows us to automatically identify existing toughness enhancement\nmechanisms, which were previously discovered through trial-and-error or\nbiomimicry. On a broader scale, our method provides a blueprint for\ncomputational design in various research areas beyond solid mechanics, such as\npolymer chemistry, fluid dynamics, meteorology, and robotics.\n", "date": "2023-02-01", "categories": []}
{"id": "2310.19274", "title": "Prediction of Effective Elastic Moduli of Rocks using Graph Neural\n  Networks", "abstract": "  This study presents a Graph Neural Networks (GNNs)-based approach for\npredicting the effective elastic moduli of rocks from their digital CT-scan\nimages. We use the Mapper algorithm to transform 3D digital rock images into\ngraph datasets, encapsulating essential geometrical information. These graphs,\nafter training, prove effective in predicting elastic moduli. Our GNN model\nshows robust predictive capabilities across various graph sizes derived from\nvarious subcube dimensions. Not only does it perform well on the test dataset,\nbut it also maintains high prediction accuracy for unseen rocks and unexplored\nsubcube sizes. Comparative analysis with Convolutional Neural Networks (CNNs)\nreveals the superior performance of GNNs in predicting unseen rock properties.\nMoreover, the graph representation of microstructures significantly reduces GPU\nmemory requirements (compared to the grid representation for CNNs), enabling\ngreater flexibility in the batch size selection. This work demonstrates the\npotential of GNN models in enhancing the prediction accuracy of rock properties\nand boosting the efficiency of digital rock analysis.\n", "date": "2023-10-30", "categories": []}
{"id": "2309.00993", "title": "A Boosted Machine Learning Framework for the Improvement of Phase and\n  Crystal Structure Prediction of High Entropy Alloys Using Thermodynamic and\n  Configurational Parameters", "abstract": "  The reason behind the remarkable properties of High-Entropy Alloys (HEAs) is\nrooted in the diverse phases and the crystal structures they contain. In the\nrealm of material informatics, employing machine learning (ML) techniques to\nclassify phases and crystal structures of HEAs has gained considerable\nsignificance. In this study, we assembled a new collection of 1345 HEAs with\nvarying compositions to predict phases. Within this collection, there were 705\nsets of data that were utilized to predict the crystal structures with the help\nof thermodynamics and electronic configuration. Our study introduces a\nmethodical framework i.e., the Pearson correlation coefficient that helps in\nselecting the strongly co-related features to increase the prediction accuracy.\nThis study employed five distinct boosting algorithms to predict phases and\ncrystal structures, offering an enhanced guideline for improving the accuracy\nof these predictions. Among all these algorithms, XGBoost gives the highest\naccuracy of prediction (94.05%) for phases and LightGBM gives the highest\naccuracy of prediction of crystal structure of the phases (90.07%). The\nquantification of the influence exerted by parameters on the model's accuracy\nwas conducted and a new approach was made to elucidate the contribution of\nindividual parameters in the process of phase prediction and crystal structure\nprediction.\n", "date": "2023-09-02", "categories": []}
{"id": "2305.01754", "title": "Single-model uncertainty quantification in neural network potentials\n  does not consistently outperform model ensembles", "abstract": "  Neural networks (NNs) often assign high confidence to their predictions, even\nfor points far out-of-distribution, making uncertainty quantification (UQ) a\nchallenge. When they are employed to model interatomic potentials in materials\nsystems, this problem leads to unphysical structures that disrupt simulations,\nor to biased statistics and dynamics that do not reflect the true physics.\nDifferentiable UQ techniques can find new informative data and drive active\nlearning loops for robust potentials. However, a variety of UQ techniques,\nincluding newly developed ones, exist for atomistic simulations and there are\nno clear guidelines for which are most effective or suitable for a given case.\nIn this work, we examine multiple UQ schemes for improving the robustness of NN\ninteratomic potentials (NNIPs) through active learning. In particular, we\ncompare incumbent ensemble-based methods against strategies that use single,\ndeterministic NNs: mean-variance estimation, deep evidential regression, and\nGaussian mixture models. We explore three datasets ranging from in-domain\ninterpolative learning to more extrapolative out-of-domain generalization\nchallenges: rMD17, ammonia inversion, and bulk silica glass. Performance is\nmeasured across multiple metrics relating model error to uncertainty. Our\nexperiments show that none of the methods consistently outperformed each other\nacross the various metrics. Ensembling remained better at generalization and\nfor NNIP robustness; MVE only proved effective for in-domain interpolation,\nwhile GMM was better out-of-domain; and evidential regression, despite its\npromise, was not the preferable alternative in any of the cases. More broadly,\ncost-effective, single deterministic models cannot yet consistently match or\noutperform ensembling for uncertainty quantification in NNIPs.\n", "date": "2023-05-02", "categories": []}
{"id": "2312.15136", "title": "Towards End-to-End Structure Solutions from Information-Compromised\n  Diffraction Data via Generative Deep Learning", "abstract": "  The revolution in materials in the past century was built on a knowledge of\nthe atomic arrangements and the structure-property relationship. The sine qua\nnon for obtaining quantitative structural information is single crystal\ncrystallography. However, increasingly we need to solve structures in cases\nwhere the information content in our input signal is significantly degraded,\nfor example, due to orientational averaging of grains, finite size effects due\nto nanostructure, and mixed signals due to sample heterogeneity. Understanding\nthe structure property relationships in such situations is, if anything, more\nimportant and insightful, yet we do not have robust approaches for\naccomplishing it. In principle, machine learning (ML) and deep learning (DL)\nare promising approaches since they augment information in the degraded input\nsignal with prior knowledge learned from large databases of already known\nstructures. Here we present a novel ML approach, a variational query-based\nmulti-branch deep neural network that has the promise to be a robust but\ngeneral tool to address this problem end-to-end. We demonstrate the approach on\ncomputed powder x-ray diffraction (PXRD), along with partial chemical\ncomposition information, as input. We choose as a structural representation a\nmodified electron density we call the Cartesian mapped electron density (CMED),\nthat straightforwardly allows our ML model to learn material structures across\ndifferent chemistries, symmetries and crystal systems. When evaluated on\ntheoretically simulated data for the cubic and trigonal crystal systems, the\nsystem achieves up to $93.4\\%$ average similarity with the ground truth on\nunseen materials, both with known and partially-known chemical composition\ninformation, showing great promise for successful structure solution even from\ndegraded and incomplete input data.\n", "date": "2023-12-23", "categories": []}
{"id": "2305.07251", "title": "Machine-learning-accelerated simulations to enable automatic surface\n  reconstruction", "abstract": "  Understanding material surfaces and interfaces is vital in applications like\ncatalysis or electronics. By combining energies from electronic structure with\nstatistical mechanics, ab initio simulations can in principle predict the\nstructure of material surfaces as a function of thermodynamic variables.\nHowever, accurate energy simulations are prohibitive when coupled to the vast\nphase space that must be statistically sampled. Here, we present a bi-faceted\ncomputational loop to predict surface phase diagrams of multi-component\nmaterials that accelerates both the energy scoring and statistical sampling\nmethods. Fast, scalable, and data-efficient machine learning interatomic\npotentials are trained on high-throughput density-functional theory\ncalculations through closed-loop active learning. Markov-chain Monte Carlo\nsampling in the semi-grand canonical ensemble is enabled by using virtual\nsurface sites. The predicted surfaces for GaN(0001), Si(111), and SrTiO3(001)\nare in agreement with past work and suggest that the proposed strategy can\nmodel complex material surfaces and discover previously unreported surface\nterminations.\n", "date": "2023-05-12", "categories": []}
{"id": "2307.10935", "title": "Inorganic synthesis-structure maps in zeolites with machine learning and\n  crystallographic distances", "abstract": "  Zeolites are inorganic materials known for their diversity of applications,\nsynthesis conditions, and resulting polymorphs. Although their synthesis is\ncontrolled both by inorganic and organic synthesis conditions, computational\nstudies of zeolite synthesis have focused mostly on organic template design. In\nthis work, we use a strong distance metric between crystal structures and\nmachine learning (ML) to create inorganic synthesis maps in zeolites. Starting\nwith 253 known zeolites, we show how the continuous distances between\nframeworks reproduce inorganic synthesis conditions from the literature without\nusing labels such as building units. An unsupervised learning analysis shows\nthat neighboring zeolites according to our metric often share similar inorganic\nsynthesis conditions, even in template-based routes. In combination with ML\nclassifiers, we find synthesis-structure relationships for 14 common inorganic\nconditions in zeolites, namely Al, B, Be, Ca, Co, F, Ga, Ge, K, Mg, Na, P, Si,\nand Zn. By explaining the model predictions, we demonstrate how\n(dis)similarities towards known structures can be used as features for the\nsynthesis space. Finally, we show how these methods can be used to predict\ninorganic synthesis conditions for unrealized frameworks in hypothetical\ndatabases and interpret the outcomes by extracting local structural patterns\nfrom zeolites. In combination with template design, this work can accelerate\nthe exploration of the space of synthesis conditions for zeolites.\n", "date": "2023-07-20", "categories": []}
{"id": "2311.16135", "title": "Use of Deep Neural Networks for Uncertain Stress Functions with\n  Extensions to Impact Mechanics", "abstract": "  Stress-strain curves, or more generally, stress functions, are an extremely\nimportant characterization of a material's mechanical properties. However,\nstress functions are often difficult to derive and are narrowly tailored to a\nspecific material. Further, large deformations, high strain-rates, temperature\nsensitivity, and effect of material parameters compound modeling challenges. We\npropose a generalized deep neural network approach to model stress as a state\nfunction with quantile regression to capture uncertainty. We extend these\nmodels to uniaxial impact mechanics using stochastic differential equations to\ndemonstrate a use case and provide a framework for implementing this\nuncertainty-aware stress function. We provide experiments benchmarking our\napproach against leading constitutive, machine learning, and transfer learning\napproaches to stress and impact mechanics modeling on publicly available and\nnewly presented data sets. We also provide a framework to optimize material\nparameters given multiple competing impact scenarios.\n", "date": "2023-11-03", "categories": []}
{"id": "2312.11487", "title": "Symbolic Learning for Material Discovery", "abstract": "  Discovering new materials is essential to solve challenges in climate change,\nsustainability and healthcare. A typical task in materials discovery is to\nsearch for a material in a database which maximises the value of a function.\nThat function is often expensive to evaluate, and can rely upon a simulation or\nan experiment. Here, we introduce SyMDis, a sample efficient optimisation\nmethod based on symbolic learning, that discovers near-optimal materials in a\nlarge database. SyMDis performs comparably to a state-of-the-art optimiser,\nwhilst learning interpretable rules to aid physical and chemical verification.\nFurthermore, the rules learned by SyMDis generalise to unseen datasets and\nreturn high performing candidates in a zero-shot evaluation, which is difficult\nto achieve with other approaches.\n", "date": "2023-11-30", "categories": []}
{"id": "2312.10996", "title": "Position Paper on Materials Design -- A Modern Approach", "abstract": "  Traditional design cycles for new materials and assemblies have two\nfundamental drawbacks. The underlying physical relationships are often too\ncomplex to be precisely calculated and described. Aside from that, many unknown\nuncertainties, such as exact manufacturing parameters or materials composition,\ndominate the real assembly behavior. Machine learning (ML) methods overcome\nthese fundamental limitations through data-driven learning. In addition, modern\napproaches can specifically increase system knowledge. Representation Learning\nallows the physical, and if necessary, even symbolic interpretation of the\nlearned solution. In this way, the most complex physical relationships can be\nconsidered and quickly described. Furthermore, generative ML approaches can\nsynthesize possible morphologies of the materials based on defined conditions\nto visualize the effects of uncertainties. This modern approach accelerates the\ndesign process for new materials and enables the prediction and interpretation\nof realistic materials behavior.\n", "date": "2023-12-18", "categories": []}
{"id": "2312.07832", "title": "Denoising diffusion-based synthetic generation of three-dimensional (3D)\n  anisotropic microstructures from two-dimensional (2D) micrographs", "abstract": "  Integrated computational materials engineering (ICME) has significantly\nenhanced the systemic analysis of the relationship between microstructure and\nmaterial properties, paving the way for the development of high-performance\nmaterials. However, analyzing microstructure-sensitive material behavior\nremains challenging due to the scarcity of three-dimensional (3D)\nmicrostructure datasets. Moreover, this challenge is amplified if the\nmicrostructure is anisotropic, as this results in anisotropic material\nproperties as well. In this paper, we present a framework for reconstruction of\nanisotropic microstructures solely based on two-dimensional (2D) micrographs\nusing conditional diffusion-based generative models (DGMs). The proposed\nframework involves spatial connection of multiple 2D conditional DGMs, each\ntrained to generate 2D microstructure samples for three different orthogonal\nplanes. The connected multiple reverse diffusion processes then enable\neffective modeling of a Markov chain for transforming noise into a 3D\nmicrostructure sample. Furthermore, a modified harmonized sampling is employed\nto enhance the sample quality while preserving the spatial connection between\nthe slices of anisotropic microstructure samples in 3D space. To validate the\nproposed framework, the 2D-to-3D reconstructed anisotropic microstructure\nsamples are evaluated in terms of both the spatial correlation function and the\nphysical material behavior. The results demonstrate that the framework is\ncapable of reproducing not only the statistical distribution of material phases\nbut also the material properties in 3D space. This highlights the potential\napplication of the proposed 2D-to-3D reconstruction framework in establishing\nmicrostructure-property linkages, which could aid high-throughput material\ndesign for future studies\n", "date": "2023-12-13", "categories": []}
{"id": "2311.13812", "title": "Mechanical Characterization and Inverse Design of Stochastic Architected\n  Metamaterials Using Neural Operators", "abstract": "  Machine learning (ML) is emerging as a transformative tool for the design of\narchitected materials, offering properties that far surpass those achievable\nthrough lab-based trial-and-error methods. However, a major challenge in\ncurrent inverse design strategies is their reliance on extensive computational\nand/or experimental datasets, which becomes particularly problematic for\ndesigning micro-scale stochastic architected materials that exhibit nonlinear\nmechanical behaviors. Here, we introduce a new end-to-end scientific ML\nframework, leveraging deep neural operators (DeepONet), to directly learn the\nrelationship between the complete microstructure and mechanical response of\narchitected metamaterials from sparse but high-quality in situ experimental\ndata. The approach facilitates the inverse design of structures tailored to\nspecific nonlinear mechanical behaviors. Results obtained from spinodal\nmicrostructures, printed using two-photon lithography, reveal that the\nprediction error for mechanical responses is within a range of 5 - 10%. Our\nwork underscores that by employing neural operators with advanced\nmicro-mechanics experimental techniques, the design of complex\nmicro-architected materials with desired properties becomes feasible, even in\nscenarios constrained by data scarcity. Our work marks a significant\nadvancement in the field of materials-by-design, potentially heralding a new\nera in the discovery and development of next-generation metamaterials with\nunparalleled mechanical characteristics derived directly from experimental\ninsights.\n", "date": "2023-11-23", "categories": []}
{"id": "2312.05410", "title": "Rethinking materials simulations: Blending direct numerical simulations\n  with neural operators", "abstract": "  Direct numerical simulations (DNS) are accurate but computationally expensive\nfor predicting materials evolution across timescales, due to the complexity of\nthe underlying evolution equations, the nature of multiscale spatio-temporal\ninteractions, and the need to reach long-time integration. We develop a new\nmethod that blends numerical solvers with neural operators to accelerate such\nsimulations. This methodology is based on the integration of a community\nnumerical solver with a U-Net neural operator, enhanced by a\ntemporal-conditioning mechanism that enables accurate extrapolation and\nefficient time-to-solution predictions of the dynamics. We demonstrate the\neffectiveness of this framework on simulations of microstructure evolution\nduring physical vapor deposition modeled via the phase-field method. Such\nsimulations exhibit high spatial gradients due to the co-evolution of different\nmaterial phases with simultaneous slow and fast materials dynamics. We\nestablish accurate extrapolation of the coupled solver with up to 16.5$\\times$\nspeed-up compared to DNS. This methodology is generalizable to a broad range of\nevolutionary models, from solid mechanics, to fluid dynamics, geophysics,\nclimate, and more.\n", "date": "2023-12-08", "categories": []}
{"id": "2312.05472", "title": "Spectroscopy-Guided Discovery of Three-Dimensional Structures of\n  Disordered Materials with Diffusion Models", "abstract": "  The ability to rapidly develop materials with desired properties has a\ntransformative impact on a broad range of emerging technologies. In this work,\nwe introduce a new framework based on the diffusion model, a recent generative\nmachine learning method to predict 3D structures of disordered materials from a\ntarget property. For demonstration, we apply the model to identify the atomic\nstructures of amorphous carbons ($a$-C) as a representative material system\nfrom the target X-ray absorption near edge structure (XANES) spectra--a common\nexperimental technique to probe atomic structures of materials. We show that\nconditional generation guided by XANES spectra reproduces key features of the\ntarget structures. Furthermore, we show that our model can steer the generative\nprocess to tailor atomic arrangements for a specific XANES spectrum. Finally,\nour generative model exhibits a remarkable scale-agnostic property, thereby\nenabling generation of realistic, large-scale structures through learning from\na small-scale dataset (i.e., with small unit cells). Our work represents a\nsignificant stride in bridging the gap between materials characterization and\natomic structure determination; in addition, it can be leveraged for materials\ndiscovery in exploring various material properties as targeted.\n", "date": "2023-12-09", "categories": []}
{"id": "2304.13038", "title": "Diffusion Probabilistic Model Based Accurate and High-Degree-of-Freedom\n  Metasurface Inverse Design", "abstract": "  Conventional meta-atom designs rely heavily on researchers' prior knowledge\nand trial-and-error searches using full-wave simulations, resulting in\ntime-consuming and inefficient processes. Inverse design methods based on\noptimization algorithms, such as evolutionary algorithms, and topological\noptimizations, have been introduced to design metamaterials. However, none of\nthese algorithms are general enough to fulfill multi-objective tasks. Recently,\ndeep learning methods represented by Generative Adversarial Networks (GANs)\nhave been applied to inverse design of metamaterials, which can directly\ngenerate high-degree-of-freedom meta-atoms based on S-parameter requirements.\nHowever, the adversarial training process of GANs makes the network unstable\nand results in high modeling costs. This paper proposes a novel metamaterial\ninverse design method based on the diffusion probability theory. By learning\nthe Markov process that transforms the original structure into a Gaussian\ndistribution, the proposed method can gradually remove the noise starting from\nthe Gaussian distribution and generate new high-degree-of-freedom meta-atoms\nthat meet S-parameter conditions, which avoids the model instability introduced\nby the adversarial training process of GANs and ensures more accurate and\nhigh-quality generation results. Experiments have proven that our method is\nsuperior to representative methods of GANs in terms of model convergence speed,\ngeneration accuracy, and quality.\n", "date": "2023-04-25", "categories": []}
{"id": "2311.15549", "title": "From Prediction to Action: Critical Role of Performance Estimation for\n  Machine-Learning-Driven Materials Discovery", "abstract": "  Materials discovery driven by statistical property models is an iterative\ndecision process, during which an initial data collection is extended with new\ndata proposed by a model-informed acquisition function--with the goal to\nmaximize a certain \"reward\" over time, such as the maximum property value\ndiscovered so far. While the materials science community achieved much progress\nin developing property models that predict well on average with respect to the\ntraining distribution, this form of in-distribution performance measurement is\nnot directly coupled with the discovery reward. This is because an iterative\ndiscovery process has a shifting reward distribution that is\nover-proportionally determined by the model performance for exceptional\nmaterials. We demonstrate this problem using the example of bulk modulus\nmaximization among double perovskite oxides. We find that the in-distribution\npredictive performance suggests random forests as superior to Gaussian process\nregression, while the results are inverse in terms of the discovery rewards. We\nargue that the lack of proper performance estimation methods from pre-computed\ndata collections is a fundamental problem for improving data-driven materials\ndiscovery, and we propose a novel such estimator that, in contrast to na\\\"ive\nreward estimation, successfully predicts Gaussian processes with the \"expected\nimprovement\" acquisition function as the best out of four options in our\ndemonstrational study for double perovskites. Importantly, it does so without\nrequiring the over thousand ab initio computations that were needed to confirm\nthis prediction.\n", "date": "2023-11-27", "categories": []}
{"id": "2312.04174", "title": "Coherent energy and force uncertainty in deep learning force fields", "abstract": "  In machine learning energy potentials for atomic systems, forces are commonly\nobtained as the negative derivative of the energy function with respect to\natomic positions. To quantify aleatoric uncertainty in the predicted energies,\na widely used modeling approach involves predicting both a mean and variance\nfor each energy value. However, this model is not differentiable under the\nusual white noise assumption, so energy uncertainty does not naturally\ntranslate to force uncertainty. In this work we propose a machine learning\npotential energy model in which energy and force aleatoric uncertainty are\nlinked through a spatially correlated noise process. We demonstrate our\napproach on an equivariant messages passing neural network potential trained on\nenergies and forces on two out-of-equilibrium molecular datasets. Furthermore,\nwe also show how to obtain epistemic uncertainties in this setting based on a\nBayesian interpretation of deep ensemble models.\n", "date": "2023-12-07", "categories": []}
{"id": "2307.05506", "title": "Data-Driven Design for Metamaterials and Multiscale Systems: A Review", "abstract": "  Metamaterials are artificial materials designed to exhibit effective material\nparameters that go beyond those found in nature. Composed of unit cells with\nrich designability that are assembled into multiscale systems, they hold great\npromise for realizing next-generation devices with exceptional, often exotic,\nfunctionalities. However, the vast design space and intricate\nstructure-property relationships pose significant challenges in their design. A\ncompelling paradigm that could bring the full potential of metamaterials to\nfruition is emerging: data-driven design. In this review, we provide a holistic\noverview of this rapidly evolving field, emphasizing the general methodology\ninstead of specific domains and deployment contexts. We organize existing\nresearch into data-driven modules, encompassing data acquisition, machine\nlearning-based unit cell design, and data-driven multiscale optimization. We\nfurther categorize the approaches within each module based on shared\nprinciples, analyze and compare strengths and applicability, explore\nconnections between different modules, and identify open research questions and\nopportunities.\n", "date": "2023-07-01", "categories": []}
{"id": "2303.02876", "title": "Metaheuristic conditional neural network for harvesting skyrmionic\n  metastable states", "abstract": "  We present a metaheuristic conditional neural-network-based method aimed at\nidentifying physically interesting metastable states in a potential energy\nsurface of high rugosity. To demonstrate how this method works, we identify and\nanalyze spin textures with topological charge $Q$ ranging from 1 to $-13$\n(where antiskyrmions have $Q<0$) in the Pd/Fe/Ir(111) system, which we model\nusing a classical atomistic spin Hamiltonian based on parameters computed from\ndensity functional theory. To facilitate the harvest of relevant spin textures,\nwe make use of the newly developed Segment Anything Model (SAM). Spin textures\nwith $Q$ ranging from $-3$ to $-6$ are further analyzed using\nfinite-temperature spin-dynamics simulations. We observe that for temperatures\nup to around 20\\,K, lifetimes longer than 200\\,ps are predicted, and that when\nthese textures decay, new topological spin textures are formed. We also find\nthat the relative stability of the spin textures depend linearly on the\ntopological charge, but only when comparing the most stable antiskyrmions for\neach topological charge. In general, the number of holes (i.e.,\nnon-self-intersecting curves that define closed domain walls in the structure)\nin the spin texture is an important predictor of stability -- the more holes,\nthe less stable is the texture. Methods for systematic identification and\ncharacterization of complex metastable skyrmionic textures -- such as the one\ndemonstrated here -- are highly relevant for advancements in the field of\ntopological spintronics.\n", "date": "2023-03-06", "categories": []}
{"id": "2302.01170", "title": "Timewarp: Transferable Acceleration of Molecular Dynamics by Learning\n  Time-Coarsened Dynamics", "abstract": "  Molecular dynamics (MD) simulation is a widely used technique to simulate\nmolecular systems, most commonly at the all-atom resolution where equations of\nmotion are integrated with timesteps on the order of femtoseconds\n($1\\textrm{fs}=10^{-15}\\textrm{s}$). MD is often used to compute equilibrium\nproperties, which requires sampling from an equilibrium distribution such as\nthe Boltzmann distribution. However, many important processes, such as binding\nand folding, occur over timescales of milliseconds or beyond, and cannot be\nefficiently sampled with conventional MD. Furthermore, new MD simulations need\nto be performed for each molecular system studied. We present Timewarp, an\nenhanced sampling method which uses a normalising flow as a proposal\ndistribution in a Markov chain Monte Carlo method targeting the Boltzmann\ndistribution. The flow is trained offline on MD trajectories and learns to make\nlarge steps in time, simulating the molecular dynamics of $10^{5} -\n10^{6}\\:\\textrm{fs}$. Crucially, Timewarp is transferable between molecular\nsystems: once trained, we show that it generalises to unseen small peptides\n(2-4 amino acids) at all-atom resolution, exploring their metastable states and\nproviding wall-clock acceleration of sampling compared to standard MD. Our\nmethod constitutes an important step towards general, transferable algorithms\nfor accelerating MD.\n", "date": "2023-02-02", "categories": []}
{"id": "2310.03121", "title": "OpenMM 8: Molecular Dynamics Simulation with Machine Learning Potentials", "abstract": "  Machine learning plays an important and growing role in molecular simulation.\nThe newest version of the OpenMM molecular dynamics toolkit introduces new\nfeatures to support the use of machine learning potentials. Arbitrary PyTorch\nmodels can be added to a simulation and used to compute forces and energy. A\nhigher-level interface allows users to easily model their molecules of interest\nwith general purpose, pretrained potential functions. A collection of optimized\nCUDA kernels and custom PyTorch operations greatly improves the speed of\nsimulations. We demonstrate these features on simulations of cyclin-dependent\nkinase 8 (CDK8) and the green fluorescent protein (GFP) chromophore in water.\nTaken together, these features make it practical to use machine learning to\nimprove the accuracy of simulations at only a modest increase in cost.\n", "date": "2023-10-04", "categories": []}
{"id": "2311.17894", "title": "Learning and Controlling Silicon Dopant Transitions in Graphene using\n  Scanning Transmission Electron Microscopy", "abstract": "  We introduce a machine learning approach to determine the transition dynamics\nof silicon atoms on a single layer of carbon atoms, when stimulated by the\nelectron beam of a scanning transmission electron microscope (STEM). Our method\nis data-centric, leveraging data collected on a STEM. The data samples are\nprocessed and filtered to produce symbolic representations, which we use to\ntrain a neural network to predict transition probabilities. These learned\ntransition dynamics are then leveraged to guide a single silicon atom\nthroughout the lattice to pre-determined target destinations. We present\nempirical analyses that demonstrate the efficacy and generality of our\napproach.\n", "date": "2023-11-21", "categories": []}
{"id": "2311.16158", "title": "CarbNN: A Novel Active Transfer Learning Neural Network To Build De Novo\n  Metal Organic Frameworks (MOFs) for Carbon Capture", "abstract": "  Over the past decade, climate change has become an increasing problem with\none of the major contributing factors being carbon dioxide (CO2) emissions;\nalmost 51% of total US carbon emissions are from factories. Current materials\nused in CO2 capture are lacking either in efficiency, sustainability, or cost.\n  Electrocatalysis of CO2 is a new approach where CO2 can be reduced and the\ncomponents used industrially as fuel, saving transportation costs, creating\nfinancial incentives. Metal Organic Frameworks (MOFs) are crystals made of\norgano-metals that adsorb, filter, and electrocatalyze CO2. The current\navailable MOFs for capture & electrocatalysis are expensive to manufacture and\ninefficient at capture. The goal therefore is to computationally design a MOF\nthat can adsorb CO2 and catalyze carbon monoxide & oxygen with low cost.\n  A novel active transfer learning neural network was developed, utilizing\ntransfer learning due to limited available data on 15 MOFs. Using the Cambridge\nStructural Database with 10,000 MOFs, the model used incremental mutations to\nfit a trained fitness hyper-heuristic function. Eventually, a Selenium MOF\n(C18MgO25Se11Sn20Zn5) was converged on. Through analysis of predictions &\nliterature, the converged MOF was shown to be more effective & more\nsynthetically accessible than existing MOFs, showing the model had an\nunderstanding of effective electrocatalytic structures in the material space.\nThis novel network can be implemented for other gas separations and catalysis\napplications that have limited training accessible datasets.\n", "date": "2023-11-09", "categories": []}
{"id": "2311.16727", "title": "Sluggish and Chemically-Biased Interstitial Diffusion in Concentrated\n  Solid Solution Alloys: Mechanisms and Methods", "abstract": "  Interstitial diffusion is a pivotal process that governs the phase stability\nand irradiation response of materials in non-equilibrium conditions. In this\nwork, we study sluggish and chemically-biased interstitial diffusion in Fe-Ni\nconcentrated solid solution alloys (CSAs) by combining machine learning (ML)\nand kinetic Monte Carlo (kMC), where ML is used to accurately and efficiently\npredict the migration energy barriers on-the-fly. The ML-kMC reproduces the\ndiffusivity that was reported by molecular dynamics results at high\ntemperatures. With this powerful tool, we find that the observed sluggish\ndiffusion and the \"Ni-Ni-Ni\"-biased diffusion in Fe-Ni alloys are ascribed to a\nunique \"Barrier Lock\" mechanism, whereas the \"Fe-Fe-Fe\"-biased diffusion is\ninfluenced by a \"Component Dominance\" mechanism. Inspired by the mentioned\nmechanisms, a practical AvgS-kMC method is proposed for conveniently and\nswiftly determining interstitial-mediated diffusivity by only relying on the\nmean energy barriers of migration patterns. Combining the AvgS-kMC with the\ndifferential evolutionary algorithm, an inverse design strategy for optimizing\nsluggish diffusion properties is applied to emphasize the crucial role of\nfavorable migration patterns.\n", "date": "2023-11-28", "categories": []}
{"id": "2311.16964", "title": "Machine learning force-field models for metallic spin glass", "abstract": "  Metallic spin glass systems, such as dilute magnetic alloys, are\ncharacterized by randomly distributed local moments coupled to each other\nthrough a long-range electron-mediated effective interaction. We present a\nscalable machine learning (ML) framework for dynamical simulations of metallic\nspin glasses. A Behler-Parrinello type neural-network model, based on the\nprinciple of locality, is developed to accurately and efficiently predict\nelectron-induced local magnetic fields that drive the spin dynamics. A crucial\ncomponent of the ML model is a proper symmetry-invariant representation of\nlocal magnetic environment which is direct input to the neural net. We develop\nsuch a magnetic descriptor by incorporating the spin degrees of freedom into\nthe atom-centered symmetry function methods which are widely used in ML\nforce-field models for quantum molecular dynamics. We apply our approach to\nstudy the relaxation dynamics of an amorphous generalization of the s-d model.\nOur work highlights the promising potential of ML models for large-scale\ndynamical modeling of itinerant magnets with quenched disorder.\n", "date": "2023-11-28", "categories": []}
{"id": "2303.16412", "title": "A Comprehensive and Versatile Multimodal Deep Learning Approach for\n  Predicting Diverse Properties of Advanced Materials", "abstract": "  We present a multimodal deep learning (MDL) framework for predicting physical\nproperties of a 10-dimensional acrylic polymer composite material by merging\nphysical attributes and chemical data. Our MDL model comprises four modules,\nincluding three generative deep learning models for material structure\ncharacterization and a fourth model for property prediction. Our approach\nhandles an 18-dimensional complexity, with 10 compositional inputs and 8\nproperty outputs, successfully predicting 913,680 property data points across\n114,210 composition conditions. This level of complexity is unprecedented in\ncomputational materials science, particularly for materials with undefined\nstructures. We propose a framework to analyze the high-dimensional information\nspace for inverse material design, demonstrating flexibility and adaptability\nto various materials and scales, provided sufficient data is available. This\nstudy advances future research on different materials and the development of\nmore sophisticated models, drawing us closer to the ultimate goal of predicting\nall properties of all materials.\n", "date": "2023-03-29", "categories": []}
{"id": "2311.00341", "title": "The Open DAC 2023 Dataset and Challenges for Sorbent Discovery in Direct\n  Air Capture", "abstract": "  New methods for carbon dioxide removal are urgently needed to combat global\nclimate change. Direct air capture (DAC) is an emerging technology to capture\ncarbon dioxide directly from ambient air. Metal-organic frameworks (MOFs) have\nbeen widely studied as potentially customizable adsorbents for DAC. However,\ndiscovering promising MOF sorbents for DAC is challenging because of the vast\nchemical space to explore and the need to understand materials as functions of\nhumidity and temperature. We explore a computational approach benefiting from\nrecent innovations in machine learning (ML) and present a dataset named Open\nDAC 2023 (ODAC23) consisting of more than 38M density functional theory (DFT)\ncalculations on more than 8,400 MOF materials containing adsorbed $CO_2$ and/or\n$H_2O$. ODAC23 is by far the largest dataset of MOF adsorption calculations at\nthe DFT level of accuracy currently available. In addition to probing\nproperties of adsorbed molecules, the dataset is a rich source of information\non structural relaxation of MOFs, which will be useful in many contexts beyond\nspecific applications for DAC. A large number of MOFs with promising properties\nfor DAC are identified directly in ODAC23. We also trained state-of-the-art ML\nmodels on this dataset to approximate calculations at the DFT level. This\nopen-source dataset and our initial ML models will provide an important\nbaseline for future efforts to identify MOFs for a wide range of applications,\nincluding DAC.\n", "date": "2023-11-01", "categories": []}
{"id": "2311.14744", "title": "Coarse-Grained Configurational Polymer Fingerprints for Property\n  Prediction using Machine Learning", "abstract": "  In this work, we present a method to generate a configurational level\nfingerprint for polymers using the Bead-Spring-Model. Unlike some of the\nprevious fingerprinting approaches that employ monomer-level information where\natomistic descriptors are computed using quantum chemistry calculations, this\napproach incorporates configurational information from a coarse-grained model\nof a long polymer chain. The proposed approach may be advantageous for the\nstudy of behavior resulting from large molecular weights. To create this\nfingerprint, we make use of two kinds of descriptors. First, we calculate\ncertain geometric descriptors like Re2, Rg2 etc. and label them as Calculated\nDescriptors. Second, we generate a set of data-driven descriptors using an\nunsupervised autoencoder model and call them Learnt Descriptors. Using a\ncombination of both of them, we are able to learn mappings from the structure\nto various properties of the polymer chain by training ML models. We test our\nfingerprint to predict the probability of occurrence of a configuration at\nequilibrium, which is approximated by a simple linear relationship between the\ninstantaneous internal energy and equilibrium average internal energy.\n", "date": "2023-11-20", "categories": []}
{"id": "2311.12856", "title": "Density of States Prediction of Crystalline Materials via Prompt-guided\n  Multi-Modal Transformer", "abstract": "  The density of states (DOS) is a spectral property of crystalline materials,\nwhich provides fundamental insights into various characteristics of the\nmaterials. While previous works mainly focus on obtaining high-quality\nrepresentations of crystalline materials for DOS prediction, we focus on\npredicting the DOS from the obtained representations by reflecting the nature\nof DOS: DOS determines the general distribution of states as a function of\nenergy. That is, DOS is not solely determined by the crystalline material but\nalso by the energy levels, which has been neglected in previous works. In this\npaper, we propose to integrate heterogeneous information obtained from the\ncrystalline materials and the energies via a multi-modal transformer, thereby\nmodeling the complex relationships between the atoms in the crystalline\nmaterials and various energy levels for DOS prediction. Moreover, we propose to\nutilize prompts to guide the model to learn the crystal structural\nsystem-specific interactions between crystalline materials and energies.\nExtensive experiments on two types of DOS, i.e., Phonon DOS and Electron DOS,\nwith various real-world scenarios demonstrate the superiority of\nDOSTransformer. The source code for DOSTransformer is available at\nhttps://github.com/HeewoongNoh/DOSTransformer.\n", "date": "2023-10-24", "categories": []}
{"id": "2311.12272", "title": "Procedural Generation of Grain Orientations using the Wave Function\n  Collapse Algorithm", "abstract": "  Statistics of grain sizes and orientations in metals correlate to the\nmaterial's mechanical properties. Reproducing representative volume elements\nfor further analysis of deformation and failure in metals, like 316L stainless\nsteel, is particularly important due to their wide use in manufacturing goods\ntoday. Two approaches, initially created for video games, were considered for\nthe procedural generation of representative grain microstructures. The first is\nthe Wave Function Collapse (WFC) algorithm, and the second is constraint\npropagation and probabilistic inference through Markov Junior, a free and\nopen-source software. This study aimed to investigate these two algorithms'\neffectiveness in using reference electron backscatter diffraction (EBSD) maps\nand recreating a statistically similar one that could be used in further\nresearch. It utilized two stainless steel EBSD maps as references to test both\nalgorithms. First, the WFC algorithm was too constricting and, thus, incapable\nof producing images that resembled EBSDs. The second, MarkovJunior, was much\nmore effective in creating a Voronoi tessellation that could be used to create\nan EBSD map in Python. When comparing the results between the reference and the\ngenerated EBSD, we discovered that the orientation and volume fractions were\nextremely similar. With the study, it was concluded that MarkovJunior is an\neffective machine learning tool that can reproduce representative grain\nmicrostructures.\n", "date": "2023-11-21", "categories": []}
{"id": "2311.11060", "title": "AIMS-EREA -- A framework for AI-accelerated Innovation of Materials for\n  Sustainability -- for Environmental Remediation and Energy Applications", "abstract": "  Many environmental remediation and energy applications (conversion and\nstorage) for sustainability need design and development of green novel\nmaterials. Discovery processes of such novel materials are time taking and\ncumbersome due to large number of possible combinations and permutations of\nmaterials structures. Often theoretical studies based on Density Functional\nTheory (DFT) and other theories, coupled with Simulations are conducted to\nnarrow down sample space of candidate materials, before conducting\nlaboratory-based synthesis and analytical process. With the emergence of\nartificial intelligence (AI), AI techniques are being tried in this process too\nto ease out simulation time and cost. However tremendous values of previously\npublished research from various parts of the world are still left as\nlabor-intensive manual effort and discretion of individual researcher and prone\nto human omissions. AIMS-EREA is our novel framework to blend best of breed of\nMaterial Science theory with power of Generative AI to give best impact and\nsmooth and quickest discovery of material for sustainability. This also helps\nto eliminate the possibility of production of hazardous residues and\nbye-products of the reactions. AIMS-EREA uses all available resources --\nPredictive and Analytical AI on large collection of chemical databases along\nwith automated intelligent assimilation of deep materials knowledge from\npreviously published research works through Generative AI. We demonstrate use\nof our own novel framework with an example, how this framework can be\nsuccessfully applied to achieve desired success in development of\nthermoelectric material for waste heat conversion.\n", "date": "2023-11-18", "categories": []}
{"id": "2306.17587", "title": "Variational principle to regularize machine-learned density functionals:\n  the non-interacting kinetic-energy functional", "abstract": "  Practical density functional theory (DFT) owes its success to the\ngroundbreaking work of Kohn and Sham that introduced the exact calculation of\nthe non-interacting kinetic energy of the electrons using an auxiliary\nmean-field system. However, the full power of DFT will not be unleashed until\nthe exact relationship between the electron density and the non-interacting\nkinetic energy is found. Various attempts have been made to approximate this\nfunctional, similar to the exchange--correlation functional, with much less\nsuccess due to the larger contribution of kinetic energy and its more non-local\nnature. In this work we propose a new and efficient regularization method to\ntrain density functionals based on deep neural networks, with particular\ninterest in the kinetic-energy functional. The method is tested on\n(effectively) one-dimensional systems, including the hydrogen chain,\nnon-interacting electrons, and atoms of the first two periods, with excellent\nresults. For the atomic systems, the generalizability of the regularization\nmethod is demonstrated by training also an exchange--correlation functional,\nand the contrasting nature of the two functionals is discussed from a\nmachine-learning perspective.\n", "date": "2023-06-30", "categories": []}
{"id": "2311.09591", "title": "Accelerating material discovery with a threshold-driven hybrid\n  acquisition policy-based Bayesian optimization", "abstract": "  Advancements in materials play a crucial role in technological progress.\nHowever, the process of discovering and developing materials with desired\nproperties is often impeded by substantial experimental costs, extensive\nresource utilization, and lengthy development periods. To address these\nchallenges, modern approaches often employ machine learning (ML) techniques\nsuch as Bayesian Optimization (BO), which streamline the search for optimal\nmaterials by iteratively selecting experiments that are most likely to yield\nbeneficial results. However, traditional BO methods, while beneficial, often\nstruggle with balancing the trade-off between exploration and exploitation,\nleading to sub-optimal performance in material discovery processes. This paper\nintroduces a novel Threshold-Driven UCB-EI Bayesian Optimization (TDUE-BO)\nmethod, which dynamically integrates the strengths of Upper Confidence Bound\n(UCB) and Expected Improvement (EI) acquisition functions to optimize the\nmaterial discovery process. Unlike the classical BO, our method focuses on\nefficiently navigating the high-dimensional material design space (MDS).\nTDUE-BO begins with an exploration-focused UCB approach, ensuring a\ncomprehensive initial sweep of the MDS. As the model gains confidence,\nindicated by reduced uncertainty, it transitions to the more exploitative EI\nmethod, focusing on promising areas identified earlier. The UCB-to-EI switching\npolicy dictated guided through continuous monitoring of the model uncertainty\nduring each step of sequential sampling results in navigating through the MDS\nmore efficiently while ensuring rapid convergence. The effectiveness of TDUE-BO\nis demonstrated through its application on three different material datasets,\nshowing significantly better approximation and optimization performance over\nthe EI and UCB-based BO methods in terms of the RMSE scores and convergence\nefficiency, respectively.\n", "date": "2023-11-16", "categories": []}
{"id": "2302.09184", "title": "Rapid Design of Top-Performing Metal-Organic Frameworks with Qualitative\n  Representations of Building Blocks", "abstract": "  Data-driven materials design often encounters challenges where systems\nrequire or possess qualitative (categorical) information. Metal-organic\nframeworks (MOFs) are an example of such material systems. The representation\nof MOFs through different building blocks makes it a challenge for designers to\nincorporate qualitative information into design optimization. Furthermore, the\nlarge number of potential building blocks leads to a combinatorial challenge,\nwith millions of possible MOFs that could be explored through time consuming\nphysics-based approaches. In this work, we integrated Latent Variable Gaussian\nProcess (LVGP) and Multi-Objective Batch-Bayesian Optimization (MOBBO) to\nidentify top-performing MOFs adaptively, autonomously, and efficiently without\nany human intervention. Our approach provides three main advantages: (i) no\nspecific physical descriptors are required and only building blocks that\nconstruct the MOFs are used in global optimization through qualitative\nrepresentations, (ii) the method is application and property independent, and\n(iii) the latent variable approach provides an interpretable model of\nqualitative building blocks with physical justification. To demonstrate the\neffectiveness of our method, we considered a design space with more than 47,000\nMOF candidates. By searching only ~1% of the design space, LVGP-MOBBO was able\nto identify all MOFs on the Pareto front and more than 97% of the 50\ntop-performing designs for the CO$_2$ working capacity and CO$_2$/N$_2$\nselectivity properties. Finally, we compared our approach with the Random\nForest algorithm and demonstrated its efficiency, interpretability, and\nrobustness.\n", "date": "2023-02-17", "categories": []}
{"id": "2307.03811", "title": "Formulation Graphs for Mapping Structure-Composition of Battery\n  Electrolytes to Device Performance", "abstract": "  Advanced computational methods are being actively sought for addressing the\nchallenges associated with discovery and development of new combinatorial\nmaterial such as formulations. A widely adopted approach involves domain\ninformed high-throughput screening of individual components that can be\ncombined into a formulation. This manages to accelerate the discovery of new\ncompounds for a target application but still leave the process of identifying\nthe right 'formulation' from the shortlisted chemical space largely a\nlaboratory experiment-driven process. We report a deep learning model,\nFormulation Graph Convolution Network (F-GCN), that can map\nstructure-composition relationship of the individual components to the property\nof liquid formulation as whole. Multiple GCNs are assembled in parallel that\nfeaturize formulation constituents domain-intuitively on the fly. The resulting\nmolecular descriptors are scaled based on respective constituent's molar\npercentage in the formulation, followed by formalizing into a combined\ndescriptor that represents a complete formulation to an external learning\narchitecture. The use case of proposed formulation learning model is\ndemonstrated for battery electrolytes by training and testing it on two\nexemplary datasets representing electrolyte formulations vs battery performance\n-- one dataset is sourced from literature about Li/Cu half-cells, while the\nother is obtained by lab-experiments related to lithium-iodide full-cell\nchemistry. The model is shown to predict the performance metrics like Coulombic\nEfficiency (CE) and specific capacity of new electrolyte formulations with\nlowest reported errors. The best performing F-GCN model uses molecular\ndescriptors derived from molecular graphs that are informed with HOMO-LUMO and\nelectric moment properties of the molecules using a knowledge transfer\ntechnique.\n", "date": "2023-07-07", "categories": []}
{"id": "2309.00305", "title": "Efficient Surrogate Models for Materials Science Simulations: Machine\n  Learning-based Prediction of Microstructure Properties", "abstract": "  Determining, understanding, and predicting the so-called structure-property\nrelation is an important task in many scientific disciplines, such as\nchemistry, biology, meteorology, physics, engineering, and materials science.\nStructure refers to the spatial distribution of, e.g., substances, material, or\nmatter in general, while property is a resulting characteristic that usually\ndepends in a non-trivial way on spatial details of the structure.\nTraditionally, forward simulations models have been used for such tasks.\nRecently, several machine learning algorithms have been applied in these\nscientific fields to enhance and accelerate simulation models or as surrogate\nmodels. In this work, we develop and investigate the applications of six\nmachine learning techniques based on two different datasets from the domain of\nmaterials science: data from a two-dimensional Ising model for predicting the\nformation of magnetic domains and data representing the evolution of dual-phase\nmicrostructures from the Cahn-Hilliard model. We analyze the accuracy and\nrobustness of all models and elucidate the reasons for the differences in their\nperformances. The impact of including domain knowledge through tailored\nfeatures is studied, and general recommendations based on the availability and\nquality of training data are derived from this.\n", "date": "2023-09-01", "categories": []}
{"id": "2311.07174", "title": "The High-dimensional Phase Diagram and the Large CALPHAD Model", "abstract": "  When alloy systems comprise more than three elements, the visualization of\nthe entire phase space becomes not only daunting but is also accompanied by a\ndata surge. Addressing this complexity, we delve into the FeNiCrMn alloy system\nand introduce the Large CALPHAD Model (LCM). The LCM acts as a computational\nconduit, capturing the entire phase space. Subsequently, this enormous data is\nsystematically structured using a high-dimensional phase diagram, aided by hash\ntables and Depth-first Search (DFS), rendering it both digestible and\nprogrammatically accessible. Remarkably, the LCM boasts a 97% classification\naccuracy and a mean square error of 4.80*10-5 in phase volume prediction. Our\nmethodology successfully delineates 51 unique phase spaces in the FeNiCrMn\nsystem, exemplifying its efficacy with the design of all 439 eutectic alloys.\nThis pioneering methodology signifies a monumental shift in alloy design\ntechniques or even multi-variable problems.\n", "date": "2023-11-13", "categories": []}
{"id": "2306.12818", "title": "StrainTensorNet: Predicting crystal structure elastic properties using\n  SE(3)-equivariant graph neural networks", "abstract": "  Accurately predicting the elastic properties of crystalline solids is vital\nfor computational materials science. However, traditional atomistic scale ab\ninitio approaches are computationally intensive, especially for studying\ncomplex materials with a large number of atoms in a unit cell. We introduce a\nnovel data-driven approach to efficiently predict the elastic properties of\ncrystal structures using SE(3)-equivariant graph neural networks (GNNs). This\napproach yields important scalar elastic moduli with the accuracy comparable to\nrecent data-driven studies. Importantly, our symmetry-aware GNNs model also\nenables the prediction of the strain energy density (SED) and the associated\nelastic constants, the fundamental tensorial quantities that are significantly\ninfluenced by a material's crystallographic group. The model consistently\ndistinguishes independent elements of SED tensors, in accordance with the\nsymmetry of the crystal structures. Finally, our deep learning model possesses\nmeaningful latent features, offering an interpretable prediction of the elastic\nproperties.\n", "date": "2023-06-22", "categories": []}
{"id": "2311.05407", "title": "Data Distillation for Neural Network Potentials toward Foundational\n  Dataset", "abstract": "  Machine learning (ML) techniques and atomistic modeling have rapidly\ntransformed materials design and discovery. Specifically, generative models can\nswiftly propose promising materials for targeted applications. However, the\npredicted properties of materials through the generative models often do not\nmatch with calculated properties through ab initio calculations. This\ndiscrepancy can arise because the generated coordinates are not fully relaxed,\nwhereas the many properties are derived from relaxed structures. Neural\nnetwork-based potentials (NNPs) can expedite the process by providing relaxed\nstructures from the initially generated ones. Nevertheless, acquiring data to\ntrain NNPs for this purpose can be extremely challenging as it needs to\nencompass previously unknown structures. This study utilized extended ensemble\nmolecular dynamics (MD) to secure a broad range of liquid- and solid-phase\nconfigurations in one of the metallic systems, nickel. Then, we could\nsignificantly reduce them through active learning without losing much accuracy.\nWe found that the NNP trained from the distilled data could predict different\nenergy-minimized closed-pack crystal structures even though those structures\nwere not explicitly part of the initial data. Furthermore, the data can be\ntranslated to other metallic systems (aluminum and niobium), without repeating\nthe sampling and distillation processes. Our approach to data acquisition and\ndistillation has demonstrated the potential to expedite NNP development and\nenhance materials design and discovery by integrating generative models.\n", "date": "2023-11-09", "categories": []}
{"id": "2306.14568", "title": "Elucidating Interfacial Dynamics of Ti-Al Systems Using Molecular\n  Dynamics Simulation and Markov State Modeling", "abstract": "  Due to their remarkable mechanical and chemical properties, Ti-Al based\nmaterials are attracting considerable interest in numerous fields of\nengineering, such as automotive, aerospace, and defense. With their low\ndensity, high strength, and resistance to corrosion and oxidation, these\nintermetallic alloys and compound metal-metallic composites have found diverse\napplications. The present study delves into the interfacial dynamics of these\nTi-Al systems, particularly focusing on the behavior of Ti and Al atoms in the\npresence of TiAl$_3$ grain boundaries under experimental heat treatment\nconditions. Using a combination of Molecular Dynamics and Markov State Model\nanalyses, we scrutinize the kinetic processes involved in the formation of\nTiAl$_3$. The Molecular Dynamics simulation indicates that at the early stage\nof heat treatment, the predominating process is the diffusion of Al atoms\ntowards the Ti surface through the TiAl$_3$ grain boundaries. The Markov State\nModeling identifies three distinct dynamic states of Al atoms within the Ti/Al\nmixture that forms during the process, each exhibiting a unique spatial\ndistribution. Using transition timescales as a qualitative measure of the\nrapidness of the dynamics, it is observed that the Al dynamics is significantly\nless rapid near the Ti surface compared to the Al surface. Put together, the\nresults offer a comprehensive understanding of the interfacial dynamics and\nreveals a three-stage diffusion mechanism. The process initiates with the\npremelting of Al, proceeds with the prevalent diffusion of Al atoms towards the\nTi surface, and eventually ceases as the Ti concentration within the mixture\nprogressively increases. The insights gained from this study could contribute\nsignificantly to the control and optimization of manufacturing processes for\nthese high-performing Ti-Al based materials.\n", "date": "2023-06-26", "categories": []}
{"id": "2306.10045", "title": "Efficient Approximations of Complete Interatomic Potentials for Crystal\n  Property Prediction", "abstract": "  We study property prediction for crystal materials. A crystal structure\nconsists of a minimal unit cell that is repeated infinitely in 3D space. How to\naccurately represent such repetitive structures in machine learning models\nremains unresolved. Current methods construct graphs by establishing edges only\nbetween nearby nodes, thereby failing to faithfully capture infinite repeating\npatterns and distant interatomic interactions. In this work, we propose several\ninnovations to overcome these limitations. First, we propose to model\nphysics-principled interatomic potentials directly instead of only using\ndistances as in many existing methods. These potentials include the Coulomb\npotential, London dispersion potential, and Pauli repulsion potential. Second,\nwe model the complete set of potentials among all atoms, instead of only\nbetween nearby atoms as in existing methods. This is enabled by our\napproximations of infinite potential summations, where we extend the Ewald\nsummation for several potential series approximations with provable error\nbounds. Finally, we propose to incorporate our computations of complete\ninteratomic potentials into message passing neural networks for representation\nlearning. We perform experiments on the JARVIS and Materials Project benchmarks\nfor evaluation. Results show that the use of interatomic potentials and\ncomplete interatomic potentials leads to consistent performance improvements\nwith reasonable computational costs. Our code is publicly available as part of\nthe AIRS library (https://github.com/divelab/AIRS/tree/main/OpenMat/PotNet).\n", "date": "2023-06-12", "categories": []}
{"id": "2307.02707", "title": "Towards Symmetry-Aware Generation of Periodic Materials", "abstract": "  We consider the problem of generating periodic materials with deep models.\nWhile symmetry-aware molecule generation has been studied extensively, periodic\nmaterials possess different symmetries, which have not been completely captured\nby existing methods. In this work, we propose SyMat, a novel material\ngeneration approach that can capture physical symmetries of periodic material\nstructures. SyMat generates atom types and lattices of materials through\ngenerating atom type sets, lattice lengths and lattice angles with a\nvariational auto-encoder model. In addition, SyMat employs a score-based\ndiffusion model to generate atom coordinates of materials, in which a novel\nsymmetry-aware probabilistic model is used in the coordinate diffusion process.\nWe show that SyMat is theoretically invariant to all symmetry transformations\non materials and demonstrate that SyMat achieves promising performance on\nrandom generation and property optimization tasks. Our code is publicly\navailable as part of the AIRS library (https://github.com/divelab/AIRS).\n", "date": "2023-07-06", "categories": []}
{"id": "2311.03139", "title": "End-to-end Material Thermal Conductivity Prediction through Machine\n  Learning", "abstract": "  We investigated the accelerated prediction of the thermal conductivity of\nmaterials through end- to-end structure-based approaches employing machine\nlearning methods. Due to the non-availability of high-quality thermal\nconductivity data, we first performed high-throughput calculations based on\nfirst principles and the Boltzmann transport equation for 225 materials,\neffectively more than doubling the size of the existing dataset. We assessed\nthe performance of state-of-the-art machine learning models for thermal\nconductivity prediction on this expanded dataset and observed that all these\nmodels suffered from overfitting. To address this issue, we introduced a novel\ngraph-based neural network model, which demonstrated more consistent and\nregularized performance across all evaluated datasets. Nevertheless, the best\nmean absolute percentage error achieved on the test dataset remained in the\nrange of 50-60%. This suggests that while these models are valuable for\nexpediting material screening, their current accuracy is still limited.\n", "date": "2023-11-06", "categories": []}
{"id": "2310.18505", "title": "Multi-fidelity Design of Porous Microstructures for Thermofluidic\n  Applications", "abstract": "  As modern electronic devices are increasingly miniaturized and integrated,\ntheir performance relies more heavily on effective thermal management.\nTwo-phase cooling methods enhanced by porous surfaces, which capitalize on\nthin-film evaporation atop structured porous surfaces, are emerging as\npotential solutions. In such porous structures, the optimum heat dissipation\ncapacity relies on two competing objectives that depend on mass and heat\ntransfer. The computational costs of evaluating these objectives, the high\ndimensionality of the design space which a voxelated microstructure\nrepresentation, and the manufacturability constraints hinder the optimization\nprocess for thermal management. We address these challenges by developing a\ndata-driven framework for designing optimal porous microstructures for cooling\napplications. In our framework we leverage spectral density functions (SDFs) to\nencode the design space via a handful of interpretable variables and, in turn,\nefficiently search it. We develop physics-based formulas to quantify the\nthermofluidic properties and feasibility of candidate designs via offline\nsimulations. To decrease the reliance on expensive simulations, we generate\nmulti-fidelity data and build emulators to find Pareto-optimal designs. We\napply our approach to a canonical problem on evaporator wick design and obtain\nfin-like topologies in the optimal microstructures which are also\ncharacteristics often observed in industrial applications.\n", "date": "2023-10-27", "categories": []}
{"id": "2310.18523", "title": "Using convolutional neural networks for stereological characterization\n  of 3D hetero-aggregates based on synthetic STEM data", "abstract": "  The structural characterization of hetero-aggregates in 3D is of great\ninterest, e.g., for deriving process-structure or structure-property\nrelationships. However, since 3D imaging techniques are often difficult to\nperform as well as time and cost intensive, a characterization of\nhetero-aggregates based on 2D image data is desirable, but often non-trivial.\nTo overcome the issues of characterizing 3D structures from 2D measurements, a\nmethod is presented that relies on machine learning combined with methods of\nspatial stochastic modeling, where the latter are utilized for the generation\nof synthetic training data. This kind of training data has the advantage that\ntime-consuming experiments for the synthesis of differently structured\nmaterials followed by their 3D imaging can be avoided. More precisely, a\nparametric stochastic 3D model is presented, from which a wide spectrum of\nvirtual hetero-aggregates can be generated. Additionally, the virtual\nstructures are passed to a physics-based simulation tool in order to generate\nvirtual scanning transmission electron microscopy (STEM) images. The preset\nparameters of the 3D model together with the simulated STEM images serve as a\ndatabase for the training of convolutional neural networks, which can be used\nto determine the parameters of the underlying 3D model and, consequently, to\npredict 3D structures of hetero-aggregates from 2D STEM images. Furthermore, an\nerror analysis is performed to evaluate the prediction power of the trained\nneural networks with respect to structural descriptors, e.g. the\nhetero-coordination number.\n", "date": "2023-10-27", "categories": []}
{"id": "2310.18552", "title": "The Role of Reference Points in Machine-Learned Atomistic Simulation\n  Models", "abstract": "  This paper introduces the Chemical Environment Modeling Theory (CEMT), a\nnovel, generalized framework designed to overcome the limitations inherent in\ntraditional atom-centered Machine Learning Force Field (MLFF) models, widely\nused in atomistic simulations of chemical systems. CEMT demonstrated enhanced\nflexibility and adaptability by allowing reference points to exist anywhere\nwithin the modeled domain and thus, enabling the study of various model\narchitectures. Utilizing Gaussian Multipole (GMP) featurization functions,\nseveral models with different reference point sets, including finite difference\ngrid-centered and bond-centered models, were tested to analyze the variance in\ncapabilities intrinsic to models built on distinct reference points. The\nresults underscore the potential of non-atom-centered reference points in force\ntraining, revealing variations in prediction accuracy, inference speed and\nlearning efficiency. Finally, a unique connection between CEMT and real-space\norbital-free finite element Density Functional Theory (FE-DFT) is established,\nand the implications include the enhancement of data efficiency and robustness.\nIt allows the leveraging of spatially-resolved energy densities and charge\ndensities from FE-DFT calculations, as well as serving as a pivotal step\ntowards integrating known quantum-mechanical laws into the architecture of ML\nmodels.\n", "date": "2023-10-28", "categories": []}
{"id": "2310.18702", "title": "Towards Combinatorial Generalization for Catalysts: A Kohn-Sham\n  Charge-Density Approach", "abstract": "  The Kohn-Sham equations underlie many important applications such as the\ndiscovery of new catalysts. Recent machine learning work on catalyst modeling\nhas focused on prediction of the energy, but has so far not yet demonstrated\nsignificant out-of-distribution generalization. Here we investigate another\napproach based on the pointwise learning of the Kohn-Sham charge-density. On a\nnew dataset of bulk catalysts with charge densities, we show density models can\ngeneralize to new structures with combinations of elements not seen at train\ntime, a form of combinatorial generalization. We show that over 80% of binary\nand ternary test cases achieve faster convergence than standard baselines in\nDensity Functional Theory, amounting to an average reduction of 13% in the\nnumber of iterations required to reach convergence, which may be of independent\ninterest. Our results suggest that density learning is a viable alternative,\ntrading greater inference costs for a step towards combinatorial\ngeneralization, a key property for applications.\n", "date": "2023-10-28", "categories": []}
{"id": "2310.18907", "title": "Topological, or Non-topological? A Deep Learning Based Prediction", "abstract": "  Prediction and discovery of new materials with desired properties are at the\nforefront of quantum science and technology research. A major bottleneck in\nthis field is the computational resources and time complexity related to\nfinding new materials from ab initio calculations. In this work, an effective\nand robust deep learning-based model is proposed by incorporating persistent\nhomology and graph neural network which offers an accuracy of 91.4% and an F1\nscore of 88.5% in classifying topological vs. non-topological materials,\noutperforming the other state-of-the-art classifier models. The incorporation\nof the graph neural network encodes the underlying relation between the atoms\ninto the model based on their own crystalline structures and thus proved to be\nan effective method to represent and process non-euclidean data like molecules\nwith a relatively shallow network. The persistent homology pipeline in the\nsuggested neural network is capable of integrating the atom-specific\ntopological information into the deep learning model, increasing robustness,\nand gain in performance. It is believed that the presented work will be an\nefficacious tool for predicting the topological class and therefore enable the\nhigh-throughput search for novel materials in this field.\n", "date": "2023-10-29", "categories": []}
{"id": "2310.16958", "title": "Transferring a molecular foundation model for polymer property\n  predictions", "abstract": "  Transformer-based large language models have remarkable potential to\naccelerate design optimization for applications such as drug development and\nmaterials discovery. Self-supervised pretraining of transformer models requires\nlarge-scale datasets, which are often sparsely populated in topical areas such\nas polymer science. State-of-the-art approaches for polymers conduct data\naugmentation to generate additional samples but unavoidably incurs extra\ncomputational costs. In contrast, large-scale open-source datasets are\navailable for small molecules and provide a potential solution to data scarcity\nthrough transfer learning. In this work, we show that using transformers\npretrained on small molecules and fine-tuned on polymer properties achieve\ncomparable accuracy to those trained on augmented polymer datasets for a series\nof benchmark prediction tasks.\n", "date": "2023-10-25", "categories": []}
{"id": "2310.15188", "title": "Deep Learning Approaches for Dynamic Mechanical Analysis of Viscoelastic\n  Fiber Composites", "abstract": "  The increased adoption of reinforced polymer (RP) composite materials, driven\nby eco-design standards, calls for a fine balance between lightness, stiffness,\nand effective vibration control. These materials are integral to enhancing\ncomfort, safety, and energy efficiency. Dynamic Mechanical Analysis (DMA)\ncharacterizes viscoelastic behavior, yet there's a growing interest in using\nMachine Learning (ML) to expedite the design and understanding of\nmicrostructures. In this paper we aim to map microstructures to their\nmechanical properties using deep neural networks, speeding up the process and\nallowing for the generation of microstructures from desired properties.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.13756", "title": "Learning Interatomic Potentials at Multiple Scales", "abstract": "  The need to use a short time step is a key limit on the speed of molecular\ndynamics (MD) simulations. Simulations governed by classical potentials are\noften accelerated by using a multiple-time-step (MTS) integrator that evaluates\ncertain potential energy terms that vary more slowly than others less\nfrequently. This approach is enabled by the simple but limiting analytic forms\nof classical potentials. Machine learning interatomic potentials (MLIPs), in\nparticular recent equivariant neural networks, are much more broadly applicable\nthan classical potentials and can faithfully reproduce the expensive but\naccurate reference electronic structure calculations used to train them. They\nstill, however, require the use of a single short time step, as they lack the\ninherent term-by-term scale separation of classical potentials. This work\nintroduces a method to learn a scale separation in complex interatomic\ninteractions by co-training two MLIPs. Initially, a small and efficient model\nis trained to reproduce short-time-scale interactions. Subsequently, a large\nand expressive model is trained jointly to capture the remaining interactions\nnot captured by the small model. When running MD, the MTS integrator then\nevaluates the smaller model for every time step and the larger model less\nfrequently, accelerating simulation. Compared to a conventionally trained MLIP,\nour approach can achieve a significant speedup (~3x in our experiments) without\na loss of accuracy on the potential energy or simulation-derived quantities.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.14029", "title": "LLM-Prop: Predicting Physical And Electronic Properties Of Crystalline\n  Solids From Their Text Descriptions", "abstract": "  The prediction of crystal properties plays a crucial role in the crystal\ndesign process. Current methods for predicting crystal properties focus on\nmodeling crystal structures using graph neural networks (GNNs). Although GNNs\nare powerful, accurately modeling the complex interactions between atoms and\nmolecules within a crystal remains a challenge. Surprisingly, predicting\ncrystal properties from crystal text descriptions is understudied, despite the\nrich information and expressiveness that text data offer. One of the main\nreasons is the lack of publicly available data for this task. In this paper, we\ndevelop and make public a benchmark dataset (called TextEdge) that contains\ntext descriptions of crystal structures with their properties. We then propose\nLLM-Prop, a method that leverages the general-purpose learning capabilities of\nlarge language models (LLMs) to predict the physical and electronic properties\nof crystals from their text descriptions. LLM-Prop outperforms the current\nstate-of-the-art GNN-based crystal property predictor by about 4% in predicting\nband gap, 3% in classifying whether the band gap is direct or indirect, and 66%\nin predicting unit cell volume. LLM-Prop also outperforms a finetuned MatBERT,\na domain-specific pre-trained BERT model, despite having 3 times fewer\nparameters. Our empirical results may highlight the current inability of GNNs\nto capture information pertaining to space group symmetry and Wyckoff sites for\naccurate crystal property prediction.\n", "date": "2023-10-21", "categories": []}
{"id": "2310.13136", "title": "Approaches for Uncertainty Quantification of AI-predicted Material\n  Properties: A Comparison", "abstract": "  The development of large databases of material properties, together with the\navailability of powerful computers, has allowed machine learning (ML) modeling\nto become a widely used tool for predicting material performances. While\nconfidence intervals are commonly reported for such ML models, prediction\nintervals, i.e., the uncertainty on each prediction, are not as frequently\navailable. Here, we investigate three easy-to-implement approaches to determine\nsuch individual uncertainty, comparing them across ten ML quantities spanning\nenergetics, mechanical, electronic, optical, and spectral properties.\nSpecifically, we focused on the Quantile approach, the direct machine learning\nof the prediction intervals and Ensemble methods.\n", "date": "2023-10-19", "categories": []}
{"id": "2306.17525", "title": "MeLM, a generative pretrained language modeling framework that solves\n  forward and inverse mechanics problems", "abstract": "  We report a flexible multi-modal mechanics language model, MeLM, applied to\nsolve various nonlinear forward and inverse problems, that can deal with a set\nof instructions, numbers and microstructure data. The framework is applied to\nvarious examples including bio-inspired hierarchical honeycomb design, carbon\nnanotube mechanics, and protein unfolding. In spite of the flexible nature of\nthe model-which allows us to easily incorporate diverse materials, scales, and\nmechanical features-it performs well across disparate forward and inverse\ntasks. Based on an autoregressive attention-model, MeLM effectively represents\na large multi-particle system consisting of hundreds of millions of neurons,\nwhere the interaction potentials are discovered through graph-forming\nself-attention mechanisms that are then used to identify relationships from\nemergent structures, while taking advantage of synergies discovered in the\ntraining data. We show that the model can solve complex degenerate mechanics\ndesign problems and determine novel material architectures across a range of\nhierarchical levels, providing an avenue for materials discovery and analysis.\nLooking beyond the demonstrations reported in this paper, we discuss other\nopportunities in applied mechanics and general considerations about the use of\nlarge language models in modeling, design, and analysis that can span a broad\nspectrum of material properties from mechanical, thermal, optical, to\nelectronic.\n", "date": "2023-06-30", "categories": []}
{"id": "2210.16190", "title": "Transferable E(3) equivariant parameterization for Hamiltonian of\n  molecules and solids", "abstract": "  Using the message-passing mechanism in machine learning (ML) instead of\nself-consistent iterations to directly build the mapping from structures to\nelectronic Hamiltonian matrices will greatly improve the efficiency of density\nfunctional theory (DFT) calculations. In this work, we proposed a general\nanalytic Hamiltonian representation in an E(3) equivariant framework, which can\nfit the ab initio Hamiltonian of molecules and solids by a complete data-driven\nmethod and are equivariant under rotation, space inversion, and time reversal\noperations. Our model reached state-of-the-art precision in the benchmark test\nand accurately predicted the electronic Hamiltonian matrices and related\nproperties of various periodic and aperiodic systems, showing high\ntransferability and generalization ability. This framework provides a general\ntransferable model that can be used to accelerate the electronic structure\ncalculations on different large systems with the same network weights trained\non small structures.\n", "date": "2022-10-28", "categories": []}
{"id": "2310.10695", "title": "Data-Driven Score-Based Models for Generating Stable Structures with\n  Adaptive Crystal Cells", "abstract": "  The discovery of new functional and stable materials is a big challenge due\nto its complexity. This work aims at the generation of new crystal structures\nwith desired properties, such as chemical stability and specified chemical\ncomposition, by using machine learning generative models. Compared to the\ngeneration of molecules, crystal structures pose new difficulties arising from\nthe periodic nature of the crystal and from the specific symmetry constraints\nrelated to the space group. In this work, score-based probabilistic models\nbased on annealed Langevin dynamics, which have shown excellent performance in\nvarious applications, are adapted to the task of crystal generation. The\nnovelty of the presented approach resides in the fact that the lattice of the\ncrystal cell is not fixed. During the training of the model, the lattice is\nlearned from the available data, whereas during the sampling of a new chemical\nstructure, two denoising processes are used in parallel to generate the lattice\nalong the generation of the atomic positions. A multigraph crystal\nrepresentation is introduced that respects symmetry constraints, yielding\ncomputational advantages and a better quality of the sampled structures. We\nshow that our model is capable of generating new candidate structures in any\nchosen chemical system and crystal group without any additional training. To\nillustrate the functionality of the proposed method, a comparison of our model\nto other recent generative models, based on descriptor-based metrics, is\nprovided.\n", "date": "2023-10-16", "categories": []}
{"id": "2310.10056", "title": "Latent Conservative Objective Models for Data-Driven Crystal Structure\n  Prediction", "abstract": "  In computational chemistry, crystal structure prediction (CSP) is an\noptimization problem that involves discovering the lowest energy stable crystal\nstructure for a given chemical formula. This problem is challenging as it\nrequires discovering globally optimal designs with the lowest energies on\ncomplex manifolds. One approach to tackle this problem involves building\nsimulators based on density functional theory (DFT) followed by running search\nin simulation, but these simulators are painfully slow. In this paper, we study\npresent and study an alternate, data-driven approach to crystal structure\nprediction: instead of directly searching for the most stable structures in\nsimulation, we train a surrogate model of the crystal formation energy from a\ndatabase of existing crystal structures, and then optimize this model with\nrespect to the parameters of the crystal structure. This surrogate model is\ntrained to be conservative so as to prevent exploitation of its errors by the\noptimizer. To handle optimization in the non-Euclidean space of crystal\nstructures, we first utilize a state-of-the-art graph diffusion auto-encoder\n(CD-VAE) to convert a crystal structure into a vector-based search space and\nthen optimize a conservative surrogate model of the crystal energy, trained on\ntop of this vector representation. We show that our approach, dubbed LCOMs\n(latent conservative objective models), performs comparably to the best current\napproaches in terms of success rate of structure prediction, while also\ndrastically reducing computational cost.\n", "date": "2023-10-16", "categories": []}
{"id": "2306.13345", "title": "Accurate melting point prediction through autonomous physics-informed\n  learning", "abstract": "  We present an algorithm for computing melting points by autonomously learning\nfrom coexistence simulations in the NPT ensemble. Given the interatomic\ninteraction model, the method makes decisions regarding the number of atoms and\ntemperature at which to conduct simulations, and based on the collected data\npredicts the melting point along with the uncertainty, which can be\nsystematically improved with more data. We demonstrate how incorporating\nphysical models of the solid-liquid coexistence evolution enhances the\nalgorithm's accuracy and enables optimal decision-making to effectively reduce\npredictive uncertainty. To validate our approach, we compare the results of 20\nmelting point calculations from the literature to the results of our\ncalculations, all conducted with same interatomic potentials. Remarkably, we\nobserve significant deviations in about one-third of the cases, underscoring\nthe need for accurate and reliable algorithms for materials property\ncalculations.\n", "date": "2023-06-23", "categories": []}
{"id": "2310.07927", "title": "Enhanced sampling of Crystal Nucleation with Graph Representation Learnt\n  Variables", "abstract": "  In this study, we present a graph neural network-based learning approach\nusing an autoencoder setup to derive low-dimensional variables from features\nobserved in experimental crystal structures. These variables are then biased in\nenhanced sampling to observe state-to-state transitions and reliable\nthermodynamic weights. Our approach uses simple convolution and pooling\nmethods. To verify the effectiveness of our protocol, we examined the\nnucleation of various allotropes and polymorphs of iron and glycine from their\nmolten states. Our graph latent variables when biased in well-tempered\nmetadynamics consistently show transitions between states and achieve accurate\nfree energy calculations in agreement with experiments, both of which are\nindicators of dependable sampling. This underscores the strength and promise of\nour graph neural net variables for improved sampling. The protocol shown here\nshould be applicable for other systems and with other sampling methods.\n", "date": "2023-10-11", "categories": []}
{"id": "2303.11699", "title": "Neural networks trained on synthetically generated crystals can extract\n  structural information from ICSD powder X-ray diffractograms", "abstract": "  Machine learning techniques have successfully been used to extract structural\ninformation such as the crystal space group from powder X-ray diffractograms.\nHowever, training directly on simulated diffractograms from databases such as\nthe ICSD is challenging due to its limited size, class-inhomogeneity, and bias\ntoward certain structure types. We propose an alternative approach of\ngenerating synthetic crystals with random coordinates by using the symmetry\noperations of each space group. Based on this approach, we demonstrate online\ntraining of deep ResNet-like models on up to a few million unique on-the-fly\ngenerated synthetic diffractograms per hour. For our chosen task of space group\nclassification, we achieved a test accuracy of 79.9% on unseen ICSD structure\ntypes from most space groups. This surpasses the 56.1% accuracy of the current\nstate-of-the-art approach of training on ICSD crystals directly. Our results\ndemonstrate that synthetically generated crystals can be used to extract\nstructural information from ICSD powder diffractograms, which makes it possible\nto apply very large state-of-the-art machine learning models in the area of\npowder X-ray diffraction. We further show first steps toward applying our\nmethodology to experimental data, where automated XRD data analysis is crucial,\nespecially in high-throughput settings. While we focused on the prediction of\nthe space group, our approach has the potential to be extended to related tasks\nin the future.\n", "date": "2023-03-21", "categories": []}
{"id": "2303.14434", "title": "Heat flux for semi-local machine-learning potentials", "abstract": "  The Green-Kubo (GK) method is a rigorous framework for heat transport\nsimulations in materials. However, it requires an accurate description of the\npotential-energy surface and carefully converged statistics. Machine-learning\npotentials can achieve the accuracy of first-principles simulations while\nallowing to reach well beyond their simulation time and length scales at a\nfraction of the cost. In this paper, we explain how to apply the GK approach to\nthe recent class of message-passing machine-learning potentials, which\niteratively consider semi-local interactions beyond the initial interaction\ncutoff. We derive an adapted heat flux formulation that can be implemented\nusing automatic differentiation without compromising computational efficiency.\nThe approach is demonstrated and validated by calculating the thermal\nconductivity of zirconium dioxide across temperatures.\n", "date": "2023-03-25", "categories": []}
{"id": "2310.06852", "title": "DeepCrysTet: A Deep Learning Approach Using Tetrahedral Mesh for\n  Predicting Properties of Crystalline Materials", "abstract": "  Machine learning (ML) is becoming increasingly popular for predicting\nmaterial properties to accelerate materials discovery. Because material\nproperties are strongly affected by its crystal structure, a key issue is\nconverting the crystal structure into the features for input to the ML model.\nCurrently, the most common method is to convert the crystal structure into a\ngraph and predicting its properties using a graph neural network (GNN). Some\nGNN models, such as crystal graph convolutional neural network (CGCNN) and\natomistic line graph neural network (ALIGNN), have achieved highly accurate\npredictions of material properties. Despite these successes, using a graph to\nrepresent a crystal structure has the notable limitation of losing the crystal\nstructure's three-dimensional (3D) information. In this work, we propose\nDeepCrysTet, a novel deep learning approach for predicting material properties,\nwhich uses crystal structures represented as a 3D tetrahedral mesh generated by\nDelaunay tetrahedralization. DeepCrysTet provides a useful framework that\nincludes a 3D mesh generation method, mesh-based feature design, and neural\nnetwork design. The experimental results using the Materials Project dataset\nshow that DeepCrysTet significantly outperforms existing GNN models in\nclassifying crystal structures and achieves state-of-the-art performance in\npredicting elastic properties.\n", "date": "2023-09-07", "categories": []}
{"id": "2310.06682", "title": "On the importance of catalyst-adsorbate 3D interactions for relaxed\n  energy predictions", "abstract": "  The use of machine learning for material property prediction and discovery\nhas traditionally centered on graph neural networks that incorporate the\ngeometric configuration of all atoms. However, in practice not all this\ninformation may be readily available, e.g.~when evaluating the potentially\nunknown binding of adsorbates to catalyst. In this paper, we investigate\nwhether it is possible to predict a system's relaxed energy in the OC20 dataset\nwhile ignoring the relative position of the adsorbate with respect to the\nelectro-catalyst. We consider SchNet, DimeNet++ and FAENet as base\narchitectures and measure the impact of four modifications on model\nperformance: removing edges in the input graph, pooling independent\nrepresentations, not sharing the backbone weights and using an attention\nmechanism to propagate non-geometric relative information. We find that while\nremoving binding site information impairs accuracy as expected, modified models\nare able to predict relaxed energies with remarkably decent MAE. Our work\nsuggests future research directions in accelerated materials discovery where\ninformation on reactant configurations can be reduced or altogether omitted.\n", "date": "2023-10-10", "categories": []}
{"id": "2310.04622", "title": "FluxGAN: A Physics-Aware Generative Adversarial Network Model for\n  Generating Microstructures That Maintain Target Heat Flux", "abstract": "  We propose a physics-aware generative adversarial network model, FluxGAN,\ncapable of simultaneously generating high-quality images of large\nmicrostructures and description of their thermal properties. During the\ntraining phase, the model learns about the relationship between the local\nstructural features and the physical processes, such as the heat flux in the\nmicrostructures, due to external temperature gradients. Once trained, the model\ngenerates new structural and associated heat flux environments, bypassing the\ncomputationally expensive modeling. Our model provides a cost effective and\nefficient approach over conventional modeling techniques, such as the finite\nelement method (FEM), for describing the thermal properties of microstructures.\nThe conventional approach requires computational modeling that scales with the\nsize of the microstructure model, therefore limiting the simulation to a given\nsize, resolution, and complexity of the model. In contrast, the FluxGAN model\nuses synthesis-by-part approach and generates arbitrary large size images at\nlow computational cost. We demonstrate that the model can be utilized to\ngenerate designs of thermal sprayed coatings that satisfies target thermal\nproperties. Furthermore, the model is capable of generating coating\nmicrostructures and physical processes in three-dimensional (3D) domain after\nbeing trained on two-dimensional (2D) examples. Our approach has the potential\nto transform the design and optimization of thermal sprayed coatings for\nvarious applications, including high-temperature and long-duration operation of\ngas turbines for aircraft or ground-based power generators.\n", "date": "2023-10-06", "categories": []}
{"id": "2310.03745", "title": "Generative Hyperelasticity with Physics-Informed Probabilistic Diffusion\n  Fields", "abstract": "  Many natural materials exhibit highly complex, nonlinear, anisotropic, and\nheterogeneous mechanical properties. Recently, it has been demonstrated that\ndata-driven strain energy functions possess the flexibility to capture the\nbehavior of these complex materials with high accuracy while satisfying\nphysics-based constraints. However, most of these approaches disregard the\nuncertainty in the estimates and the spatial heterogeneity of these materials.\nIn this work, we leverage recent advances in generative models to address these\nissues. We use as building block neural ordinary equations (NODE) that -- by\nconstruction -- create polyconvex strain energy functions, a key property of\nrealistic hyperelastic material models. We combine this approach with\nprobabilistic diffusion models to generate new samples of strain energy\nfunctions. This technique allows us to sample a vector of Gaussian white noise\nand translate it to NODE parameters thereby representing plausible strain\nenergy functions. We extend our approach to spatially correlated diffusion\nresulting in heterogeneous material properties for arbitrary geometries. We\nextensively test our method with synthetic and experimental data on biological\ntissues and run finite element simulations with various degrees of spatial\nheterogeneity. We believe this approach is a major step forward including\nuncertainty in predictive, data-driven models of hyperelasticity\n", "date": "2023-09-11", "categories": []}
{"id": "2308.13208", "title": "Physics-inspired Equivariant Descriptors of Non-bonded Interactions", "abstract": "  One essential ingredient in many machine learning (ML) based methods for\natomistic modeling of materials and molecules is the use of locality. While\nallowing better system-size scaling, this systematically neglects long-range\n(LR) effects, such as electrostatics or dispersion interaction. We present an\nextension of the long distance equivariant (LODE) framework that can handle\ndiverse LR interactions in a consistent way, and seamlessly integrates with\npreexisting methods by building new sets of atom centered features. We provide\na direct physical interpretation of these using the multipole expansion, which\nallows for simpler and more efficient implementations. The framework is applied\nto simple toy systems as proof of concept, and a heterogeneous set of molecular\ndimers to push the method to its limits. By generalizing LODE to arbitrary\nasymptotic behaviors, we provide a coherent approach to treat arbitrary two-\nand many-body non-bonded interactions in the data-driven modeling of matter.\n", "date": "2023-08-25", "categories": []}
{"id": "2310.02904", "title": "Spline-based neural network interatomic potentials: blending classical\n  and machine learning models", "abstract": "  While machine learning (ML) interatomic potentials (IPs) are able to achieve\naccuracies nearing the level of noise inherent in the first-principles data to\nwhich they are trained, it remains to be shown if their increased complexities\nare strictly necessary for constructing high-quality IPs. In this work, we\nintroduce a new MLIP framework which blends the simplicity of spline-based MEAM\n(s-MEAM) potentials with the flexibility of a neural network (NN) architecture.\nThe proposed framework, which we call the spline-based neural network potential\n(s-NNP), is a simplified version of the traditional NNP that can be used to\ndescribe complex datasets in a computationally efficient manner. We demonstrate\nhow this framework can be used to probe the boundary between classical and ML\nIPs, highlighting the benefits of key architectural changes. Furthermore, we\nshow that using spline filters for encoding atomic environments results in a\nreadily interpreted embedding layer which can be coupled with modifications to\nthe NN to incorporate expected physical behaviors and improve overall\ninterpretability. Finally, we test the flexibility of the spline filters,\nobserving that they can be shared across multiple chemical systems in order to\nprovide a convenient reference point from which to begin performing\ncross-system analyses.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.00475", "title": "Generative Design of inorganic compounds using deep diffusion language\n  models", "abstract": "  Due to the vast chemical space, discovering materials with a specific\nfunction is challenging. Chemical formulas are obligated to conform to a set of\nexacting criteria such as charge neutrality, balanced electronegativity,\nsynthesizability, and mechanical stability. In response to this formidable\ntask, we introduce a deep learning-based generative model for material\ncomposition and structure design by learning and exploiting explicit and\nimplicit chemical knowledge. Our pipeline first uses deep diffusion language\nmodels as the generator of compositions and then applies a template-based\ncrystal structure prediction algorithm to predict their corresponding\nstructures, which is then followed by structure relaxation using a universal\ngraph neural network-based potential. The density functional theory (DFT)\ncalculations of the formation energies and energy-above-the-hull analysis are\nused to validate new structures generated through our pipeline. Based on the\nDFT calculation results, six new materials, including Ti2HfO5, TaNbP, YMoN2,\nTaReO4, HfTiO2, and HfMnO2, with formation energy less than zero have been\nfound. Remarkably, among these, four materials, namely Ti2$HfO5, TaNbP, YMoN2,\nand TaReO4, exhibit an e-above-hull energy of less than 0.3 eV. These findings\nhave proved the effectiveness of our approach.\n", "date": "2023-09-30", "categories": []}
{"id": "2310.01117", "title": "Predicting emergence of crystals from amorphous matter with deep\n  learning", "abstract": "  Crystallization of the amorphous phases into metastable crystals plays a\nfundamental role in the formation of new matter, from geological to biological\nprocesses in nature to synthesis and development of new materials in the\nlaboratory. Predicting the outcome of such phase transitions reliably would\nenable new research directions in these areas, but has remained beyond reach\nwith molecular modeling or ab-initio methods. Here, we show that\ncrystallization products of amorphous phases can be predicted in any inorganic\nchemistry by sampling the crystallization pathways of their local structural\nmotifs at the atomistic level using universal deep learning potentials. We show\nthat this approach identifies the crystal structures of polymorphs that\ninitially nucleate from amorphous precursors with high accuracy across a\ndiverse set of material systems, including polymorphic oxides, nitrides,\ncarbides, fluorides, chlorides, chalcogenides, and metal alloys. Our results\ndemonstrate that Ostwald's rule of stages can be exploited mechanistically at\nthe molecular level to predictably access new metastable crystals from the\namorphous phase in material synthesis.\n", "date": "2023-10-02", "categories": []}
{"id": "2309.16374", "title": "MHG-GNN: Combination of Molecular Hypergraph Grammar with Graph Neural\n  Network", "abstract": "  Property prediction plays an important role in material discovery. As an\ninitial step to eventually develop a foundation model for material science, we\nintroduce a new autoencoder called the MHG-GNN, which combines graph neural\nnetwork (GNN) with Molecular Hypergraph Grammar (MHG). Results on a variety of\nproperty prediction tasks with diverse materials show that MHG-GNN is\npromising.\n", "date": "2023-09-28", "categories": []}
{"id": "2309.14450", "title": "Learning dislocation dynamics mobility laws from large-scale MD\n  simulations", "abstract": "  The computational method of discrete dislocation dynamics (DDD), used as a\ncoarse-grained model of true atomistic dynamics of lattice dislocations, has\nbecome of powerful tool to study metal plasticity arising from the collective\nbehavior of dislocations. As a mesoscale approach, motion of dislocations in\nthe DDD model is prescribed via the mobility law; a function which specifies\nhow dislocation lines should respond to the driving force. However, the\ndevelopment of traditional hand-crafted mobility laws can be a cumbersome task\nand may involve detrimental simplifications. Here we introduce a\nmachine-learning (ML) framework to streamline the development of data-driven\nmobility laws which are modeled as graph neural networks (GNN) trained on\nlarge-scale Molecular Dynamics (MD) simulations of crystal plasticity. We\nillustrate our approach on BCC tungsten and demonstrate that our GNN mobility\nimplemented in large-scale DDD simulations accurately reproduces the\nchallenging tension/compression asymmetry observed in ground-truth MD\nsimulations while correctly predicting the flow stress at lower straining rate\nconditions unseen during training, thereby demonstrating the ability of our\nmethod to learn relevant dislocation physics. Our DDD+ML approach opens new\npromising avenues to improve fidelity of the DDD model and to incorporate more\ncomplex dislocation motion behaviors in an automated way, providing a faithful\nproxy for dislocation dynamics several orders of magnitude faster than\nground-truth MD simulations.\n", "date": "2023-09-25", "categories": []}
{"id": "2308.14035", "title": "Multi-plane denoising diffusion-based dimensionality expansion for\n  2D-to-3D reconstruction of microstructures with harmonized sampling", "abstract": "  Acquiring reliable microstructure datasets is a pivotal step toward the\nsystematic design of materials with the aid of integrated computational\nmaterials engineering (ICME) approaches. However, obtaining three-dimensional\n(3D) microstructure datasets is often challenging due to high experimental\ncosts or technical limitations, while acquiring two-dimensional (2D)\nmicrographs is comparatively easier. To deal with this issue, this study\nproposes a novel framework for 2D-to-3D reconstruction of microstructures\ncalled Micro3Diff using diffusion-based generative models (DGMs). Specifically,\nthis approach solely requires pre-trained DGMs for the generation of 2D\nsamples, and dimensionality expansion (2D-to-3D) takes place only during the\ngeneration process (i.e., reverse diffusion process). The proposed framework\nincorporates a new concept referred to as multi-plane denoising diffusion,\nwhich transforms noisy samples (i.e., latent variables) from different planes\ninto the data structure while maintaining spatial connectivity in 3D space.\nFurthermore, a harmonized sampling process is developed to address possible\ndeviations from the reverse Markov chain of DGMs during the dimensionality\nexpansion. Combined, we demonstrate the feasibility of Micro3Diff in\nreconstructing 3D samples with connected slices that maintain morphologically\nequivalence to the original 2D images. To validate the performance of\nMicro3Diff, various types of microstructures (synthetic and experimentally\nobserved) are reconstructed, and the quality of the generated samples is\nassessed both qualitatively and quantitatively. The successful reconstruction\noutcomes inspire the potential utilization of Micro3Diff in upcoming ICME\napplications while achieving a breakthrough in comprehending and manipulating\nthe latent space of DGMs.\n", "date": "2023-08-27", "categories": []}
{"id": "2309.13626", "title": "Crack-Net: Prediction of Crack Propagation in Composites", "abstract": "  Computational solid mechanics has become an indispensable approach in\nengineering, and numerical investigation of fracture in composites is essential\nas composites are widely used in structural applications. Crack evolution in\ncomposites is the bridge to elucidate the relationship between the\nmicrostructure and fracture performance, but crack-based finite element methods\nare computationally expensive and time-consuming, limiting their application in\ncomputation-intensive scenarios. Here we propose a deep learning framework\ncalled Crack-Net, which incorporates the relationship between crack evolution\nand stress response to predict the fracture process in composites. Trained on a\nhigh-precision fracture development dataset generated using the phase field\nmethod, Crack-Net demonstrates a remarkable capability to accurately forecast\nthe long-term evolution of crack growth patterns and the stress-strain curve\nfor a given composite design. The Crack-Net captures the essential principle of\ncrack growth, which enables it to handle more complex microstructures such as\nbinary co-continuous structures. Moreover, transfer learning is adopted to\nfurther improve the generalization ability of Crack-Net for composite materials\nwith reinforcements of different strengths. The proposed Crack-Net holds great\npromise for practical applications in engineering and materials science, in\nwhich accurate and efficient fracture prediction is crucial for optimizing\nmaterial performance and microstructural design.\n", "date": "2023-09-24", "categories": []}
{"id": "2304.05422", "title": "Differentiable graph-structured models for inverse design of lattice\n  materials", "abstract": "  Architected materials possessing physico-chemical properties adaptable to\ndisparate environmental conditions embody a disruptive new domain of materials\nscience. Fueled by advances in digital design and fabrication, materials shaped\ninto lattice topologies enable a degree of property customization not afforded\nto bulk materials. A promising venue for inspiration toward their design is in\nthe irregular micro-architectures of nature. However, the immense design\nvariability unlocked by such irregularity is challenging to probe analytically.\nHere, we propose a new computational approach using graph-based representation\nfor regular and irregular lattice materials. Our method uses differentiable\nmessage passing algorithms to calculate mechanical properties, therefore\nallowing automatic differentiation with surrogate derivatives to adjust both\ngeometric structure and local attributes of individual lattice elements to\nachieve inversely designed materials with desired properties. We further\nintroduce a graph neural network surrogate model for structural analysis at\nscale. The methodology is generalizable to any system representable as\nheterogeneous graphs.\n", "date": "2023-04-11", "categories": []}
{"id": "2309.12323", "title": "Evaluating the diversity and utility of materials proposed by generative\n  models", "abstract": "  Generative machine learning models can use data generated by scientific\nmodeling to create large quantities of novel material structures. Here, we\nassess how one state-of-the-art generative model, the physics-guided crystal\ngeneration model (PGCGM), can be used as part of the inverse design process. We\nshow that the default PGCGM's input space is not smooth with respect to\nparameter variation, making material optimization difficult and limited. We\nalso demonstrate that most generated structures are predicted to be\nthermodynamically unstable by a separate property-prediction model, partially\ndue to out-of-domain data challenges. Our findings suggest how generative\nmodels might be improved to enable better inverse design.\n", "date": "2023-08-09", "categories": []}
{"id": "2309.10600", "title": "Neural Metamaterial Networks for Nonlinear Material Design", "abstract": "  Nonlinear metamaterials with tailored mechanical properties have applications\nin engineering, medicine, robotics, and beyond. While modeling their\nmacromechanical behavior is challenging in itself, finding structure parameters\nthat lead to ideal approximation of high-level performance goals is a\nchallenging task. In this work, we propose Neural Metamaterial Networks (NMN)\n-- smooth neural representations that encode the nonlinear mechanics of entire\nmetamaterial families. Given structure parameters as input, NMN return\ncontinuously differentiable strain energy density functions, thus guaranteeing\nconservative forces by construction. Though trained on simulation data, NMN do\nnot inherit the discontinuities resulting from topological changes in finite\nelement meshes. They instead provide a smooth map from parameter to performance\nspace that is fully differentiable and thus well-suited for gradient-based\noptimization. On this basis, we formulate inverse material design as a\nnonlinear programming problem that leverages neural networks for both objective\nfunctions and constraints. We use this approach to automatically design\nmaterials with desired strain-stress curves, prescribed directional stiffness\nand Poisson ratio profiles. We furthermore conduct ablation studies on network\nnonlinearities and show the advantages of our approach compared to native-scale\noptimization.\n", "date": "2023-09-15", "categories": []}
{"id": "2210.06310", "title": "Deep learning extraction of band structure parameters from density of\n  states: a case study on trilayer graphene", "abstract": "  The development of two-dimensional materials has resulted in a diverse range\nof novel, high-quality compounds with increasing complexity. A key requirement\nfor a comprehensive quantitative theory is the accurate determination of these\nmaterials' band structure parameters. However, this task is challenging due to\nthe intricate band structures and the indirect nature of experimental probes.\nIn this work, we introduce a general framework to derive band structure\nparameters from experimental data using deep neural networks. We applied our\nmethod to the penetration field capacitance measurement of trilayer graphene,\nan effective probe of its density of states. First, we demonstrate that a\ntrained deep network gives accurate predictions for the penetration field\ncapacitance as a function of tight-binding parameters. Next, we use the fast\nand accurate predictions from the trained network to automatically determine\ntight-binding parameters directly from experimental data, with extracted\nparameters being in a good agreement with values in the literature. We conclude\nby discussing potential applications of our method to other materials and\nexperimental techniques beyond penetration field capacitance.\n", "date": "2022-10-12", "categories": []}
{"id": "2211.16486", "title": "AdsorbML: A Leap in Efficiency for Adsorption Energy Calculations using\n  Generalizable Machine Learning Potentials", "abstract": "  Computational catalysis is playing an increasingly significant role in the\ndesign of catalysts across a wide range of applications. A common task for many\ncomputational methods is the need to accurately compute the adsorption energy\nfor an adsorbate and a catalyst surface of interest. Traditionally, the\nidentification of low energy adsorbate-surface configurations relies on\nheuristic methods and researcher intuition. As the desire to perform\nhigh-throughput screening increases, it becomes challenging to use heuristics\nand intuition alone. In this paper, we demonstrate machine learning potentials\ncan be leveraged to identify low energy adsorbate-surface configurations more\naccurately and efficiently. Our algorithm provides a spectrum of trade-offs\nbetween accuracy and efficiency, with one balanced option finding the lowest\nenergy configuration 87.36% of the time, while achieving a 2000x speedup in\ncomputation. To standardize benchmarking, we introduce the Open Catalyst Dense\ndataset containing nearly 1,000 diverse surfaces and 100,000 unique\nconfigurations.\n", "date": "2022-11-29", "categories": []}
{"id": "2208.08236", "title": "DPA-1: Pretraining of Attention-based Deep Potential Model for Molecular\n  Simulation", "abstract": "  Machine learning assisted modeling of the inter-atomic potential energy\nsurface (PES) is revolutionizing the field of molecular simulation. With the\naccumulation of high-quality electronic structure data, a model that can be\npretrained on all available data and finetuned on downstream tasks with a small\nadditional effort would bring the field to a new stage. Here we propose DPA-1,\na Deep Potential model with a novel attention mechanism, which is highly\neffective for representing the conformation and chemical spaces of atomic\nsystems and learning the PES. We tested DPA-1 on a number of systems and\nobserved superior performance compared with existing benchmarks. When\npretrained on large-scale datasets containing 56 elements, DPA-1 can be\nsuccessfully applied to various downstream tasks with a great improvement of\nsample efficiency. Surprisingly, for different elements, the learned type\nembedding parameters form a $spiral$ in the latent space and have a natural\ncorrespondence with their positions on the periodic table, showing interesting\ninterpretability of the pretrained DPA-1 model.\n", "date": "2022-08-17", "categories": []}
{"id": "2309.06613", "title": "Unsupervised Learning of Nanoindentation Data to Infer Microstructural\n  Details of Complex Materials", "abstract": "  In this study, Cu-Cr composites were studied by nanoindentation. Arrays of\nindents were placed over large areas of the samples resulting in datasets\nconsisting of several hundred measurements of Young's modulus and hardness at\nvarying indentation depths. The unsupervised learning technique, Gaussian\nmixture model, was employed to analyze the data, which helped to determine the\nnumber of \"mechanical phases\" and the respective mechanical properties.\nAdditionally, a cross-validation approach was introduced to infer whether the\ndata quantity was adequate and to suggest the amount of data required for\nreliable predictions -- one of the often encountered but difficult to resolve\nissues in machine learning of materials science problems.\n", "date": "2023-09-12", "categories": []}
{"id": "2309.06710", "title": "Crystal structure prediction using neural network potential and\n  age-fitness Pareto genetic algorithm", "abstract": "  While crystal structure prediction (CSP) remains a longstanding challenge, we\nintroduce ParetoCSP, a novel algorithm for CSP, which combines a\nmulti-objective genetic algorithm (MOGA) with a neural network inter-atomic\npotential (IAP) model to find energetically optimal crystal structures given\nchemical compositions. We enhance the NSGA-III algorithm by incorporating the\ngenotypic age as an independent optimization criterion and employ the M3GNet\nuniversal IAP to guide the GA search. Compared to GN-OA, a state-of-the-art\nneural potential based CSP algorithm, ParetoCSP demonstrated significantly\nbetter predictive capabilities, outperforming by a factor of $2.562$ across\n$55$ diverse benchmark structures, as evaluated by seven performance metrics.\nTrajectory analysis of the traversed structures of all algorithms shows that\nParetoCSP generated more valid structures than other algorithms, which helped\nguide the GA to search more effectively for the optimal structures\n", "date": "2023-09-13", "categories": []}
{"id": "2303.05911", "title": "Lifelong Machine Learning Potentials", "abstract": "  Machine learning potentials (MLPs) trained on accurate quantum chemical data\ncan retain the high accuracy, while inflicting little computational demands. On\nthe downside, they need to be trained for each individual system. In recent\nyears, a vast number of MLPs has been trained from scratch because learning\nadditional data typically requires to train again on all data to not forget\npreviously acquired knowledge. Additionally, most common structural descriptors\nof MLPs cannot represent efficiently a large number of different chemical\nelements. In this work, we tackle these problems by introducing\nelement-embracing atom-centered symmetry functions (eeACSFs) which combine\nstructural properties and element information from the periodic table. These\neeACSFs are a key for our development of a lifelong machine learning potential\n(lMLP). Uncertainty quantification can be exploited to transgress a fixed,\npre-trained MLP to arrive at a continuously adapting lMLP, because a predefined\nlevel of accuracy can be ensured. To extend the applicability of an lMLP to new\nsystems, we apply continual learning strategies to enable autonomous and\non-the-fly training on a continuous stream of new data. For the training of\ndeep neural networks, we propose the continual resilient (CoRe) optimizer and\nincremental learning strategies relying on rehearsal of data, regularization of\nparameters, and the architecture of the model.\n", "date": "2023-03-10", "categories": []}
{"id": "2309.05934", "title": "MatSciML: A Broad, Multi-Task Benchmark for Solid-State Materials\n  Modeling", "abstract": "  We propose MatSci ML, a novel benchmark for modeling MATerials SCIence using\nMachine Learning (MatSci ML) methods focused on solid-state materials with\nperiodic crystal structures. Applying machine learning methods to solid-state\nmaterials is a nascent field with substantial fragmentation largely driven by\nthe great variety of datasets used to develop machine learning models. This\nfragmentation makes comparing the performance and generalizability of different\nmethods difficult, thereby hindering overall research progress in the field.\nBuilding on top of open-source datasets, including large-scale datasets like\nthe OpenCatalyst, OQMD, NOMAD, the Carolina Materials Database, and Materials\nProject, the MatSci ML benchmark provides a diverse set of materials systems\nand properties data for model training and evaluation, including simulated\nenergies, atomic forces, material bandgaps, as well as classification data for\ncrystal symmetries via space groups. The diversity of properties in MatSci ML\nmakes the implementation and evaluation of multi-task learning algorithms for\nsolid-state materials possible, while the diversity of datasets facilitates the\ndevelopment of new, more generalized algorithms and methods across multiple\ndatasets. In the multi-dataset learning setting, MatSci ML enables researchers\nto combine observations from multiple datasets to perform joint prediction of\ncommon properties, such as energy and forces. Using MatSci ML, we evaluate the\nperformance of different graph neural networks and equivariant point cloud\nnetworks on several benchmark tasks spanning single task, multitask, and\nmulti-data learning scenarios. Our open-source code is available at\nhttps://github.com/IntelLabs/matsciml.\n", "date": "2023-09-12", "categories": []}
{"id": "2309.06348", "title": "Band-gap regression with architecture-optimized message-passing neural\n  networks", "abstract": "  Graph-based neural networks and, specifically, message-passing neural\nnetworks (MPNNs) have shown great potential in predicting physical properties\nof solids. In this work, we train an MPNN to first classify materials through\ndensity functional theory data from the AFLOW database as being metallic or\nsemiconducting/insulating. We then perform a neural-architecture search to\nexplore the model architecture and hyperparameter space of MPNNs to predict the\nband gaps of the materials identified as non-metals. The parameters in the\nsearch include the number of message-passing steps, latent size, and\nactivation-function, among others. The top-performing models from the search\nare pooled into an ensemble that significantly outperforms existing models from\nthe literature. Uncertainty quantification is evaluated with Monte-Carlo\nDropout and ensembling, with the ensemble method proving superior. The domain\nof applicability of the ensemble model is analyzed with respect to the crystal\nsystems, the inclusion of a Hubbard parameter in the density functional\ncalculations, and the atomic species building up the materials.\n", "date": "2023-09-12", "categories": []}
{"id": "2309.04478", "title": "Multimodal machine learning for materials science: composition-structure\n  bimodal learning for experimentally measured properties", "abstract": "  The widespread application of multimodal machine learning models like GPT-4\nhas revolutionized various research fields including computer vision and\nnatural language processing. However, its implementation in materials\ninformatics remains underexplored, despite the presence of materials data\nacross diverse modalities, such as composition and structure. The effectiveness\nof machine learning models trained on large calculated datasets depends on the\naccuracy of calculations, while experimental datasets often have limited data\navailability and incomplete information. This paper introduces a novel approach\nto multimodal machine learning in materials science via composition-structure\nbimodal learning. The proposed COmposition-Structure Bimodal Network (COSNet)\nis designed to enhance learning and predictions of experimentally measured\nmaterials properties that have incomplete structure information. Bimodal\nlearning significantly reduces prediction errors across distinct materials\nproperties including Li conductivity in solid electrolyte, band gap, refractive\nindex, dielectric constant, energy, and magnetic moment, surpassing\ncomposition-only learning methods. Furthermore, we identified that data\naugmentation based on modal availability plays a pivotal role in the success of\nbimodal learning.\n", "date": "2023-08-04", "categories": []}
{"id": "2309.04482", "title": "Addressing the Accuracy-Cost Tradeoff in Material Property Prediction: A\n  Teacher-Student Strategy", "abstract": "  Deep learning has revolutionized the process of new material discovery, with\nstate-of-the-art models now able to predict material properties based solely on\nchemical compositions, thus eliminating the necessity for material structures.\nHowever, this cost-effective method has led to a trade-off in model accuracy.\nSpecifically, the accuracy of Chemical Composition-based Property Prediction\nModels (CPMs) significantly lags behind that of Structure-based Property\nPrediction Models (SPMs). To tackle this challenge, we propose an innovative\nTeacher-Student (T-S) strategy, where a pre-trained SPM serves as the 'teacher'\nto enhance the accuracy of the CPM. Leveraging the T-S strategy, T-S CrabNet\nhas risen to become the most accurate model among current CPMs. Initially, we\ndemonstrated the universality of this strategy. On the Materials Project (MP)\nand Jarvis datasets, we validated the effectiveness of the T-S strategy in\nboosting the accuracy of CPMs with two distinct network structures, namely\nCrabNet and Roost. This led to CrabNet, under the guidance of the T-S strategy,\nemerging as the most accurate model among the current CPMs. Moreover, this\nstrategy shows remarkable efficacy in small datasets. When predicting the\nformation energy on a small MP dataset comprising merely 5% of the samples, the\nT-S strategy boosted CrabNet's accuracy by 37.1%, exceeding the enhancement\neffect of the T-S strategy on the whole dataset.\n", "date": "2023-08-22", "categories": []}
{"id": "2304.06044", "title": "Learning solution of nonlinear constitutive material models using\n  physics-informed neural networks: COMM-PINN", "abstract": "  We applied physics-informed neural networks to solve the constitutive\nrelations for nonlinear, path-dependent material behavior. As a result, the\ntrained network not only satisfies all thermodynamic constraints but also\ninstantly provides information about the current material state (i.e., free\nenergy, stress, and the evolution of internal variables) under any given\nloading scenario without requiring initial data. One advantage of this work is\nthat it bypasses the repetitive Newton iterations needed to solve nonlinear\nequations in complex material models. Additionally, strategies are provided to\nreduce the required order of derivative for obtaining the tangent operator. The\ntrained model can be directly used in any finite element package (or other\nnumerical methods) as a user-defined material model. However, challenges remain\nin the proper definition of collocation points and in integrating several\nnon-equality constraints that become active or non-active simultaneously. We\ntested this methodology on rate-independent processes such as the classical von\nMises plasticity model with a nonlinear hardening law, as well as local damage\nmodels for interface cracking behavior with a nonlinear softening law. In order\nto demonstrate the applicability of the methodology in handling complex path\ndependency in a three-dimensional (3D) scenario, we tested the approach using\nthe equations governing a damage model for a three-dimensional interface model.\nSuch models are frequently employed for intergranular fracture at grain\nboundaries. We have observed a perfect agreement between the results obtained\nthrough the proposed methodology and those obtained using the classical\napproach. Furthermore, the proposed approach requires significantly less effort\nin terms of implementation and computing time compared to the traditional\nmethods.\n", "date": "2023-04-10", "categories": []}
{"id": "2309.01271", "title": "Bayesian inference of composition-dependent phase diagrams", "abstract": "  Phase diagrams serve as a highly informative tool for materials design,\nencapsulating information about the phases that a material can manifest under\nspecific conditions. In this work, we develop a method in which Bayesian\ninference is employed to combine thermodynamic data from molecular dynamics\n(MD), melting point simulations, and phonon calculations, process these data,\nand yield a temperature-concentration phase diagram. The employed Bayesian\nframework yields us not only the free energies of different phases as functions\nof temperature and concentration but also the uncertainties of these free\nenergies originating from statistical errors inherent to finite-length MD\ntrajectories. Furthermore, it extrapolates the results of the finite-atom\ncalculations to the infinite-atom limit and facilitates the choice of\ntemperature, chemical potentials, and the number of atoms conducting the next\nsimulation with which will be the most efficient in reducing the uncertainty of\nthe phase diagram. The developed algorithm was successfully tested on two\nbinary systems, Ge-Si and K-Na, in the full range of concentrations and\ntemperatures.\n", "date": "2023-09-03", "categories": []}
{"id": "2308.16259", "title": "Materials Informatics Transformer: A Language Model for Interpretable\n  Materials Properties Prediction", "abstract": "  Recently, the remarkable capabilities of large language models (LLMs) have\nbeen illustrated across a variety of research domains such as natural language\nprocessing, computer vision, and molecular modeling. We extend this paradigm by\nutilizing LLMs for material property prediction by introducing our model\nMaterials Informatics Transformer (MatInFormer). Specifically, we introduce a\nnovel approach that involves learning the grammar of crystallography through\nthe tokenization of pertinent space group information. We further illustrate\nthe adaptability of MatInFormer by incorporating task-specific data pertaining\nto Metal-Organic Frameworks (MOFs). Through attention visualization, we uncover\nthe key features that the model prioritizes during property prediction. The\neffectiveness of our proposed model is empirically validated across 14 distinct\ndatasets, hereby underscoring its potential for high throughput screening\nthrough accurate material property prediction.\n", "date": "2023-08-30", "categories": []}
{"id": "2309.00585", "title": "PolyGET: Accelerating Polymer Simulations by Accurate and Generalizable\n  Forcefield with Equivariant Transformer", "abstract": "  Polymer simulation with both accuracy and efficiency is a challenging task.\nMachine learning (ML) forcefields have been developed to achieve both the\naccuracy of ab initio methods and the efficiency of empirical force fields.\nHowever, existing ML force fields are usually limited to single-molecule\nsettings, and their simulations are not robust enough. In this paper, we\npresent PolyGET, a new framework for Polymer Forcefields with Generalizable\nEquivariant Transformers. PolyGET is designed to capture complex quantum\ninteractions between atoms and generalize across various polymer families,\nusing a deep learning model called Equivariant Transformers. We propose a new\ntraining paradigm that focuses exclusively on optimizing forces, which is\ndifferent from existing methods that jointly optimize forces and energy. This\nsimple force-centric objective function avoids competing objectives between\nenergy and forces, thereby allowing for learning a unified forcefield ML model\nover different polymer families. We evaluated PolyGET on a large-scale dataset\nof 24 distinct polymer types and demonstrated state-of-the-art performance in\nforce accuracy and robust MD simulations. Furthermore, PolyGET can simulate\nlarge polymers with high fidelity to the reference ab initio DFT method while\nbeing able to generalize to unseen polymers.\n", "date": "2023-09-01", "categories": []}
{"id": "2210.17484", "title": "The Open MatSci ML Toolkit: A Flexible Framework for Machine Learning in\n  Materials Science", "abstract": "  We present the Open MatSci ML Toolkit: a flexible, self-contained, and\nscalable Python-based framework to apply deep learning models and methods on\nscientific data with a specific focus on materials science and the OpenCatalyst\nDataset. Our toolkit provides: 1. A scalable machine learning workflow for\nmaterials science leveraging PyTorch Lightning, which enables seamless scaling\nacross different computation capabilities (laptop, server, cluster) and\nhardware platforms (CPU, GPU, XPU). 2. Deep Graph Library (DGL) support for\nrapid graph neural network prototyping and development. By publishing and\nsharing this toolkit with the research community via open-source release, we\nhope to: 1. Lower the entry barrier for new machine learning researchers and\npractitioners that want to get started with the OpenCatalyst dataset, which\npresently comprises the largest computational materials science dataset. 2.\nEnable the scientific community to apply advanced machine learning tools to\nhigh-impact scientific challenges, such as modeling of materials behavior for\nclean energy applications. We demonstrate the capabilities of our framework by\nenabling three new equivariant neural network models for multiple OpenCatalyst\ntasks and arrive at promising results for compute scaling and model\nperformance.\n", "date": "2022-10-31", "categories": []}
{"id": "2308.16886", "title": "Prediction of Diblock Copolymer Morphology via Machine Learning", "abstract": "  A machine learning approach is presented to accelerate the computation of\nblock polymer morphology evolution for large domains over long timescales. The\nstrategy exploits the separation of characteristic times between coarse-grained\nparticle evolution on the monomer scale and slow morphological evolution over\nmesoscopic scales. In contrast to empirical continuum models, the proposed\napproach learns stochastically driven defect annihilation processes directly\nfrom particle-based simulations. A UNet architecture that respects different\nboundary conditions is adopted, thereby allowing periodic and fixed substrate\nboundary conditions of arbitrary shape. Physical concepts are also introduced\nvia the loss function and symmetries are incorporated via data augmentation.\nThe model is validated using three different use cases. Explainable artificial\nintelligence methods are applied to visualize the morphology evolution over\ntime. This approach enables the generation of large system sizes and long\ntrajectories to investigate defect densities and their evolution under\ndifferent types of confinement. As an application, we demonstrate the\nimportance of accessing late-stage morphologies for understanding particle\ndiffusion inside a single block. This work has implications for directed\nself-assembly and materials design in micro-electronics, battery materials, and\nmembranes.\n", "date": "2023-08-31", "categories": []}
{"id": "2308.13687", "title": "1.5 million materials narratives generated by chatbots", "abstract": "  The advent of artificial intelligence (AI) has enabled a comprehensive\nexploration of materials for various applications. However, AI models often\nprioritize frequently encountered materials in the scientific literature,\nlimiting the selection of suitable candidates based on inherent physical and\nchemical properties. To address this imbalance, we have generated a dataset of\n1,494,017 natural language-material paragraphs based on combined OQMD,\nMaterials Project, JARVIS, COD and AFLOW2 databases, which are dominated by ab\ninitio calculations and tend to be much more evenly distributed on the periodic\ntable. The generated text narratives were then polled and scored by both human\nexperts and ChatGPT-4, based on three rubrics: technical accuracy, language and\nstructure, and relevance and depth of content, showing similar scores but with\nhuman-scored depth of content being the most lagging. The merger of\nmulti-modality data sources and large language model (LLM) holds immense\npotential for AI frameworks to help the exploration and discovery of\nsolid-state materials for specific applications.\n", "date": "2023-08-25", "categories": []}
{"id": "2308.01423", "title": "ChatMOF: An Autonomous AI System for Predicting and Generating\n  Metal-Organic Frameworks", "abstract": "  ChatMOF is an autonomous Artificial Intelligence (AI) system that is built to\npredict and generate metal-organic frameworks (MOFs). By leveraging a\nlarge-scale language model (GPT-4 and GPT-3.5-turbo), ChatMOF extracts key\ndetails from textual inputs and delivers appropriate responses, thus\neliminating the necessity for rigid structured queries. The system is comprised\nof three core components (i.e. an agent, a toolkit, and an evaluator) and it\nforms a robust pipeline that manages a variety of tasks, including data\nretrieval, property prediction, and structure generations. The study further\nexplores the merits and constraints of using large language models (LLMs) AI\nsystem in material sciences using and showcases its transformative potential\nfor future advancements.\n", "date": "2023-08-01", "categories": []}
{"id": "2207.03928", "title": "Accelerating Material Design with the Generative Toolkit for Scientific\n  Discovery", "abstract": "  With the growing availability of data within various scientific domains,\ngenerative models hold enormous potential to accelerate scientific discovery.\nThey harness powerful representations learned from datasets to speed up the\nformulation of novel hypotheses with the potential to impact material discovery\nbroadly. We present the Generative Toolkit for Scientific Discovery (GT4SD).\nThis extensible open-source library enables scientists, developers, and\nresearchers to train and use state-of-the-art generative models to accelerate\nscientific discovery focused on material design.\n", "date": "2022-07-08", "categories": []}
{"id": "2308.11080", "title": "Stress representations for tensor basis neural networks: alternative\n  formulations to Finger-Rivlin-Ericksen", "abstract": "  Data-driven constitutive modeling frameworks based on neural networks and\nclassical representation theorems have recently gained considerable attention\ndue to their ability to easily incorporate constitutive constraints and their\nexcellent generalization performance. In these models, the stress prediction\nfollows from a linear combination of invariant-dependent coefficient functions\nand known tensor basis generators. However, thus far the formulations have been\nlimited to stress representations based on the classical Rivlin and Ericksen\nform, while the performance of alternative representations has yet to be\ninvestigated. In this work, we survey a variety of tensor basis neural network\nmodels for modeling hyperelastic materials in a finite deformation context,\nincluding a number of so far unexplored formulations which use theoretically\nequivalent invariants and generators to Finger-Rivlin-Ericksen. Furthermore, we\ncompare potential-based and coefficient-based approaches, as well as different\ncalibration techniques. Nine variants are tested against both noisy and\nnoiseless datasets for three different materials. Theoretical and practical\ninsights into the performance of each formulation are given.\n", "date": "2023-08-21", "categories": []}
{"id": "2308.10818", "title": "Interpretable Ensemble Learning for Materials Property Prediction with\n  Classical Interatomic Potentials: Carbon as an Example", "abstract": "  Machine learning (ML) is widely used to explore crystal materials and predict\ntheir properties. However, the training is time-consuming for deep-learning\nmodels, and the regression process is a black box that is hard to interpret.\nAlso, the preprocess to transfer a crystal structure into the input of ML,\ncalled descriptor, needs to be designed carefully. To efficiently predict\nimportant properties of materials, we propose an approach based on ensemble\nlearning consisting of regression trees to predict formation energy and elastic\nconstants based on small-size datasets of carbon allotropes as an example.\nWithout using any descriptor, the inputs are the properties calculated by\nmolecular dynamics with 9 different classical interatomic potentials. Overall,\nthe results from ensemble learning are more accurate than those from classical\ninteratomic potentials, and ensemble learning can capture the relatively\naccurate properties from the 9 classical potentials as criteria for predicting\nthe final properties.\n", "date": "2023-07-24", "categories": []}
{"id": "2308.08934", "title": "On Data Imbalance in Molecular Property Prediction with Pre-training", "abstract": "  Revealing and analyzing the various properties of materials is an essential\nand critical issue in the development of materials, including batteries,\nsemiconductors, catalysts, and pharmaceuticals. Traditionally, these properties\nhave been determined through theoretical calculations and simulations. However,\nit is not practical to perform such calculations on every single candidate\nmaterial. Recently, a combination method of the theoretical calculation and\nmachine learning has emerged, that involves training machine learning models on\na subset of theoretical calculation results to construct a surrogate model that\ncan be applied to the remaining materials. On the other hand, a technique\ncalled pre-training is used to improve the accuracy of machine learning models.\nPre-training involves training the model on pretext task, which is different\nfrom the target task, before training the model on the target task. This\nprocess aims to extract the input data features, stabilizing the learning\nprocess and improving its accuracy. However, in the case of molecular property\nprediction, there is a strong imbalance in the distribution of input data and\nfeatures, which may lead to biased learning towards frequently occurring data\nduring pre-training. In this study, we propose an effective pre-training method\nthat addresses the imbalance in input data. We aim to improve the final\naccuracy by modifying the loss function of the existing representative\npre-training method, node masking, to compensate the imbalance. We have\ninvestigated and assessed the impact of our proposed imbalance compensation on\npre-training and the final prediction accuracy through experiments and\nevaluations using benchmark of molecular property prediction models.\n", "date": "2023-08-17", "categories": []}
{"id": "2308.08129", "title": "Is Self-Supervised Pretraining Good for Extrapolation in Molecular\n  Property Prediction?", "abstract": "  The prediction of material properties plays a crucial role in the development\nand discovery of materials in diverse applications, such as batteries,\nsemiconductors, catalysts, and pharmaceuticals. Recently, there has been a\ngrowing interest in employing data-driven approaches by using machine learning\ntechnologies, in combination with conventional theoretical calculations. In\nmaterial science, the prediction of unobserved values, commonly referred to as\nextrapolation, is particularly critical for property prediction as it enables\nresearchers to gain insight into materials beyond the limits of available data.\nHowever, even with the recent advancements in powerful machine learning models,\naccurate extrapolation is still widely recognized as a significantly\nchallenging problem. On the other hand, self-supervised pretraining is a\nmachine learning technique where a model is first trained on unlabeled data\nusing relatively simple pretext tasks before being trained on labeled data for\ntarget tasks. As self-supervised pretraining can effectively utilize material\ndata without observed property values, it has the potential to improve the\nmodel's extrapolation ability. In this paper, we clarify how such\nself-supervised pretraining can enhance extrapolation performance.We propose an\nexperimental framework for the demonstration and empirically reveal that while\nmodels were unable to accurately extrapolate absolute property values,\nself-supervised pretraining enables them to learn relative tendencies of\nunobserved property values and improve extrapolation performance.\n", "date": "2023-08-16", "categories": []}
{"id": "2212.11317", "title": "End-to-end AI framework for interpretable prediction of molecular and\n  crystal properties", "abstract": "  We introduce an end-to-end computational framework that allows for\nhyperparameter optimization using the DeepHyper library, accelerated model\ntraining, and interpretable AI inference. The framework is based on\nstate-of-the-art AI models including CGCNN, PhysNet, SchNet, MPNN,\nMPNN-transformer, and TorchMD-NET. We employ these AI models along with the\nbenchmark QM9, hMOF, and MD17 datasets to showcase how the models can predict\nuser-specified material properties within modern computing environments. We\ndemonstrate transferable applications in the modeling of small molecules,\ninorganic crystals and nanoporous metal organic frameworks with a unified,\nstandalone framework. We have deployed and tested this framework in the\nThetaGPU supercomputer at the Argonne Leadership Computing Facility, and in the\nDelta supercomputer at the National Center for Supercomputing Applications to\nprovide researchers with modern tools to conduct accelerated AI-driven\ndiscovery in leadership-class computing environments. We release these digital\nassets as open source scientific software in GitLab, and ready-to-use Jupyter\nnotebooks in Google Colab.\n", "date": "2022-12-21", "categories": []}
{"id": "2305.11122", "title": "Autonomous sputter synthesis of thin film nitrides with composition\n  controlled by Bayesian optimization of optical plasma emission", "abstract": "  Autonomous experimentation has emerged as an efficient approach to accelerate\nthe pace of materials discovery. Although instruments for autonomous synthesis\nhave become popular in molecular and polymer science, solution processing of\nhybrid materials and nanoparticles, examples of autonomous tools for physical\nvapor deposition are scarce yet important for the semiconductor industry. Here,\nwe report the design and implementation of an autonomous workflow for sputter\ndeposition of thin films with controlled composition, leveraging a highly\nautomated sputtering reactor custom-controlled by Python, optical emission\nspectroscopy (OES), and a Bayesian optimization algorithm. We modeled film\ncomposition, measured by x-ray fluorescence, as a linear function of emission\nlines monitored during the co-sputtering from elemental Zn and Ti targets in\nN$_2$ atmosphere. A Bayesian control algorithm, informed by OES, navigates the\nspace of sputtering power to fabricate films with user-defined composition, by\nminimizing the absolute error between desired and measured emission signals. We\nvalidated our approach by autonomously fabricating Zn$_x$Ti$_{1-x}$N$_y$ films\nwith deviations from the targeted cation composition within relative 3.5 %,\neven for 15 nm thin films, demonstrating that the proposed approach can\nreliably synthesize thin films with specific composition and minimal human\ninterference. Moreover, the proposed method can be extended to more difficult\nsynthesis experiments where plasma intensity depends non-linearly on pressure,\nor the elemental sticking coefficients strongly depend on the substrate\ntemperature.\n", "date": "2023-05-18", "categories": []}
{"id": "2210.10507", "title": "Predicting Oxide Glass Properties with Low Complexity Neural Network and\n  Physical and Chemical Descriptors", "abstract": "  Due to their disordered structure, glasses present a unique challenge in\npredicting the composition-property relationships. Recently, several attempts\nhave been made to predict the glass properties using machine learning\ntechniques. However, these techniques have the limitations, namely, (i)\npredictions are limited to the components that are present in the original\ndataset, and (ii) predictions towards the extreme values of the properties,\nimportant regions for new materials discovery, are not very reliable due to the\nsparse datapoints in this region. To address these challenges, here we present\na low complexity neural network (LCNN) that provides improved performance in\npredicting the properties of oxide glasses. In addition, we combine the LCNN\nwith physical and chemical descriptors that allow the development of universal\nmodels that can provide predictions for components beyond the training set. By\ntraining on a large dataset (~50000) of glass components, we show the LCNN\noutperforms state-of-the-art algorithms such as XGBoost. In addition, we\ninterpret the LCNN models using Shapely additive explanations to gain insights\ninto the role played by the descriptors in governing the property. Finally, we\ndemonstrate the universality of the LCNN models by predicting the properties\nfor glasses with new components that were not present in the original training\nset. Altogether, the present approach provides a promising direction towards\naccelerated discovery of novel glass compositions.\n", "date": "2022-10-19", "categories": []}
{"id": "2308.03915", "title": "Predicting and explaining nonlinear material response using deep\n  Physically Guided Neural Networks with Internal Variables", "abstract": "  Nonlinear materials are often difficult to model with classical state model\ntheory because they have a complex and sometimes inaccurate physical and\nmathematical description or we simply do not know how to describe such\nmaterials in terms of relations between external and internal variables. In\nmany disciplines, Neural Network methods have arisen as powerful tools to\nidentify very complex and non-linear correlations. In this work, we use the\nvery recently developed concept of Physically Guided Neural Networks with\nInternal Variables (PGNNIV) to discover constitutive laws using a model-free\napproach and training solely with measured force-displacement data. PGNNIVs\nmake a particular use of the physics of the problem to enforce constraints on\nspecific hidden layers and are able to make predictions without internal\nvariable data. We demonstrate that PGNNIVs are capable of predicting both\ninternal and external variables under unseen load scenarios, regardless of the\nnature of the material considered (linear, with hardening or softening behavior\nand hyperelastic), unravelling the constitutive law of the material hence\nexplaining its nature altogether, placing the method in what is known as\neXplainable Artificial Intelligence (XAI).\n", "date": "2023-08-07", "categories": []}
{"id": "2308.01932", "title": "Investigation on Machine Learning Based Approaches for Estimating the\n  Critical Temperature of Superconductors", "abstract": "  Superconductors have been among the most fascinating substances, as the\nfundamental concept of superconductivity as well as the correlation of critical\ntemperature and superconductive materials have been the focus of extensive\ninvestigation since their discovery. However, superconductors at normal\ntemperatures have yet to be identified. Additionally, there are still many\nunknown factors and gaps of understanding regarding this unique phenomenon,\nparticularly the connection between superconductivity and the fundamental\ncriteria to estimate the critical temperature. To bridge the gap, numerous\nmachine learning techniques have been established to estimate critical\ntemperatures as it is extremely challenging to determine. Furthermore, the need\nfor a sophisticated and feasible method for determining the temperature range\nthat goes beyond the scope of the standard empirical formula appears to be\nstrongly emphasized by various machine-learning approaches. This paper uses a\nstacking machine learning approach to train itself on the complex\ncharacteristics of superconductive materials in order to accurately predict\ncritical temperatures. In comparison to other previous accessible research\ninvestigations, this model demonstrated a promising performance with an RMSE of\n9.68 and an R2 score of 0.922. The findings presented here could be a viable\ntechnique to shed new insight on the efficient implementation of the stacking\nensemble method with hyperparameter optimization (HPO).\n", "date": "2023-08-02", "categories": []}
{"id": "2308.02165", "title": "Diffusion probabilistic models enhance variational autoencoder for\n  crystal structure generative modeling", "abstract": "  The crystal diffusion variational autoencoder (CDVAE) is a machine learning\nmodel that leverages score matching to generate realistic crystal structures\nthat preserve crystal symmetry. In this study, we leverage novel diffusion\nprobabilistic (DP) models to denoise atomic coordinates rather than adopting\nthe standard score matching approach in CDVAE. Our proposed DP-CDVAE model can\nreconstruct and generate crystal structures whose qualities are statistically\ncomparable to those of the original CDVAE. Furthermore, notably, when comparing\nthe carbon structures generated by the DP-CDVAE model with relaxed structures\nobtained from density functional theory calculations, we find that the DP-CDVAE\ngenerated structures are remarkably closer to their respective ground states.\nThe energy differences between these structures and the true ground states are,\non average, 68.1 meV/atom lower than those generated by the original CDVAE.\nThis significant improvement in the energy accuracy highlights the\neffectiveness of the DP-CDVAE model in generating crystal structures that\nbetter represent their ground-state configurations.\n", "date": "2023-08-04", "categories": []}
{"id": "2303.12360", "title": "Automatically Predict Material Properties with Microscopic Image Example\n  Polymer Compatibility", "abstract": "  Many material properties are manifested in the morphological appearance and\ncharacterized with microscopic image, such as scanning electron microscopy\n(SEM). Polymer miscibility is a key physical quantity of polymer material and\ncommonly and intuitively judged by SEM images. However, human observation and\njudgement for the images is time-consuming, labor-intensive and hard to be\nquantified. Computer image recognition with machine learning method can make up\nthe defects of artificial judging, giving accurate and quantitative judgement.\nWe achieve automatic miscibility recognition utilizing convolution neural\nnetwork and transfer learning method, and the model obtains up to 94% accuracy.\nWe also put forward a quantitative criterion for polymer miscibility with this\nmodel. The proposed method can be widely applied to the quantitative\ncharacterization of the microstructure and properties of various materials.\n", "date": "2023-03-22", "categories": []}
{"id": "2212.07959", "title": "Scalable Bayesian Uncertainty Quantification for Neural Network\n  Potentials: Promise and Pitfalls", "abstract": "  Neural network (NN) potentials promise highly accurate molecular dynamics\n(MD) simulations within the computational complexity of classical MD force\nfields. However, when applied outside their training domain, NN potential\npredictions can be inaccurate, increasing the need for Uncertainty\nQuantification (UQ). Bayesian modeling provides the mathematical framework for\nUQ, but classical Bayesian methods based on Markov chain Monte Carlo (MCMC) are\ncomputationally intractable for NN potentials. By training graph NN potentials\nfor coarse-grained systems of liquid water and alanine dipeptide, we\ndemonstrate here that scalable Bayesian UQ via stochastic gradient MCMC\n(SG-MCMC) yields reliable uncertainty estimates for MD observables. We show\nthat cold posteriors can reduce the required training data size and that for\nreliable UQ, multiple Markov chains are needed. Additionally, we find that\nSG-MCMC and the Deep Ensemble method achieve comparable results, despite\nshorter training and less hyperparameter tuning of the latter. We show that\nboth methods can capture aleatoric and epistemic uncertainty reliably, but not\nsystematic uncertainty, which needs to be minimized by adequate modeling to\nobtain accurate credible intervals for MD observables. Our results represent a\nstep towards accurate UQ that is of vital importance for trustworthy NN\npotential-based MD simulations required for decision-making in practice.\n", "date": "2022-12-15", "categories": []}
{"id": "2210.01705", "title": "Tensor-reduced atomic density representations", "abstract": "  Density based representations of atomic environments that are invariant under\nEuclidean symmetries have become a widely used tool in the machine learning of\ninteratomic potentials, broader data-driven atomistic modelling and the\nvisualisation and analysis of materials datasets.The standard mechanism used to\nincorporate chemical element information is to create separate densities for\neach element and form tensor products between them. This leads to a steep\nscaling in the size of the representation as the number of elements increases.\nGraph neural networks, which do not explicitly use density representations,\nescape this scaling by mapping the chemical element information into a fixed\ndimensional space in a learnable way. We recast this approach as tensor\nfactorisation by exploiting the tensor structure of standard neighbour density\nbased descriptors. In doing so, we form compact tensor-reduced representations\nwhose size does not depend on the number of chemical elements, but remain\nsystematically convergeable and are therefore applicable to a wide range of\ndata analysis and regression tasks.\n", "date": "2022-10-02", "categories": []}
{"id": "2307.13581", "title": "Comparing Forward and Inverse Design Paradigms: A Case Study on\n  Refractory High-Entropy Alloys", "abstract": "  The rapid design of advanced materials is a topic of great scientific\ninterest. The conventional, ``forward'' paradigm of materials design involves\nevaluating multiple candidates to determine the best candidate that matches the\ntarget properties. However, recent advances in the field of deep learning have\ngiven rise to the possibility of an ``inverse'' design paradigm for advanced\nmaterials, wherein a model provided with the target properties is able to find\nthe best candidate. Being a relatively new concept, there remains a need to\nsystematically evaluate how these two paradigms perform in practical\napplications. Therefore, the objective of this study is to directly,\nquantitatively compare the forward and inverse design modeling paradigms. We do\nso by considering two case studies of refractory high-entropy alloy design with\ndifferent objectives and constraints and comparing the inverse design method to\nother forward schemes like localized forward search, high throughput screening,\nand multi objective optimization.\n", "date": "2023-07-25", "categories": []}
{"id": "2302.13380", "title": "Closed-loop Error Correction Learning Accelerates Experimental Discovery\n  of Thermoelectric Materials", "abstract": "  The exploration of thermoelectric materials is challenging considering the\nlarge materials space, combined with added exponential degrees of freedom\ncoming from doping and the diversity of synthetic pathways. Here we seek to\nincorporate historical data and update and refine it using experimental\nfeedback by employing error-correction learning (ECL). We thus learn from prior\ndatasets and then adapt the model to differences in synthesis and\ncharacterization that are otherwise difficult to parameterize. We then apply\nthis strategy to discovering thermoelectric materials where we prioritize\nsynthesis at temperatures < 300{\\deg}C. We document a previously unreported\nchemical family of thermoelectric materials, PbSe:SnSb, finding that the best\ncandidate in this chemical family, 2 wt% SnSb doped PbSe, exhibits a power\nfactor more than 2x that of PbSe. Our investigations show that our closed-loop\nexperimentation strategy reduces the required number of experiments to find an\noptimized material by as much as 3x compared to high-throughput searches\npowered by state-of-the-art machine learning models. We also observe that this\nimprovement is dependent on the accuracy of prior in a manner that exhibits\ndiminishing returns, and after a certain accuracy is reached, it is factors\nassociated with experimental pathways that dictate the trends.\n", "date": "2023-02-26", "categories": []}
{"id": "2209.14803", "title": "polyBERT: A chemical language model to enable fully machine-driven\n  ultrafast polymer informatics", "abstract": "  Polymers are a vital part of everyday life. Their chemical universe is so\nlarge that it presents unprecedented opportunities as well as significant\nchallenges to identify suitable application-specific candidates. We present a\ncomplete end-to-end machine-driven polymer informatics pipeline that can search\nthis space for suitable candidates at unprecedented speed and accuracy. This\npipeline includes a polymer chemical fingerprinting capability called polyBERT\n(inspired by Natural Language Processing concepts), and a multitask learning\napproach that maps the polyBERT fingerprints to a host of properties. polyBERT\nis a chemical linguist that treats the chemical structure of polymers as a\nchemical language. The present approach outstrips the best presently available\nconcepts for polymer property prediction based on handcrafted fingerprint\nschemes in speed by two orders of magnitude while preserving accuracy, thus\nmaking it a strong candidate for deployment in scalable architectures including\ncloud infrastructures.\n", "date": "2022-09-29", "categories": []}
{"id": "2303.03493", "title": "Machine learning for phase ordering dynamics of charge density waves", "abstract": "  We present a machine learning (ML) framework for large-scale dynamical\nsimulations of charge density wave (CDW) states. The charge modulation in a CDW\nstate is often accompanied by a concomitant structural distortion, and the\nadiabatic evolution of a CDW order is governed by the dynamics of the lattice\ndistortion. Calculation of the electronic contribution to the driving forces,\nhowever, is computationally very expensive for large systems. Assuming the\nprinciple of locality for electron systems, a neural-network model is developed\nto accurately and efficiently predict local electronic forces with input from\nneighborhood configurations. Importantly, the ML model makes possible a linear\ncomplexity algorithm for dynamical simulations of CDWs. As a demonstration, we\napply our approach to investigate the phase ordering dynamics of the Holstein\nmodel, a canonical system of CDW order. Our large-scale simulations uncover an\nintriguing growth of the CDW domains that deviates significantly from the\nexpected Allen-Cahn law for phase ordering of Ising-type order parameter field.\nThis anomalous domain-growth could be attributed to the complex structure of\ndomain-walls in this system. Our work highlights the promising potential of\nML-based force-field models for dynamical simulations of functional electronic\nmaterials.\n", "date": "2023-03-06", "categories": []}
{"id": "2307.07912", "title": "Predicting mechanical properties of Carbon Nanotube (CNT) images Using\n  Multi-Layer Synthetic Finite Element Model Simulations", "abstract": "  We present a pipeline for predicting mechanical properties of\nvertically-oriented carbon nanotube (CNT) forest images using a deep learning\nmodel for artificial intelligence (AI)-based materials discovery. Our approach\nincorporates an innovative data augmentation technique that involves the use of\nmulti-layer synthetic (MLS) or quasi-2.5D images which are generated by\nblending 2D synthetic images. The MLS images more closely resemble 3D synthetic\nand real scanning electron microscopy (SEM) images of CNTs but without the\ncomputational cost of performing expensive 3D simulations or experiments.\nMechanical properties such as stiffness and buckling load for the MLS images\nare estimated using a physics-based model. The proposed deep learning\narchitecture, CNTNeXt, builds upon our previous CNTNet neural network, using a\nResNeXt feature representation followed by random forest regression estimator.\nOur machine learning approach for predicting CNT physical properties by\nutilizing a blended set of synthetic images is expected to outperform single\nsynthetic image-based learning when it comes to predicting mechanical\nproperties of real scanning electron microscopy images. This has the potential\nto accelerate understanding and control of CNT forest self-assembly for diverse\napplications.\n", "date": "2023-07-16", "categories": []}
{"id": "2307.05521", "title": "Toward High-Performance Energy and Power Battery Cells with Machine\n  Learning-based Optimization of Electrode Manufacturing", "abstract": "  The optimization of the electrode manufacturing process is important for\nupscaling the application of Lithium Ion Batteries (LIBs) to cater for growing\nenergy demand. In particular, LIB manufacturing is very important to be\noptimized because it determines the practical performance of the cells when the\nlatter are being used in applications such as electric vehicles. In this study,\nwe tackled the issue of high-performance electrodes for desired battery\napplication conditions by proposing a powerful data-driven approach supported\nby a deterministic machine learning (ML)-assisted pipeline for bi-objective\noptimization of the electrochemical performance. This ML pipeline allows the\ninverse design of the process parameters to adopt in order to manufacture\nelectrodes for energy or power applications. The latter work is an analogy to\nour previous work that supported the optimization of the electrode\nmicrostructures for kinetic, ionic, and electronic transport properties\nimprovement. An electrochemical pseudo-two-dimensional model is fed with the\nelectrode properties characterizing the electrode microstructures generated by\nmanufacturing simulations and used to simulate the electrochemical\nperformances. Secondly, the resulting dataset was used to train a deterministic\nML model to implement fast bi-objective optimizations to identify optimal\nelectrodes. Our results suggested a high amount of active material, combined\nwith intermediate values of solid content in the slurry and calendering degree,\nto achieve the optimal electrodes.\n", "date": "2023-07-07", "categories": []}
{"id": "2307.05378", "title": "M$^2$Hub: Unlocking the Potential of Machine Learning for Materials\n  Discovery", "abstract": "  We introduce M$^2$Hub, a toolkit for advancing machine learning in materials\ndiscovery. Machine learning has achieved remarkable progress in modeling\nmolecular structures, especially biomolecules for drug discovery. However, the\ndevelopment of machine learning approaches for modeling materials structures\nlag behind, which is partly due to the lack of an integrated platform that\nenables access to diverse tasks for materials discovery. To bridge this gap,\nM$^2$Hub will enable easy access to materials discovery tasks, datasets,\nmachine learning methods, evaluations, and benchmark results that cover the\nentire workflow. Specifically, the first release of M$^2$Hub focuses on three\nkey stages in materials discovery: virtual screening, inverse design, and\nmolecular simulation, including 9 datasets that covers 6 types of materials\nwith 56 tasks across 8 types of material properties. We further provide 2\nsynthetic datasets for the purpose of generative tasks on materials. In\naddition to random data splits, we also provide 3 additional data partitions to\nreflect the real-world materials discovery scenarios. State-of-the-art machine\nlearning methods (including those are suitable for materials structures but\nnever compared in the literature) are benchmarked on representative tasks. Our\ncodes and library are publicly available at https://github.com/yuanqidu/M2Hub.\n", "date": "2023-06-14", "categories": []}
{"id": "2307.05380", "title": "Optimized Crystallographic Graph Generation for Material Science", "abstract": "  Graph neural networks are widely used in machine learning applied to\nchemistry, and in particular for material science discovery. For crystalline\nmaterials, however, generating graph-based representation from geometrical\ninformation for neural networks is not a trivial task. The periodicity of\ncrystalline needs efficient implementations to be processed in real-time under\na massively parallel environment. With the aim of training graph-based\ngenerative models of new material discovery, we propose an efficient tool to\ngenerate cutoff graphs and k-nearest-neighbours graphs of periodic structures\nwithin GPU optimization. We provide pyMatGraph a Pytorch-compatible framework\nto generate graphs in real-time during the training of neural network\narchitecture. Our tool can update a graph of a structure, making generative\nmodels able to update the geometry and process the updated graph during the\nforward propagation on the GPU side. Our code is publicly available at\nhttps://github.com/aklipf/mat-graph.\n", "date": "2023-06-07", "categories": []}
{"id": "2307.05390", "title": "CrysMMNet: Multimodal Representation for Crystal Property Prediction", "abstract": "  Machine Learning models have emerged as a powerful tool for fast and accurate\nprediction of different crystalline properties. Exiting state-of-the-art models\nrely on a single modality of crystal data i.e. crystal graph structure, where\nthey construct multi-graph by establishing edges between nearby atoms in 3D\nspace and apply GNN to learn materials representation. Thereby, they encode\nlocal chemical semantics around the atoms successfully but fail to capture\nimportant global periodic structural information like space group number,\ncrystal symmetry, rotational information, etc, which influence different\ncrystal properties. In this work, we leverage textual descriptions of materials\nto model global structural information into graph structure and learn a more\nrobust and enriched representation of crystalline materials. To this effect, we\nfirst curate a textual dataset for crystalline material databases containing\ndescriptions of each material. Further, we propose CrysMMNet, a simple\nmulti-modal framework, which fuses both structural and textual representation\ntogether to generate a joint multimodal representation of crystalline\nmaterials. We conduct extensive experiments on two benchmark datasets across\nten different properties to show that CrysMMNet outperforms existing\nstate-of-the-art baseline methods with a good margin. We also observe that\nfusing the textual representation with crystal graph structure provides\nconsistent improvement for all the SOTA GNN models compared to their own\nvanilla versions. We have shared the textual dataset, that we have curated for\nboth the benchmark material databases, with the community for future use.\n", "date": "2023-06-09", "categories": []}
{"id": "2307.04351", "title": "MD-HIT: Machine learning for materials property prediction with dataset\n  redundancy control", "abstract": "  Materials datasets are usually featured by the existence of many redundant\n(highly similar) materials due to the tinkering material design practice over\nthe history of materials research. For example, the materials project database\nhas many perovskite cubic structure materials similar to SrTiO$_3$. This sample\nredundancy within the dataset makes the random splitting of machine learning\nmodel evaluation to fail so that the ML models tend to achieve over-estimated\npredictive performance which is misleading for the materials science community.\nThis issue is well known in the field of bioinformatics for protein function\nprediction, in which a redundancy reduction procedure (CD-Hit) is always\napplied to reduce the sample redundancy by ensuring no pair of samples has a\nsequence similarity greater than a given threshold. This paper surveys the\noverestimated ML performance in the literature for both composition based and\nstructure based material property prediction. We then propose a material\ndataset redundancy reduction algorithm called MD-HIT and evaluate it with\nseveral composition and structure based distance threshold sfor reducing data\nset sample redundancy. We show that with this control, the predicted\nperformance tends to better reflect their true prediction capability. Our\nMD-hit code can be freely accessed at https://github.com/usccolumbia/MD-HIT\n", "date": "2023-07-10", "categories": []}
{"id": "2208.09273", "title": "Atomistic structure search using local surrogate mode", "abstract": "  We describe a local surrogate model for use in conjunction with global\nstructure search methods. The model follows the Gaussian approximation\npotential (GAP) formalism and is based on a the smooth overlap of atomic\npositions descriptor with sparsification in terms of a reduced number of local\nenvironments using mini-batch $k$-means. The model is implemented in the\nAtomistic Global Optimization X framework and used as a partial replacement of\nthe local relaxations in basin hopping structure search. The approach is shown\nto be robust for a wide range of atomistic system including molecules,\nnano-particles, surface supported clusters and surface thin films. The benefits\nin a structure search context of a local surrogate model are demonstrated. This\nincludes the ability to transfer learning from smaller systems as well as the\npossibility to perform concurrent multi-stoichiometry searches.\n", "date": "2022-08-19", "categories": []}
{"id": "2212.10478", "title": "Machine Learning and Polymer Self-Consistent Field Theory in Two Spatial\n  Dimensions", "abstract": "  A computational framework that leverages data from self-consistent field\ntheory simulations with deep learning to accelerate the exploration of\nparameter space for block copolymers is presented. This is a substantial\ntwo-dimensional extension of the framework introduced in [1]. Several\ninnovations and improvements are proposed. (1) A Sobolev space-trained,\nconvolutional neural network (CNN) is employed to handle the exponential\ndimension increase of the discretized, local average monomer density fields and\nto strongly enforce both spatial translation and rotation invariance of the\npredicted, field-theoretic intensive Hamiltonian. (2) A generative adversarial\nnetwork (GAN) is introduced to efficiently and accurately predict saddle point,\nlocal average monomer density fields without resorting to gradient descent\nmethods that employ the training set. This GAN approach yields important\nsavings of both memory and computational cost. (3) The proposed machine\nlearning framework is successfully applied to 2D cell size optimization as a\nclear illustration of its broad potential to accelerate the exploration of\nparameter space for discovering polymer nanostructures. Extensions to\nthree-dimensional phase discovery appear to be feasible.\n", "date": "2022-12-16", "categories": []}
{"id": "2306.16270", "title": "S2SNet: A Pretrained Neural Network for Superconductivity Discovery", "abstract": "  Superconductivity allows electrical current to flow without any energy loss,\nand thus making solids superconducting is a grand goal of physics, material\nscience, and electrical engineering. More than 16 Nobel Laureates have been\nawarded for their contribution to superconductivity research. Superconductors\nare valuable for sustainable development goals (SDGs), such as climate change\nmitigation, affordable and clean energy, industry, innovation and\ninfrastructure, and so on. However, a unified physics theory explaining all\nsuperconductivity mechanism is still unknown. It is believed that\nsuperconductivity is microscopically due to not only molecular compositions but\nalso the geometric crystal structure. Hence a new dataset, S2S, containing both\ncrystal structures and superconducting critical temperature, is built upon\nSuperCon and Material Project. Based on this new dataset, we propose a novel\nmodel, S2SNet, which utilizes the attention mechanism for superconductivity\nprediction. To overcome the shortage of data, S2SNet is pre-trained on the\nwhole Material Project dataset with Masked-Language Modeling (MLM). S2SNet\nmakes a new state-of-the-art, with out-of-sample accuracy of 92% and Area Under\nCurve (AUC) of 0.92. To the best of our knowledge, S2SNet is the first work to\npredict superconductivity with only information of crystal structures. This\nwork is beneficial to superconductivity discovery and further SDGs. Code and\ndatasets are available in https://github.com/zjuKeLiu/S2SNet\n", "date": "2023-06-28", "categories": []}
{"id": "2306.10406", "title": "Human-In-the-Loop for Bayesian Autonomous Materials Phase Mapping", "abstract": "  Autonomous experimentation (AE) combines machine learning and research\nhardware automation in a closed loop, guiding subsequent experiments toward\nuser goals. As applied to materials research, AE can accelerate materials\nexploration, reducing time and cost compared to traditional Edisonian studies.\nAdditionally, integrating knowledge from diverse sources including theory,\nsimulations, literature, and domain experts can boost AE performance. Domain\nexperts may provide unique knowledge addressing tasks that are difficult to\nautomate. Here, we present a set of methods for integrating human input into an\nautonomous materials exploration campaign for composition-structure phase\nmapping. The methods are demonstrated on x-ray diffraction data collected from\na thin film ternary combinatorial library. At any point during the campaign,\nthe user can choose to provide input by indicating regions-of-interest, likely\nphase regions, and likely phase boundaries based on their prior knowledge\n(e.g., knowledge of the phase map of a similar material system), along with\nquantifying their certainty. The human input is integrated by defining a set of\nprobabilistic priors over the phase map. Algorithm output is a probabilistic\ndistribution over potential phase maps, given the data, model, and human input.\nWe demonstrate a significant improvement in phase mapping performance given\nappropriate human input.\n", "date": "2023-06-17", "categories": []}
{"id": "2306.10766", "title": "Substitutional Alloying Using Crystal Graph Neural Networks", "abstract": "  Materials discovery, especially for applications that require extreme\noperating conditions, requires extensive testing that naturally limits the\nability to inquire the wealth of possible compositions. Machine Learning (ML)\nhas nowadays a well established role in facilitating this effort in systematic\nways. The increasing amount of available accurate DFT data represents a solid\nbasis upon which new ML models can be trained and tested. While conventional\nmodels rely on static descriptors, generally suitable for a limited class of\nsystems, the flexibility of Graph Neural Networks (GNNs) allows for direct\nlearning representations on graphs, such as the ones formed by crystals. We\nutilize crystal graph neural networks (CGNN) to predict crystal properties with\nDFT level accuracy, through graphs with encoding of the atomic (node/vertex),\nbond (edge), and global state attributes. In this work, we aim at testing the\nability of the CGNN MegNet framework in predicting a number of properties of\nsystems previously unseen from the model, obtained by adding a substitutional\ndefect in bulk crystals that are included in the training set. We perform DFT\nvalidation to assess the accuracy in the prediction of formation energies and\nstructural features (such as elastic moduli). Using CGNNs, one may identify\npromising paths in alloy discovery.\n", "date": "2023-06-19", "categories": []}
{"id": "2302.12881", "title": "Denoising diffusion algorithm for inverse design of microstructures with\n  fine-tuned nonlinear material properties", "abstract": "  In this paper, we introduce a denoising diffusion algorithm to discover\nmicrostructures with nonlinear fine-tuned properties. Denoising diffusion\nprobabilistic models are generative models that use diffusion-based dynamics to\ngradually denoise images and generate realistic synthetic samples. By learning\nthe reverse of a Markov diffusion process, we design an artificial intelligence\nto efficiently manipulate the topology of microstructures to generate a massive\nnumber of prototypes that exhibit constitutive responses sufficiently close to\ndesignated nonlinear constitutive responses. To identify the subset of\nmicrostructures with sufficiently precise fine-tuned properties, a\nconvolutional neural network surrogate is trained to replace high-fidelity\nfinite element simulations to filter out prototypes outside the admissible\nrange. The results of this study indicate that the denoising diffusion process\nis capable of creating microstructures of fine-tuned nonlinear material\nproperties within the latent space of the training data. More importantly, the\nresulting algorithm can be easily extended to incorporate additional\ntopological and geometric modifications by introducing high-dimensional\nstructures embedded in the latent space. The algorithm is tested on the\nopen-source mechanical MNIST data set. Consequently, this algorithm is not only\ncapable of performing inverse design of nonlinear effective media but also\nlearns the nonlinear structure-property map to quantitatively understand the\nmultiscale interplay among the geometry and topology and their effective\nmacroscopic properties.\n", "date": "2023-02-24", "categories": []}
{"id": "2211.04796", "title": "Physics-separating artificial neural networks for predicting initial\n  stages of Al sputtering and thin film deposition in Ar plasma discharges", "abstract": "  Simulations of Al thin film sputter depositions rely on accurate plasma and\nsurface interaction models. Establishing the latter commonly requires a higher\nlevel of abstraction and means to dismiss the fundamental atomic fidelity.\nPrevious works on sputtering processes addressed this issue by establishing\nmachine learning surrogate models, which include a basic surface state (i.e.,\nstoichiometry) as static input. In this work, an evolving surface state and\ndefect structure are introduced to jointly describe sputtering and growth with\nphysics-separating artificial neural networks. The data describing the\nplasma-surface interactions stem from hybrid reactive molecular\ndynamics/time-stamped force bias Monte Carlo simulations of Al neutrals and\nAr$^+$ ions impinging onto Al(001) surfaces. It is demonstrated that the\nfundamental processes are comprehensively described by taking the surface state\nas well as defect structure into account. Hence, a machine learning\nplasma-surface interaction surrogate model is established that resolves the\ninherent kinetics with high physical fidelity. The resulting model is not\nrestricted to input from modeling and simulation, but may similarly be applied\nto experimental input data.\n", "date": "2022-11-09", "categories": []}
{"id": "2302.02303", "title": "Precursor recommendation for inorganic synthesis by machine learning\n  materials similarity from scientific literature", "abstract": "  Synthesis prediction is a key accelerator for the rapid design of advanced\nmaterials. However, determining synthesis variables such as the choice of\nprecursor materials is challenging for inorganic materials because the sequence\nof reactions during heating is not well understood. In this work, we use a\nknowledge base of 29,900 solid-state synthesis recipes, text-mined from the\nscientific literature, to automatically learn which precursors to recommend for\nthe synthesis of a novel target material. The data-driven approach learns\nchemical similarity of materials and refers the synthesis of a new target to\nprecedent synthesis procedures of similar materials, mimicking human synthesis\ndesign. When proposing five precursor sets for each of 2,654 unseen test target\nmaterials, the recommendation strategy achieves a success rate of at least 82%.\nOur approach captures decades of heuristic synthesis data in a mathematical\nform, making it accessible for use in recommendation engines and autonomous\nlaboratories.\n", "date": "2023-02-05", "categories": []}
{"id": "2306.05344", "title": "A Crystal-Specific Pre-Training Framework for Crystal Material Property\n  Prediction", "abstract": "  Crystal property prediction is a crucial aspect of developing novel\nmaterials. However, there are two technical challenges to be addressed for\nspeeding up the investigation of crystals. First, labeling crystal properties\nis intrinsically difficult due to the high cost and time involved in physical\nsimulations or lab experiments. Second, crystals adhere to a specific quantum\nchemical principle known as periodic invariance, which is often not captured by\nexisting machine learning methods. To overcome these challenges, we propose the\ncrystal-specific pre-training framework for learning crystal representations\nwith self-supervision. The framework designs a mutex mask strategy for\nenhancing representation learning so as to alleviate the limited labels\navailable for crystal property prediction. Moreover, we take into account the\nspecific periodic invariance in crystal structures by developing a periodic\ninvariance multi-graph module and periodic attribute learning within our\nframework. This framework has been tested on eight different tasks. The\nexperimental results on these tasks show that the framework achieves promising\nprediction performance and is able to outperform recent strong baselines.\n", "date": "2023-06-08", "categories": []}
{"id": "2306.04365", "title": "Edge conductivity in PtSe$_2$ nanostructures", "abstract": "  PtSe$_2$ is a promising 2D material for nanoelectromechanical sensing and\nphotodetection in the infrared regime. One of its most compelling features is\nthe facile synthesis at temperatures below 500 {\\deg}C, which is compatible\nwith current back-end-of-line semiconductor processing. However, this process\ngenerates polycrystalline thin films with nanoflake-like domains of 5 to 100 nm\nsize. To investigate the lateral quantum confinement effect in this size\nregime, we train a deep neural network to obtain an interatomic potential at\nDFT accuracy and use that to model ribbons, surfaces, nanoflakes, and\nnanoplatelets of PtSe$_2$ with lateral widths between 5 to 15 nm. We determine\nwhich edge terminations are the most stable and find evidence that the\nelectrical conductivity is localized on the edges for lateral sizes below 10\nnm. This suggests that the transport channels in thin films of PtSe$_2$ might\nbe dominated by networks of edges, instead of transport through the layers\nthemselves.\n", "date": "2023-06-07", "categories": []}
{"id": "2306.04510", "title": "Unified Model for Crystalline Material Generation", "abstract": "  One of the greatest challenges facing our society is the discovery of new\ninnovative crystal materials with specific properties. Recently, the problem of\ngenerating crystal materials has received increasing attention, however, it\nremains unclear to what extent, or in what way, we can develop generative\nmodels that consider both the periodicity and equivalence geometric of crystal\nstructures. To alleviate this issue, we propose two unified models that act at\nthe same time on crystal lattice and atomic positions using periodic\nequivariant architectures. Our models are capable to learn any arbitrary\ncrystal lattice deformation by lowering the total energy to reach thermodynamic\nstability. Code and data are available at https://github.com/aklipf/GemsNet.\n", "date": "2023-06-07", "categories": []}
{"id": "2306.03109", "title": "Machine Learning Force Fields with Data Cost Aware Training", "abstract": "  Machine learning force fields (MLFF) have been proposed to accelerate\nmolecular dynamics (MD) simulation, which finds widespread applications in\nchemistry and biomedical research. Even for the most data-efficient MLFFs,\nreaching chemical accuracy can require hundreds of frames of force and energy\nlabels generated by expensive quantum mechanical algorithms, which may scale as\n$O(n^3)$ to $O(n^7)$, with $n$ proportional to the number of basis functions.\nTo address this issue, we propose a multi-stage computational framework --\nASTEROID, which lowers the data cost of MLFFs by leveraging a combination of\ncheap inaccurate data and expensive accurate data. The motivation behind\nASTEROID is that inaccurate data, though incurring large bias, can help capture\nthe sophisticated structures of the underlying force field. Therefore, we first\ntrain a MLFF model on a large amount of inaccurate training data, employing a\nbias-aware loss function to prevent the model from overfitting tahe potential\nbias of this data. We then fine-tune the obtained model using a small amount of\naccurate training data, which preserves the knowledge learned from the\ninaccurate training data while significantly improving the model's accuracy.\nMoreover, we propose a variant of ASTEROID based on score matching for the\nsetting where the inaccurate training data are unlabeled. Extensive experiments\non MD datasets and downstream tasks validate the efficacy of ASTEROID. Our code\nand data are available at https://github.com/abukharin3/asteroid.\n", "date": "2023-06-05", "categories": []}
{"id": "2306.00797", "title": "Microstructure quality control of steels using deep learning", "abstract": "  In quality control, microstructures are investigated rigorously to ensure\nstructural integrity, exclude the presence of critical volume defects, and\nvalidate the formation of the target microstructure. For quenched,\nhierarchically-structured steels, the morphology of the bainitic and\nmartensitic microstructures are of major concern to guarantee the reliability\nof the material under service conditions. Therefore, industries conduct small\nsample-size inspections of materials cross-sections through metallographers to\nvalidate the needle morphology of such microstructures. We demonstrate\nround-robin test results revealing that this visual grading is afflicted by\npronounced subjectivity despite the thorough training of personnel. Instead, we\npropose a deep learning image classification approach that distinguishes steels\nbased on their microstructure type and classifies their needle length alluding\nto the ISO 643 grain size assessment standard. This classification approach\nfacilitates the reliable, objective, and automated classification of\nhierarchically structured steels. Specifically, an accuracy of 96% and roughly\n91% is attained for the distinction of martensite/bainite subtypes and needle\nlength, respectively. This is achieved on an image dataset that contains\nsignificant variance and labeling noise as it is acquired over more than ten\nyears from multiple plants, alloys, etchant applications, and light optical\nmicroscopes by many metallographers (raters). Interpretability analysis gives\ninsights into the decision-making of these models and allows for estimating\ntheir generalization capability.\n", "date": "2023-06-01", "categories": []}
{"id": "2206.08841", "title": "Random projections and Kernelised Leave One Cluster Out\n  Cross-Validation: Universal baselines and evaluation tools for supervised\n  machine learning for materials properties", "abstract": "  With machine learning being a popular topic in current computational\nmaterials science literature, creating representations for compounds has become\ncommon place. These representations are rarely compared, as evaluating their\nperformance - and the performance of the algorithms that they are used with -\nis non-trivial. With many materials datasets containing bias and skew caused by\nthe research process, leave one cluster out cross validation (LOCO-CV) has been\nintroduced as a way of measuring the performance of an algorithm in predicting\npreviously unseen groups of materials. This raises the question of the impact,\nand control, of the range of cluster sizes on the LOCO-CV measurement outcomes.\nWe present a thorough comparison between composition-based representations, and\ninvestigate how kernel approximation functions can be used to better separate\ndata to enhance LOCO-CV applications.\n  We find that domain knowledge does not improve machine learning performance\nin most tasks tested, with band gap prediction being the notable exception. We\nalso find that the radial basis function improves the linear separability of\nchemical datasets in all 10 datasets tested and provide a framework for the\napplication of this function in the LOCO-CV process to improve the outcome of\nLOCO-CV measurements regardless of machine learning algorithm, choice of\nmetric, and choice of compound representation. We recommend kernelised LOCO-CV\nas a training paradigm for those looking to measure the extrapolatory power of\nan algorithm on materials data.\n", "date": "2022-06-17", "categories": []}
{"id": "2305.15410", "title": "Machine learning-assisted close-set X-ray diffraction phase\n  identification of transition metals", "abstract": "  Machine learning has been applied to the problem of X-ray diffraction phase\nprediction with promising results. In this paper, we describe a method for\nusing machine learning to predict crystal structure phases from X-ray\ndiffraction data of transition metals and their oxides. We evaluate the\nperformance of our method and compare the variety of its settings. Our results\ndemonstrate that the proposed machine learning framework achieves competitive\nperformance. This demonstrates the potential for machine learning to\nsignificantly impact the field of X-ray diffraction and crystal structure\ndetermination. Open-source implementation:\nhttps://github.com/maxnygma/NeuralXRD.\n", "date": "2023-04-28", "categories": []}
{"id": "2209.01948", "title": "A smooth basis for atomistic machine learning", "abstract": "  Machine learning frameworks based on correlations of interatomic positions\nbegin with a discretized description of the density of other atoms in the\nneighbourhood of each atom in the system. Symmetry considerations support the\nuse of spherical harmonics to expand the angular dependence of this density,\nbut there is as yet no clear rationale to choose one radial basis over another.\nHere we investigate the basis that results from the solution of the Laplacian\neigenvalue problem within a sphere around the atom of interest. We show that\nthis generates the smoothest possible basis of a given size within the sphere,\nand that a tensor product of Laplacian eigenstates also provides the smoothest\npossible basis for expanding any higher-order correlation of the atomic density\nwithin the appropriate hypersphere. We consider several unsupervised metrics of\nthe quality of a basis for a given dataset, and show that the Laplacian\neigenstate basis has a performance that is much better than some widely used\nbasis sets and is competitive with data-driven bases that numerically optimize\neach metric. In supervised machine learning tests, we find that the optimal\nfunction smoothness of the Laplacian eigenstates leads to comparable or better\nperformance than can be obtained from a data-driven basis of a similar size\nthat has been optimized to describe the atom-density correlation for the\nspecific dataset. We conclude that the smoothness of the basis functions is a\nkey and hitherto largely overlooked aspect of successful atomic density\nrepresentations.\n", "date": "2022-09-05", "categories": []}
{"id": "2208.03296", "title": "Accelerating discrete dislocation dynamics simulations with graph neural\n  networks", "abstract": "  Discrete dislocation dynamics (DDD) is a widely employed computational method\nto study plasticity at the mesoscale that connects the motion of dislocation\nlines to the macroscopic response of crystalline materials. However, the\ncomputational cost of DDD simulations remains a bottleneck that limits its\nrange of applicability. Here, we introduce a new DDD-GNN framework in which the\nexpensive time-integration of dislocation motion is entirely substituted by a\ngraph neural network (GNN) model trained on DDD trajectories. As a first\napplication, we demonstrate the feasibility and potential of our method on a\nsimple yet relevant model of a dislocation line gliding through an array of\nobstacles. We show that the DDD-GNN model is stable and reproduces very well\nunseen ground-truth DDD simulation responses for a range of straining rates and\nobstacle densities, without the need to explicitly compute nodal forces or\ndislocation mobilities during time-integration. Our approach opens new\npromising avenues to accelerate DDD simulations and to incorporate more complex\ndislocation motion behaviors.\n", "date": "2022-08-05", "categories": []}
{"id": "2305.06925", "title": "Accurate Surface and Finite Temperature Bulk Properties of Lithium Metal\n  at Large Scales using Machine Learning Interaction Potentials", "abstract": "  The properties of lithium metal are key parameters in the design of lithium\nion and lithium metal batteries. They are difficult to probe experimentally due\nto the high reactivity and low melting point of lithium as well as the\nmicroscopic scales at which lithium exists in batteries where it is found to\nhave enhanced strength, with implications for dendrite suppression strategies.\nComputationally, there is a lack of empirical potentials that are consistently\nquantitatively accurate across all properties and ab-initio calculations are\ntoo costly. In this work, we train Machine Learning Interaction Potentials\n(MLIPs) on Density Functional Theory (DFT) data to state-of-the-art accuracy in\nreproducing experimental and ab-initio results across a wide range of\nsimulations at large length and time scales. We accurately predict\nthermodynamic properties, phonon spectra, temperature dependence of elastic\nconstants and various surface properties inaccessible using DFT. We establish\nthat there exists a Bell-Evans-Polanyi relation correlating the self-adsorption\nenergy and the minimum surface diffusion barrier for high Miller index facets.\n", "date": "2023-04-24", "categories": []}
{"id": "2206.13727", "title": "Persistent homology-based descriptor for machine-learning potential of\n  amorphous structures", "abstract": "  High-accuracy prediction of the physical properties of amorphous materials is\nchallenging in condensed-matter physics. A promising method to achieve this is\nmachine-learning potentials, which is an alternative to computationally\ndemanding ab initio calculations. When applying machine-learning potentials,\nthe construction of descriptors to represent atomic configurations is crucial.\nThese descriptors should be invariant to symmetry operations. Handcrafted\nrepresentations using a smooth overlap of atomic positions and graph neural\nnetworks (GNN) are examples of methods used for constructing symmetry-invariant\ndescriptors. In this study, we propose a novel descriptor based on a\npersistence diagram (PD), a two-dimensional representation of persistent\nhomology (PH). First, we demonstrated that the normalized two-dimensional\nhistogram obtained from PD could predict the average energy per atom of\namorphous carbon (aC) at various densities, even when using a simple model.\nSecond, an analysis of the dimensional reduction results of the descriptor\nspaces revealed that PH can be used to construct descriptors with\ncharacteristics similar to those of a latent space in a GNN. These results\nindicate that PH is a promising method for constructing descriptors suitable\nfor machine-learning potentials without hyperparameter tuning and deep-learning\ntechniques.\n", "date": "2022-06-28", "categories": []}
{"id": "2302.05313", "title": "Discovery of sparse hysteresis models for piezoelectric materials", "abstract": "  This article presents an approach for modelling hysteresis in piezoelectric\nmaterials, that leverages recent advancements in machine learning, particularly\nin sparse-regression techniques. While sparse regression has previously been\nused to model various scientific and engineering phenomena, its application to\nnonlinear hysteresis modelling in piezoelectric materials has yet to be\nexplored. The study employs the least-squares algorithm with a sequential\nthreshold to model the dynamic system responsible for hysteresis, resulting in\na concise model that accurately predicts hysteresis for both simulated and\nexperimental piezoelectric material data. Several numerical experiments are\nperformed, including learning butterfly-shaped hysteresis and modelling\nreal-world hysteresis data for a piezoelectric actuator. The presented approach\nis compared to traditional regression-based and neural network methods,\ndemonstrating its efficiency and robustness. Source code is available at\nhttps://github.com/chandratue/SmartHysteresis\n", "date": "2023-02-10", "categories": []}
{"id": "2305.01101", "title": "Leveraging Language Representation for Material Recommendation, Ranking,\n  and Exploration", "abstract": "  Data-driven approaches for material discovery and design have been\naccelerated by emerging efforts in machine learning. However, general\nrepresentations of crystals to explore the vast material search space remain\nlimited. We introduce a material discovery framework that uses natural language\nembeddings derived from language models as representations of compositional and\nstructural features. The discovery framework consists of a joint scheme that\nfirst recalls relevant candidates, and next ranks the candidates based on\nmultiple target properties. The contextual knowledge encoded in language\nrepresentations conveys information about material properties and structures,\nenabling both representational similarity analysis for recall, and multi-task\nlearning to share information across related properties. By applying the\nframework to thermoelectrics, we demonstrate diversified recommendations of\nprototype structures and identify under-studied high-performance material\nspaces. The recommended materials are corroborated by first-principles\ncalculations and experiments, revealing novel materials with potential high\nperformance. Our framework provides a task-agnostic means for effective\nmaterial recommendation and can be applied to various material systems.\n", "date": "2023-05-01", "categories": []}
{"id": "2305.12010", "title": "Chemellia: An Ecosystem for Atomistic Scientific Machine Learning", "abstract": "  Chemellia is an open-source framework for atomistic machine learning in the\nJulia programming language. The framework takes advantage of Julia's high speed\nas well as the ability to share and reuse code and interfaces through the\nparadigm of multiple dispatch. Chemellia is designed to make use of existing\ninterfaces and avoid ``reinventing the wheel'' wherever possible. A key aspect\nof the Chemellia ecosystem is the ChemistryFeaturization interface for defining\nand encoding features -- it is designed to maximize interoperability between\nfeaturization schemes and elements thereof, to maintain provenance of encoded\nfeatures, and to ensure easy decodability and reconfigurability to enable\nfeature engineering experiments. This embodies the overall design principles of\nthe Chemellia ecosystem: separation of concerns, interoperability, and\ntransparency. We illustrate these principles by discussing the implementation\nof crystal graph convolutional neural networks for material property\nprediction.\n", "date": "2023-05-19", "categories": []}
{"id": "2305.12060", "title": "Mechanical Property Design of Bio-compatible Mg alloys using\n  Machine-Learning Algorithms", "abstract": "  Magnesium alloys are attractive options for temporary bio-implants because of\ntheir biocompatibility, controlled corrosion rate, and similarity to natural\nbone in terms of stiffness and density. Nevertheless, their low mechanical\nstrength hinders their use as cardiovascular stents and bone substitutes. While\nit is possible to engineer alloys with the desired mechanical strength,\noptimizing the mechanical properties of biocompatible magnesium alloys using\nconventional experimental methods is time-consuming and expensive. Therefore,\nArtificial Intelligence (AI) can be leveraged to streamline the alloy design\nprocess and reduce the required time. In this study, a machine learning model\nwas developed to predict the yield strength (YS) of biocompatible magnesium\nalloys with an $R^2$ accuracy of 91\\%. The predictive model was then validated\nusing the CALPHAD technique and thermodynamics calculations. Next, the\npredictive model was employed as the fitness function of a genetic algorithm to\noptimize the alloy composition for high-strength biocompatible magnesium\nimplants. As a result, two alloys were proposed and synthesized, exhibiting YS\nvalues of 108 and 113 MPa, respectively. These values were substantially higher\nthan those of conventional magnesium biocompatible alloys and closer to the YS\nand compressive strength of natural bone. Finally, the synthesized alloys were\nsubjected to microstructure analysis and mechanical property testing to\nvalidate and evaluate the performance of the proposed AI-based alloy design\napproach for creating alloys with specific properties suitable for diverse\napplications.\n", "date": "2023-05-20", "categories": []}
{"id": "2305.11805", "title": "PANNA 2.0: Efficient neural network interatomic potentials and new\n  architectures", "abstract": "  We present the latest release of PANNA 2.0 (Properties from Artificial Neural\nNetwork Architectures), a code for the generation of neural network interatomic\npotentials based on local atomic descriptors and multilayer perceptrons. Built\non a new back end, this new release of PANNA features improved tools for\ncustomizing and monitoring network training, better GPU support including a\nfast descriptor calculator, new plugins for external codes and a new\narchitecture for the inclusion of long-range electrostatic interactions through\na variational charge equilibration scheme. We present an overview of the main\nfeatures of the new code, and several benchmarks comparing the accuracy of\nPANNA models to the state of the art, on commonly used benchmarks as well as\nricher datasets.\n", "date": "2023-05-19", "categories": []}
{"id": "2211.08179", "title": "Artificial intelligence approaches for materials-by-design of energetic\n  materials: state-of-the-art, challenges, and future directions", "abstract": "  Artificial intelligence (AI) is rapidly emerging as an enabling tool for\nsolving various complex materials design problems. This paper aims to review\nrecent advances in AI-driven materials-by-design and their applications to\nenergetic materials (EM). Trained with data from numerical simulations and/or\nphysical experiments, AI models can assimilate trends and patterns within the\ndesign parameter space, identify optimal material designs (micro-morphologies,\ncombinations of materials in composites, etc.), and point to designs with\nsuperior/targeted property and performance metrics. We review approaches\nfocusing on such capabilities with respect to the three main stages of\nmaterials-by-design, namely representation learning of microstructure\nmorphology (i.e., shape descriptors), structure-property-performance (S-P-P)\nlinkage estimation, and optimization/design exploration. We provide a\nperspective view of these methods in terms of their potential, practicality,\nand efficacy towards the realization of materials-by-design. Specifically,\nmethods in the literature are evaluated in terms of their capacity to learn\nfrom a small/limited number of data, computational complexity,\ngeneralizability/scalability to other material species and operating\nconditions, interpretability of the model predictions, and the burden of\nsupervision/data annotation. Finally, we suggest a few promising future\nresearch directions for EM materials-by-design, such as meta-learning, active\nlearning, Bayesian learning, and semi-/weakly-supervised learning, to bridge\nthe gap between machine learning research and EM research.\n", "date": "2022-11-15", "categories": []}
{"id": "2302.03598", "title": "Data-driven anisotropic finite viscoelasticity using neural ordinary\n  differential equations", "abstract": "  We develop a fully data-driven model of anisotropic finite viscoelasticity\nusing neural ordinary differential equations as building blocks. We replace the\nHelmholtz free energy function and the dissipation potential with data-driven\nfunctions that a priori satisfy physics-based constraints such as objectivity\nand the second law of thermodynamics. Our approach enables modeling\nviscoelastic behavior of materials under arbitrary loads in three-dimensions\neven with large deformations and large deviations from the thermodynamic\nequilibrium. The data-driven nature of the governing potentials endows the\nmodel with much needed flexibility in modeling the viscoelastic behavior of a\nwide class of materials. We train the model using stress-strain data from\nbiological and synthetic materials including humain brain tissue, blood clots,\nnatural rubber and human myocardium and show that the data-driven method\noutperforms traditional, closed-form models of viscoelasticity.\n", "date": "2023-01-11", "categories": []}
{"id": "2305.03797", "title": "Materials Informatics: An Algorithmic Design Rule", "abstract": "  Materials informatics, data-enabled investigation, is a \"fourth paradigm\" in\nmaterials science research after the conventional empirical approach,\ntheoretical science, and computational research. Materials informatics has two\nessential ingredients: fingerprinting materials proprieties and the theory of\nstatistical inference and learning. We have researched the organic\nsemiconductor's enigmas through the materials informatics approach. By applying\ndiverse neural network topologies, logical axiom, and inferencing information\nscience, we have developed data-driven procedures for novel organic\nsemiconductor discovery for the semiconductor industry and knowledge extraction\nfor the materials science community. We have reviewed and corresponded with\nvarious algorithms for the neural network design topology for the materials\ninformatics dataset.\n", "date": "2023-05-05", "categories": []}
{"id": "2212.02421", "title": "Score-based denoising for atomic structure identification", "abstract": "  We propose an effective method for removing thermal vibrations that\ncomplicate the task of analyzing complex dynamics in atomistic simulation of\ncondensed matter. Our method iteratively subtracts thermal noises or\nperturbations in atomic positions using a denoising score function trained on\nsynthetically noised but otherwise perfect crystal lattices. The resulting\ndenoised structures clearly reveal underlying crystal order while retaining\ndisorder associated with crystal defects. Purely geometric, agnostic to\ninteratomic potentials, and trained without inputs from explicit simulations,\nour denoiser can be applied to simulation data generated from vastly different\ninteratomic interactions. The denoiser is shown to improve existing\nclassification methods such as common neighbor analysis and polyhedral template\nmatching, reaching perfect classification accuracy on a recent benchmark\ndataset of thermally perturbed structures up to the melting point. Demonstrated\nhere in a wide variety of atomistic simulation contexts, the denoiser is\ngeneral, robust, and readily extendable to delineate order from disorder in\nstructurally and chemically complex materials.\n", "date": "2022-12-05", "categories": []}
{"id": "2305.00003", "title": "Neural Network Accelerated Process Design of Polycrystalline\n  Microstructures", "abstract": "  Computational experiments are exploited in finding a well-designed processing\npath to optimize material structures for desired properties. This requires\nunderstanding the interplay between the processing-(micro)structure-property\nlinkages using a multi-scale approach that connects the macro-scale (process\nparameters) to meso (homogenized properties) and micro (crystallographic\ntexture) scales. Due to the nature of the problem's multi-scale modeling setup,\npossible processing path choices could grow exponentially as the decision tree\nbecomes deeper, and the traditional simulators' speed reaches a critical\ncomputational threshold. To lessen the computational burden for predicting\nmicrostructural evolution under given loading conditions, we develop a neural\nnetwork (NN)-based method with physics-infused constraints. The NN aims to\nlearn the evolution of microstructures under each elementary process. Our\nmethod is effective and robust in finding optimal processing paths. In this\nstudy, our NN-based method is applied to maximize the homogenized stiffness of\na Copper microstructure, and it is found to be 686 times faster while achieving\n0.053% error in the resulting homogenized stiffness compared to the traditional\nfinite element simulator on a 10-process experiment.\n", "date": "2023-04-11", "categories": []}
{"id": "2209.11234", "title": "Artificial Intelligence in Material Engineering: A review on\n  applications of AI in Material Engineering", "abstract": "  The role of artificial intelligence (AI) in material science and engineering\n(MSE) is becoming increasingly important as AI technology advances. The\ndevelopment of high-performance computing has made it possible to test deep\nlearning (DL) models with significant parameters, providing an opportunity to\novercome the limitation of traditional computational methods, such as density\nfunctional theory (DFT), in property prediction. Machine learning (ML)-based\nmethods are faster and more accurate than DFT-based methods. Furthermore, the\ngenerative adversarial networks (GANs) have facilitated the generation of\nchemical compositions of inorganic materials without using crystal structure\ninformation. These developments have significantly impacted material\nengineering (ME) and research. Some of the latest developments in AI in ME\nherein are reviewed. First, the development of AI in the critical areas of ME,\nsuch as in material processing, the study of structure and material property,\nand measuring the performance of materials in various aspects, is discussed.\nThen, the significant methods of AI and their uses in MSE, such as graph neural\nnetwork, generative models, transfer of learning, etc. are discussed. The use\nof AI to analyze the results from existing analytical instruments is also\ndiscussed. Finally, AI's advantages, disadvantages, and future in ME are\ndiscussed.\n", "date": "2022-09-15", "categories": []}
{"id": "2209.01307", "title": "TransPolymer: a Transformer-based language model for polymer property\n  predictions", "abstract": "  Accurate and efficient prediction of polymer properties is of great\nsignificance in polymer design. Conventionally, expensive and time-consuming\nexperiments or simulations are required to evaluate polymer functions.\nRecently, Transformer models, equipped with self-attention mechanisms, have\nexhibited superior performance in natural language processing. However, such\nmethods have not been investigated in polymer sciences. Herein, we report\nTransPolymer, a Transformer-based language model for polymer property\nprediction. Our proposed polymer tokenizer with chemical awareness enables\nlearning representations from polymer sequences. Rigorous experiments on ten\npolymer property prediction benchmarks demonstrate the superior performance of\nTransPolymer. Moreover, we show that TransPolymer benefits from pretraining on\nlarge unlabeled dataset via Masked Language Modeling. Experimental results\nfurther manifest the important role of self-attention in modeling polymer\nsequences. We highlight this model as a promising computational tool for\npromoting rational polymer design and understanding structure-property\nrelationships from a data science view.\n", "date": "2022-09-03", "categories": []}
{"id": "2207.05013", "title": "Boosting Heterogeneous Catalyst Discovery by Structurally Constrained\n  Deep Learning Models", "abstract": "  The discovery of new catalysts is one of the significant topics of\ncomputational chemistry as it has the potential to accelerate the adoption of\nrenewable energy sources. Recently developed deep learning approaches such as\ngraph neural networks (GNNs) open new opportunity to significantly extend scope\nfor modelling novel high-performance catalysts. Nevertheless, the graph\nrepresentation of particular crystal structure is not a straightforward task\ndue to the ambiguous connectivity schemes and numerous embeddings of nodes and\nedges. Here we present embedding improvement for GNN that has been modified by\nVoronoi tesselation and is able to predict the energy of catalytic systems\nwithin Open Catalyst Project dataset. Enrichment of the graph was calculated\nvia Voronoi tessellation and the corresponding contact solid angles and types\n(direct or indirect) were considered as features of edges and Voronoi volumes\nwere used as node characteristics. The auxiliary approach was enriching node\nrepresentation by intrinsic atomic properties (electronegativity, period and\ngroup position). Proposed modifications allowed us to improve the mean absolute\nerror of the original model and the final error equals to 651 meV per atom on\nthe Open Catalyst Project dataset and 6 meV per atom on the intermetallics\ndataset. Also, by consideration of additional dataset, we show that a sensible\nchoice of data can decrease the error to values above physically-based 20 meV\nper atom threshold.\n", "date": "2022-07-11", "categories": []}
{"id": "2304.07445", "title": "A framework for fully autonomous design of materials via multiobjective\n  optimization and active learning: challenges and next steps", "abstract": "  In order to deploy machine learning in a real-world self-driving laboratory\nwhere data acquisition is costly and there are multiple competing design\ncriteria, systems need to be able to intelligently sample while balancing\nperformance trade-offs and constraints. For these reasons, we present an active\nlearning process based on multiobjective black-box optimization with\ncontinuously updated machine learning models. This workflow is built on\nopen-source technologies for real-time data streaming and modular\nmultiobjective optimization software development. We demonstrate a proof of\nconcept for this workflow through the autonomous operation of a continuous-flow\nchemistry laboratory, which identifies ideal manufacturing conditions for the\nelectrolyte 2,2,2-trifluoroethyl methyl carbonate.\n", "date": "2023-04-15", "categories": []}
{"id": "2211.07881", "title": "ET-AL: Entropy-Targeted Active Learning for Bias Mitigation in Materials\n  Data", "abstract": "  Growing materials data and data-driven informatics drastically promote the\ndiscovery and design of materials. While there are significant advancements in\ndata-driven models, the quality of data resources is less studied despite its\nhuge impact on model performance. In this work, we focus on data bias arising\nfrom uneven coverage of materials families in existing knowledge. Observing\ndifferent diversities among crystal systems in common materials databases, we\npropose an information entropy-based metric for measuring this bias. To\nmitigate the bias, we develop an entropy-targeted active learning (ET-AL)\nframework, which guides the acquisition of new data to improve the diversity of\nunderrepresented crystal systems. We demonstrate the capability of ET-AL for\nbias mitigation and the resulting improvement in downstream machine learning\nmodels. This approach is broadly applicable to data-driven materials discovery,\nincluding autonomous data acquisition and dataset trimming to reduce bias, as\nwell as data-driven informatics in other scientific domains.\n", "date": "2022-11-15", "categories": []}
{"id": "2303.07000", "title": "Predicting Density of States via Multi-modal Transformer", "abstract": "  The density of states (DOS) is a spectral property of materials, which\nprovides fundamental insights on various characteristics of materials. In this\npaper, we propose a model to predict the DOS by reflecting the nature of DOS:\nDOS determines the general distribution of states as a function of energy.\nSpecifically, we integrate the heterogeneous information obtained from the\ncrystal structure and the energies via multi-modal transformer, thereby\nmodeling the complex relationships between the atoms in the crystal structure,\nand various energy levels. Extensive experiments on two types of DOS, i.e.,\nPhonon DOS and Electron DOS, with various real-world scenarios demonstrate the\nsuperiority of DOSTransformer. The source code for DOSTransformer is available\nat https://github.com/HeewoongNoh/DOSTransformer.\n", "date": "2023-03-13", "categories": []}
{"id": "2301.03729", "title": "Evaluating the Transferability of Machine-Learned Force Fields for\n  Material Property Modeling", "abstract": "  Machine-learned force fields have generated significant interest in recent\nyears as a tool for molecular dynamics (MD) simulations, with the aim of\ndeveloping accurate and efficient models that can replace classical interatomic\npotentials. However, before these models can be confidently applied to\nmaterials simulations, they must be thoroughly tested and validated. The\nexisting tests on the radial distribution function and mean-squared\ndisplacements are insufficient in assessing the transferability of these\nmodels. Here we present a more comprehensive set of benchmarking tests for\nevaluating the transferability of machine-learned force fields. We use a graph\nneural network (GNN)-based force field coupled with the OpenMM package to carry\nout MD simulations for Argon as a test case. Our tests include computational\nX-ray photon correlation spectroscopy (XPCS) signals, which capture the density\nfluctuation at various length scales in the liquid phase, as well as phonon\ndensity-of-state in the solid phase and the liquid-solid phase transition\nbehavior. Our results show that the model can accurately capture the behavior\nof the solid phase only when the configurations from the solid phase are\nincluded in the training dataset. This underscores the importance of\nappropriately selecting the training data set when developing machine-learned\nforce fields. The tests presented in this work provide a necessary foundation\nfor the development and application of machine-learned force fields for\nmaterials simulations.\n", "date": "2023-01-10", "categories": []}
{"id": "2304.01120", "title": "Synthesis parameter effect detection using quantitative representations\n  and high dimensional distribution distances", "abstract": "  Detection of effects of the parameters of the synthetic process on the\nmicrostructure of materials is an important, yet elusive goal of materials\nscience. We develop a method for detecting effects based on copula theory, high\ndimensional distribution distances, and permutational statistics to analyze a\ndesigned experiment synthesizing plutonium oxide from Pu(III) Oxalate. We\ndetect effects of strike order and oxalic acid feed on the microstructure of\nthe resulting plutonium oxide, which match the literature well. We also detect\nexcess bivariate effects between the pairs of acid concentration, strike order\nand precipitation temperature.\n", "date": "2023-04-03", "categories": []}
{"id": "2303.18055", "title": "Deep neural operator for learning transient response of interpenetrating\n  phase composites subject to dynamic loading", "abstract": "  Additive manufacturing has been recognized as an industrial technological\nrevolution for manufacturing, which allows fabrication of materials with\ncomplex three-dimensional (3D) structures directly from computer-aided design\nmodels. The mechanical properties of interpenetrating phase composites (IPCs),\nespecially response to dynamic loading, highly depend on their 3D structures.\nIn general, for each specified structural design, it could take hours or days\nto perform either finite element analysis (FEA) or experiments to test the\nmechanical response of IPCs to a given dynamic load. To accelerate the\nphysics-based prediction of mechanical properties of IPCs for various\nstructural designs, we employ a deep neural operator (DNO) to learn the\ntransient response of IPCs under dynamic loading as surrogate of physics-based\nFEA models. We consider a 3D IPC beam formed by two metals with a ratio of\nYoung's modulus of 2.7, wherein random blocks of constituent materials are used\nto demonstrate the generality and robustness of the DNO model. To obtain FEA\nresults of IPC properties, 5,000 random time-dependent strain loads generated\nby a Gaussian process kennel are applied to the 3D IPC beam, and the reaction\nforces and stress fields inside the IPC beam under various loading are\ncollected. Subsequently, the DNO model is trained using an incremental learning\nmethod with sequence-to-sequence training implemented in JAX, leading to a 100X\nspeedup compared to widely used vanilla deep operator network models. After an\noffline training, the DNO model can act as surrogate of physics-based FEA to\npredict the transient mechanical response in terms of reaction force and stress\ndistribution of the IPCs to various strain loads in one second at an accuracy\nof 98%. Also, the learned operator is able to provide extended prediction of\nthe IPC beam subject to longer random strain loads at a reasonably well\naccuracy.\n", "date": "2023-03-30", "categories": []}
{"id": "2303.15663", "title": "Predicting Thermoelectric Power Factor of Bismuth Telluride During Laser\n  Powder Bed Fusion Additive Manufacturing", "abstract": "  An additive manufacturing (AM) process, like laser powder bed fusion, allows\nfor the fabrication of objects by spreading and melting powder in layers until\na freeform part shape is created. In order to improve the properties of the\nmaterial involved in the AM process, it is important to predict the material\ncharacterization property as a function of the processing conditions. In\nthermoelectric materials, the power factor is a measure of how efficiently the\nmaterial can convert heat to electricity. While earlier works have predicted\nthe material characterization properties of different thermoelectric materials\nusing various techniques, implementation of machine learning models to predict\nthe power factor of bismuth telluride (Bi2Te3) during the AM process has not\nbeen explored. This is important as Bi2Te3 is a standard material for low\ntemperature applications. Thus, we used data about manufacturing processing\nparameters involved and in-situ sensor monitoring data collected during AM of\nBi2Te3, to train different machine learning models in order to predict its\nthermoelectric power factor. We implemented supervised machine learning\ntechniques using 80% training and 20% test data and further used the\npermutation feature importance method to identify important processing\nparameters and in-situ sensor features which were best at predicting power\nfactor of the material. Ensemble-based methods like random forest, AdaBoost\nclassifier, and bagging classifier performed the best in predicting power\nfactor with the highest accuracy of 90% achieved by the bagging classifier\nmodel. Additionally, we found the top 15 processing parameters and in-situ\nsensor features to characterize the material manufacturing property like power\nfactor. These features could further be optimized to maximize power factor of\nthe thermoelectric material and improve the quality of the products built using\nthis material.\n", "date": "2023-03-28", "categories": []}
{"id": "2303.16088", "title": "GNN-Assisted Phase Space Integration with Application to Atomistics", "abstract": "  Overcoming the time scale limitations of atomistics can be achieved by\nswitching from the state-space representation of Molecular Dynamics (MD) to a\nstatistical-mechanics-based representation in phase space, where approximations\nsuch as maximum-entropy or Gaussian phase packets (GPP) evolve the atomistic\nensemble in a time-coarsened fashion. In practice, this requires the\ncomputation of expensive high-dimensional integrals over all of phase space of\nan atomistic ensemble. This, in turn, is commonly accomplished efficiently by\nlow-order numerical quadrature. We show that numerical quadrature in this\ncontext, unfortunately, comes with a set of inherent problems, which corrupt\nthe accuracy of simulations -- especially when dealing with crystal lattices\nwith imperfections. As a remedy, we demonstrate that Graph Neural Networks,\ntrained on Monte-Carlo data, can serve as a replacement for commonly used\nnumerical quadrature rules, overcoming their deficiencies and significantly\nimproving the accuracy. This is showcased by three benchmarks: the thermal\nexpansion of copper, the martensitic phase transition of iron, and the energy\nof grain boundaries. We illustrate the benefits of the proposed technique over\nclassically used third- and fifth-order Gaussian quadrature, we highlight the\nimpact on time-coarsened atomistic predictions, and we discuss the\ncomputational efficiency. The latter is of general importance when performing\nfrequent evaluation of phase space or other high-dimensional integrals, which\nis why the proposed framework promises applications beyond the scope of\natomistics.\n", "date": "2023-03-20", "categories": []}
{"id": "2303.14223", "title": "Machine Guided Discovery of Novel Carbon Capture Solvents", "abstract": "  The increasing importance of carbon capture technologies for deployment in\nremediating CO2 emissions, and thus the necessity to improve capture materials\nto allow scalability and efficiency, faces the challenge of materials\ndevelopment, which can require substantial costs and time. Machine learning\noffers a promising method for reducing the time and resource burdens of\nmaterials development through efficient correlation of structure-property\nrelationships to allow down-selection and focusing on promising candidates.\nTowards demonstrating this, we have developed an end-to-end \"discovery cycle\"\nto select new aqueous amines compatible with the commercially viable acid gas\nscrubbing carbon capture. We combine a simple, rapid laboratory assay for CO2\nabsorption with a machine learning based molecular fingerprinting model\napproach. The prediction process shows 60% accuracy against experiment for both\nmaterial parameters and 80% for a single parameter on an external test set. The\ndiscovery cycle determined several promising amines that were verified\nexperimentally, and which had not been applied to carbon capture previously. In\nthe process we have compiled a large, single-source data set for carbon capture\namines and produced an open source machine learning tool for the identification\nof amine molecule candidates\n(https://github.com/IBM/Carbon-capture-fingerprint-generation).\n", "date": "2023-03-24", "categories": []}
{"id": "2303.08291", "title": "Machine Learning Approaches in Agile Manufacturing with Recycled\n  Materials for Sustainability", "abstract": "  It is important to develop sustainable processes in materials science and\nmanufacturing that are environmentally friendly. AI can play a significant role\nin decision support here as evident from our earlier research leading to tools\ndeveloped using our proposed machine learning based approaches. Such tools\nserved the purpose of computational estimation and expert systems. This\nresearch addresses environmental sustainability in materials science via\ndecision support in agile manufacturing using recycled and reclaimed materials.\nIt is a safe and responsible way to turn a specific waste stream to value-added\nproducts. We propose to use data-driven methods in AI by applying machine\nlearning models for predictive analysis to guide decision support in\nmanufacturing. This includes harnessing artificial neural networks to study\nparameters affecting heat treatment of materials and impacts on their\nproperties; deep learning via advances such as convolutional neural networks to\nexplore grain size detection; and other classifiers such as Random Forests to\nanalyze phrase fraction detection. Results with all these methods seem\npromising to embark on further work, e.g. ANN yields accuracy around 90\\% for\npredicting micro-structure development as per quench tempering, a heat\ntreatment process. Future work entails several challenges: investigating\nvarious computer vision models (VGG, ResNet etc.) to find optimal accuracy,\nefficiency and robustness adequate for sustainable processes; creating\ndomain-specific tools using machine learning for decision support in agile\nmanufacturing; and assessing impacts on sustainability with metrics\nincorporating the appropriate use of recycled materials as well as the\neffectiveness of developed products. Our work makes impacts on green technology\nfor smart manufacturing, and is motivated by related work in the highly\ninteresting realm of AI for materials science.\n", "date": "2023-03-15", "categories": []}
{"id": "2303.12164", "title": "Viscoelastic Constitutive Artificial Neural Networks (vCANNs) $-$ a\n  framework for data-driven anisotropic nonlinear finite viscoelasticity", "abstract": "  The constitutive behavior of polymeric materials is often modeled by finite\nlinear viscoelastic (FLV) or quasi-linear viscoelastic (QLV) models. These\npopular models are simplifications that typically cannot accurately capture the\nnonlinear viscoelastic behavior of materials. For example, the success of\nattempts to capture strain rate-dependent behavior has been limited so far. To\novercome this problem, we introduce viscoelastic Constitutive Artificial Neural\nNetworks (vCANNs), a novel physics-informed machine learning framework for\nanisotropic nonlinear viscoelasticity at finite strains. vCANNs rely on the\nconcept of generalized Maxwell models enhanced with nonlinear strain\n(rate)-dependent properties represented by neural networks. The flexibility of\nvCANNs enables them to automatically identify accurate and sparse constitutive\nmodels of a broad range of materials. To test vCANNs, we trained them on\nstress-strain data from Polyvinyl Butyral, the electro-active polymers VHB 4910\nand 4905, and a biological tissue, the rectus abdominis muscle. Different\nloading conditions were considered, including relaxation tests, cyclic\ntension-compression tests, and blast loads. We demonstrate that vCANNs can\nlearn to capture the behavior of all these materials accurately and\ncomputationally efficiently without human guidance.\n", "date": "2023-03-21", "categories": []}
{"id": "2303.10722", "title": "Q-RBSA: High-Resolution 3D EBSD Map Generation Using An Efficient\n  Quaternion Transformer Network", "abstract": "  Gathering 3D material microstructural information is time-consuming,\nexpensive, and energy-intensive. Acquisition of 3D data has been accelerated by\ndevelopments in serial sectioning instrument capabilities; however, for\ncrystallographic information, the electron backscatter diffraction (EBSD)\nimaging modality remains rate limiting. We propose a physics-based efficient\ndeep learning framework to reduce the time and cost of collecting 3D EBSD maps.\nOur framework uses a quaternion residual block self-attention network (QRBSA)\nto generate high-resolution 3D EBSD maps from sparsely sectioned EBSD maps. In\nQRBSA, quaternion-valued convolution effectively learns local relations in\norientation space, while self-attention in the quaternion domain captures\nlong-range correlations. We apply our framework to 3D data collected from\ncommercially relevant titanium alloys, showing both qualitatively and\nquantitatively that our method can predict missing samples (EBSD information\nbetween sparsely sectioned mapping points) as compared to high-resolution\nground truth 3D EBSD maps.\n", "date": "2023-03-19", "categories": []}
{"id": "2303.08541", "title": "Adapting U-Net for linear elastic stress estimation in polycrystal Zr\n  microstructures", "abstract": "  A variant of the U-Net convolutional neural network architecture is proposed\nto estimate linear elastic compatibility stresses in a-Zr (hcp) polycrystalline\ngrain structures. Training data was generated using VGrain software with a\nregularity alpha of 0.73 and uniform random orientation for the grain\nstructures and ABAQUS to evaluate the stress welds using the finite element\nmethod. The initial dataset contains 200 samples with 20 held from training for\nvalidation. The network gives speedups of around 200x to 6000x using a CPU or\nGPU, with signifcant memory savings, compared to finite element analysis with a\nmodest reduction in accuracy of up to 10%. Network performance is not\ncorrelated with grain structure regularity or texture, showing generalisation\nof the network beyond the training set to arbitrary Zr crystal structures.\nPerformance when trained with 200 and 400 samples was measured, finding an\nimprovement in accuracy of approximately 10% when the size of the dataset was\ndoubled.\n", "date": "2023-03-15", "categories": []}
{"id": "2206.08917", "title": "The Open Catalyst 2022 (OC22) Dataset and Challenges for Oxide\n  Electrocatalysts", "abstract": "  The development of machine learning models for electrocatalysts requires a\nbroad set of training data to enable their use across a wide variety of\nmaterials. One class of materials that currently lacks sufficient training data\nis oxides, which are critical for the development of OER catalysts. To address\nthis, we developed the OC22 dataset, consisting of 62,331 DFT relaxations\n(~9,854,504 single point calculations) across a range of oxide materials,\ncoverages, and adsorbates. We define generalized total energy tasks that enable\nproperty prediction beyond adsorption energies; we test baseline performance of\nseveral graph neural networks; and we provide pre-defined dataset splits to\nestablish clear benchmarks for future efforts. In the most general task,\nGemNet-OC sees a ~36% improvement in energy predictions when combining the\nchemically dissimilar OC20 and OC22 datasets via fine-tuning. Similarly, we\nachieved a ~19% improvement in total energy predictions on OC20 and a ~9%\nimprovement in force predictions in OC22 when using joint training. We\ndemonstrate the practical utility of a top performing model by capturing\nliterature adsorption energies and important OER scaling relationships. We\nexpect OC22 to provide an important benchmark for models seeking to incorporate\nintricate long-range electrostatic and magnetic interactions in oxide surfaces.\nDataset and baseline models are open sourced, and a public leaderboard is\navailable to encourage continued community developments on the total energy\ntasks and data.\n", "date": "2022-06-17", "categories": []}
{"id": "2303.03403", "title": "DA-VEGAN: Differentiably Augmenting VAE-GAN for microstructure\n  reconstruction from extremely small data sets", "abstract": "  Microstructure reconstruction is an important and emerging field of research\nand an essential foundation to improving inverse computational materials\nengineering (ICME). Much of the recent progress in the field is made based on\ngenerative adversarial networks (GANs). Although excellent results have been\nachieved throughout a variety of materials, challenges remain regarding the\ninterpretability of the model's latent space as well as the applicability to\nextremely small data sets. The present work addresses these issues by\nintroducing DA-VEGAN, a model with two central innovations. First, a\n$\\beta$-variational autoencoder is incorporated into a hybrid GAN architecture\nthat allows to penalize strong nonlinearities in the latent space by an\nadditional parameter, $\\beta$. Secondly, a custom differentiable data\naugmentation scheme is developed specifically for this architecture. The\ndifferentiability allows the model to learn from extremely small data sets\nwithout mode collapse or deteriorated sample quality. An extensive validation\non a variety of structures demonstrates the potential of the method and future\ndirections of investigation are discussed.\n", "date": "2023-02-17", "categories": []}
{"id": "2303.03748", "title": "Computing formation enthalpies through an explainable machine learning\n  method: the case of Lanthanide Orthophosphates solid solutions", "abstract": "  In the last decade, the use of Machine and Deep Learning (MDL) methods in\nCondensed Matter physics has seen a steep increase in the number of problems\ntackled and methods employed. A number of distinct MDL approaches have been\nemployed in many different topics; from prediction of materials properties to\ncomputation of Density Functional Theory potentials and inter-atomic force\nfields. In many cases the result is a surrogate model which returns promising\npredictions but is opaque on the inner mechanisms of its success. On the other\nhand, the typical practitioner looks for answers that are explainable and\nprovide a clear insight on the mechanisms governing a physical phenomena. In\nthis work, we describe a proposal to use a sophisticated combination of\ntraditional Machine Learning methods to obtain an explainable model that\noutputs an explicit functional formulation for the material property of\ninterest. We demonstrate the effectiveness of our methodology in deriving a new\nhighly accurate expression for the enthalpy of formation of solid solutions of\nlanthanides orthophosphates.\n", "date": "2023-03-07", "categories": []}
{"id": "2211.15415", "title": "Machine Learning for Screening Large Organic Molecules", "abstract": "  Organic semiconductors are promising materials for cheap, scalable and\nsustainable electronics, light-emitting diodes and photovoltaics. For organic\nphotovoltaic cells, it is a challenge to find compounds with suitable\nproperties in the vast chemical compound space. For example, the ionization\nenergy should fit to the optical spectrum of sun light, and the energy levels\nmust allow efficient charge transport. Here, a machine-learning model is\ndeveloped for rapidly and accurately estimating the HOMO and LUMO energies of a\ngiven molecular structure. It is build upon the SchNet model (Sch\\\"utt et al.\n(2018)) and augmented with a `Set2Set' readout module (Vinyals et al. (2016)).\nThe Set2Set module has more expressive power than sum and average aggregation\nand is more suitable for the complex quantities under consideration. Most\nprevious models have been trained and evaluated on rather small molecules.\nTherefore, the second contribution is extending the scope of machine-learning\nmethods by adding also larger molecules from other sources and establishing a\nconsistent train/validation/test split. As a third contribution, we make a\nmultitask ansatz to resolve the problem of different sources coming at\ndifferent levels of theory. All three contributions in conjunction bring the\naccuracy of the model close to chemical accuracy.\n", "date": "2022-11-23", "categories": []}
{"id": "2209.13269", "title": "Evolution TANN and the identification of internal variables and\n  evolution equations in solid mechanics", "abstract": "  Data-driven and deep learning approaches have demonstrated to have the\npotential of replacing classical constitutive models for complex materials.\nYet, the necessity of structuring constitutive models with an incremental\nformulation has given rise to data-driven approaches where physical quantities,\ne.g. deformation, blend with artificial, non-physical ones, such as the\nincrements in deformation and time. Neural networks and the consequent\nconstitutive models depend, thus, on the particular incremental formulation,\nfail in identifying material representations locally in time, and suffer from\npoor generalization. Herein, we propose a new approach which allows, for the\nfirst time, to decouple the material representation from the incremental\nformulation. Inspired by the Thermodynamics-based Artificial Neural Networks\n(TANN) and the theory of the internal variables, the evolution TANN (eTANN) are\ncontinuous-time and, therefore, independent of the aforementioned artificial\nquantities. Key feature of the proposed approach is the identification of the\nevolution equations of the internal variables in the form of ordinary\ndifferential equations, rather than in an incremental discrete-time form. In\nthis work, we focus attention to juxtapose and show how the various general\nnotions of solid mechanics are implemented in eTANN. The capabilities as well\nas the scalability of the proposed approach are demonstrated through several\napplications involving a broad spectrum of complex material behaviors, from\nplasticity to damage and viscosity (and combination of them). Finally, we show\nthat the proposed approach can be used to speed-up multiscale analyses, by\nvirtue of asymptotic homogenization. eTANN provide excellent results compared\nto detailed fine-scale simulations and offer the possibility not only to\ndescribe the average macroscopic material behavior, but also micromechanical,\ncomplex mechanisms.\n", "date": "2022-09-27", "categories": []}
{"id": "2302.12545", "title": "Hybrid machine-learned homogenization: Bayesian data mining and\n  convolutional neural networks", "abstract": "  Beyond the generally deployed features for microstructure property prediction\nthis study aims to improve the machine learned prediction by developing novel\nfeature descriptors. Therefore, Bayesian infused data mining is conducted to\nacquire samples containing characteristics inexplicable to the current feature\nset, and suitable feature descriptors to describe these characteristics are\nproposed. The iterative development of feature descriptors resulted in 37 novel\nfeatures, being able to reduce the prediction error by roughly one third. To\nfurther improve the predictive model, convolutional neural networks (Conv Nets)\nare deployed to generate auxiliary features in a supervised machine learning\nmanner. The Conv Nets were able to outperform the feature based approach. A key\ningredient for that is a newly proposed data augmentation scheme and the\ndevelopment of so-called deep inception modules. A combination of the feature\nbased approach and the convolutional neural network leads to a hybrid neural\nnetwork: A parallel deployment of the both neural network archetypes in a\nsingle model achieved a relative rooted mean squared error below 1%, more than\nhalving the error compared to prior models operating on the same data. The\nhybrid neural network was found powerful enough to be extended to predict\nvariable material parameters, from a low to high phase contrast, while allowing\nfor arbitrary microstructure geometry at the same time.\n", "date": "2023-02-24", "categories": []}
{"id": "2210.08343", "title": "Modular machine learning-based elastoplasticity: generalization in the\n  context of limited data", "abstract": "  The development of accurate constitutive models for materials that undergo\npath-dependent processes continues to be a complex challenge in computational\nsolid mechanics. Challenges arise both in considering the appropriate model\nassumptions and from the viewpoint of data availability, verification, and\nvalidation. Recently, data-driven modeling approaches have been proposed that\naim to establish stress-evolution laws that avoid user-chosen functional forms\nby relying on machine learning representations and algorithms. However, these\napproaches not only require a significant amount of data but also need data\nthat probes the full stress space with a variety of complex loading paths.\nFurthermore, they rarely enforce all necessary thermodynamic principles as hard\nconstraints. Hence, they are in particular not suitable for low-data or\nlimited-data regimes, where the first arises from the cost of obtaining the\ndata and the latter from the experimental limitations of obtaining labeled\ndata, which is commonly the case in engineering applications. In this work, we\ndiscuss a hybrid framework that can work on a variable amount of data by\nrelying on the modularity of the elastoplasticity formulation where each\ncomponent of the model can be chosen to be either a classical phenomenological\nor a data-driven model depending on the amount of available information and the\ncomplexity of the response. The method is tested on synthetic uniaxial data\ncoming from simulations as well as cyclic experimental data for structural\nmaterials. The discovered material models are found to not only interpolate\nwell but also allow for accurate extrapolation in a thermodynamically\nconsistent manner far outside the domain of the training data. Training aspects\nand details of the implementation of these models into Finite Element\nsimulations are discussed and analyzed.\n", "date": "2022-10-15", "categories": []}
{"id": "2302.06454", "title": "Outlier-Based Domain of Applicability Identification for Materials\n  Property Prediction Models", "abstract": "  Machine learning models have been widely applied for material property\nprediction. However, practical application of these models can be hindered by a\nlack of information about how well they will perform on previously unseen types\nof materials. Because machine learning model predictions depend on the quality\nof the available training data, different domains of the material feature space\nare predicted with different accuracy levels by such models. The ability to\nidentify such domains enables the ability to find the confidence level of each\nprediction, to determine when and how the model should be employed depending on\nthe prediction accuracy requirements of different tasks, and to improve the\nmodel for domains with high errors. In this work, we propose a method to find\ndomains of applicability using a large feature space and also introduce\nanalysis techniques to gain more insight into the detected domains and\nsubdomains.\n", "date": "2023-01-17", "categories": []}
{"id": "2209.07679", "title": "Learning Pair Potentials using Differentiable Simulations", "abstract": "  Learning pair interactions from experimental or simulation data is of great\ninterest for molecular simulations. We propose a general stochastic method for\nlearning pair interactions from data using differentiable simulations\n(DiffSim). DiffSim defines a loss function based on structural observables,\nsuch as the radial distribution function, through molecular dynamics (MD)\nsimulations. The interaction potentials are then learned directly by stochastic\ngradient descent, using backpropagation to calculate the gradient of the\nstructural loss metric with respect to the interaction potential through the MD\nsimulation. This gradient-based method is flexible and can be configured to\nsimulate and optimize multiple systems simultaneously. For example, it is\npossible to simultaneously learn potentials for different temperatures or for\ndifferent compositions. We demonstrate the approach by recovering simple pair\npotentials, such as Lennard-Jones systems, from radial distribution functions.\nWe find that DiffSim can be used to probe a wider functional space of pair\npotentials compared to traditional methods like Iterative Boltzmann Inversion.\nWe show that our methods can be used to simultaneously fit potentials for\nsimulations at different compositions and temperatures to improve the\ntransferability of the learned potentials.\n", "date": "2022-09-16", "categories": []}
{"id": "2302.02555", "title": "Proposing Novel Extrapolative Compounds by Nested Variational\n  Autoencoders", "abstract": "  Materials informatics (MI), which uses artificial intelligence and data\nanalysis techniques to improve the efficiency of materials development, is\nattracting increasing interest from industry. One of its main applications is\nthe rapid development of new high-performance compounds. Recently, several deep\ngenerative models have been proposed to suggest candidate compounds that are\nexpected to satisfy the desired performance. However, they usually have the\nproblem of requiring a large amount of experimental datasets for training to\nachieve sufficient accuracy. In actual cases, it is often possible to\naccumulate only about 1000 experimental data at most. Therefore, the authors\nproposed a deep generative model with nested two variational autoencoders\n(VAEs). The outer VAE learns the structural features of compounds using\nlarge-scale public data, while the inner VAE learns the relationship between\nthe latent variables of the outer VAE and the properties from small-scale\nexperimental data. To generate high performance compounds beyond the range of\nthe training data, the authors also proposed a loss function that amplifies the\ncorrelation between a component of latent variables of the inner VAE and\nmaterial properties. The results indicated that this loss function contributes\nto improve the probability of generating high-performance candidates.\nFurthermore, as a result of verification test with an actual customer in\nchemical industry, it was confirmed that the proposed method is effective in\nreducing the number of experiments to $1/4$ compared to a conventional method.\n", "date": "2023-02-06", "categories": []}
{"id": "2302.02595", "title": "Clarifying Trust of Materials Property Predictions using Neural Networks\n  with Distribution-Specific Uncertainty Quantification", "abstract": "  It is critical that machine learning (ML) model predictions be trustworthy\nfor high-throughput catalyst discovery approaches. Uncertainty quantification\n(UQ) methods allow estimation of the trustworthiness of an ML model, but these\nmethods have not been well explored in the field of heterogeneous catalysis.\nHerein, we investigate different UQ methods applied to a crystal graph\nconvolutional neural network (CGCNN) to predict adsorption energies of\nmolecules on alloys from the Open Catalyst 2020 (OC20) dataset, the largest\nexisting heterogeneous catalyst dataset. We apply three UQ methods to the\nadsorption energy predictions, namely k-fold ensembling, Monte Carlo dropout,\nand evidential regression. The effectiveness of each UQ method is assessed\nbased on accuracy, sharpness, dispersion, calibration, and tightness.\nEvidential regression is demonstrated to be a powerful approach for rapidly\nobtaining tunable, competitively trustworthy UQ estimates for heterogeneous\ncatalysis applications when using neural networks. Recalibration of model\nuncertainties is shown to be essential in practical screening applications of\ncatalysts using uncertainties.\n", "date": "2023-02-06", "categories": []}
{"id": "2302.01486", "title": "Xtal2DoS: Attention-based Crystal to Sequence Learning for Density of\n  States Prediction", "abstract": "  Modern machine learning techniques have been extensively applied to materials\nscience, especially for property prediction tasks. A majority of these methods\naddress scalar property predictions, while more challenging spectral properties\nremain less emphasized. We formulate a crystal-to-sequence learning task and\npropose a novel attention-based learning method, Xtal2DoS, which decodes the\nsequential representation of the material density of states (DoS) properties by\nincorporating the learned atomic embeddings through attention networks.\nExperiments show Xtal2DoS is faster than the existing models, and consistently\noutperforms other state-of-the-art methods on four metrics for two fundamental\nspectral properties, phonon and electronic DoS.\n", "date": "2023-02-03", "categories": []}
{"id": "2301.05824", "title": "Discovery of 2D materials using Transformer Network based Generative\n  Design", "abstract": "  Two-dimensional (2D) materials have wide applications in superconductors,\nquantum, and topological materials. However, their rational design is not well\nestablished, and currently less than 6,000 experimentally synthesized 2D\nmaterials have been reported. Recently, deep learning, data-mining, and density\nfunctional theory (DFT)-based high-throughput calculations are widely performed\nto discover potential new materials for diverse applications. Here we propose a\ngenerative material design pipeline, namely material transformer\ngenerator(MTG), for large-scale discovery of hypothetical 2D materials. We\ntrain two 2D materials composition generators using self-learning neural\nlanguage models based on Transformers with and without transfer learning. The\nmodels are then used to generate a large number of candidate 2D compositions,\nwhich are fed to known 2D materials templates for crystal structure prediction.\nNext, we performed DFT computations to study their thermodynamic stability\nbased on energy-above-hull and formation energy. We report four new\nDFT-verified stable 2D materials with zero e-above-hull energies, including\nNiCl$_4$, IrSBr, CuBr$_3$, and CoBrCl. Our work thus demonstrates the potential\nof our MTG generative materials design pipeline in the discovery of novel 2D\nmaterials and other functional materials.\n", "date": "2023-01-14", "categories": []}
{"id": "2301.04261", "title": "Towards Microstructural State Variables in Materials Systems", "abstract": "  The vast combination of material properties seen in nature are achieved by\nthe complexity of the material microstructure. Advanced characterization and\nphysics based simulation techniques have led to generation of extremely large\nmicrostructural datasets. There is a need for machine learning techniques that\ncan manage data complexity by capturing the maximal amount of information about\nthe microstructure using the least number of variables. This paper aims to\nformulate dimensionality and state variable estimation techniques focused on\nreducing microstructural image data. It is shown that local dimensionality\nestimation based on nearest neighbors tend to give consistent dimension\nestimates for natural images for all p-Minkowski distances. However, it is\nfound that dimensionality estimates have a systematic error for low-bit depth\nmicrostructural images. The use of Manhattan distance to alleviate this issue\nis demonstrated. It is also shown that stacked autoencoders can reconstruct the\ngenerator space of high dimensional microstructural data and provide a sparse\nset of state variables to fully describe the variability in material\nmicrostructures.\n", "date": "2023-01-11", "categories": []}
{"id": "2301.04540", "title": "Towards a unified nonlocal, peridynamics framework for the\n  coarse-graining of molecular dynamics data with fractures", "abstract": "  Molecular dynamics (MD) has served as a powerful tool for designing materials\nwith reduced reliance on laboratory testing. However, the use of MD directly to\ntreat the deformation and failure of materials at the mesoscale is still\nlargely beyond reach. Herein, we propose a learning framework to extract a\nperidynamic model as a mesoscale continuum surrogate from MD simulated material\nfracture datasets. Firstly, we develop a novel coarse-graining method, to\nautomatically handle the material fracture and its corresponding\ndiscontinuities in MD displacement dataset. Inspired by the Weighted\nEssentially Non-Oscillatory scheme, the key idea lies at an adaptive procedure\nto automatically choose the locally smoothest stencil, then reconstruct the\ncoarse-grained material displacement field as piecewise smooth solutions\ncontaining discontinuities. Then, based on the coarse-grained MD data, a\ntwo-phase optimization-based learning approach is proposed to infer the optimal\nperidynamics model with damage criterion. In the first phase, we identify the\noptimal nonlocal kernel function from datasets without material damage, to\ncapture the material stiffness properties. Then, in the second phase, the\nmaterial damage criterion is learnt as a smoothed step function from the data\nwith fractures. As a result, a peridynamics surrogate is obtained. Our\nperidynamics surrogate model can be employed in further prediction tasks with\ndifferent grid resolutions from training, and hence allows for substantial\nreductions in computational cost compared with MD. We illustrate the efficacy\nof the proposed approach with several numerical tests for single layer\ngraphene. Our tests show that the proposed data-driven model is robust and\ngeneralizable: it is capable in modeling the initialization and growth of\nfractures under discretization and loading settings that are different from the\nones used during training.\n", "date": "2023-01-11", "categories": []}
{"id": "2301.02738", "title": "LS-DYNA Machine Learning-based Multiscale Method for Nonlinear Modeling\n  of Short Fiber-Reinforced Composites", "abstract": "  Short-fiber-reinforced composites (SFRC) are high-performance engineering\nmaterials for lightweight structural applications in the automotive and\nelectronics industries. Typically, SFRC structures are manufactured by\ninjection molding, which induces heterogeneous microstructures, and the\nresulting nonlinear anisotropic behaviors are challenging to predict by\nconventional micromechanical analyses. In this work, we present a machine\nlearning-based multiscale method by integrating injection molding-induced\nmicrostructures, material homogenization, and Deep Material Network (DMN) in\nthe finite element simulation software LS-DYNA for structural analysis of SFRC.\nDMN is a physics-embedded machine learning model that learns the microscale\nmaterial morphologies hidden in representative volume elements of composites\nthrough offline training. By coupling DMN with finite elements, we have\ndeveloped a highly accurate and efficient data-driven approach, which predicts\nnonlinear behaviors of composite materials and structures at a computational\nspeed orders-of-magnitude faster than the high-fidelity direct numerical\nsimulation. To model industrial-scale SFRC products, transfer learning is\nutilized to generate a unified DMN database, which effectively captures the\neffects of injection molding-induced fiber orientations and volume fractions on\nthe overall composite properties. Numerical examples are presented to\ndemonstrate the promising performance of this LS-DYNA machine learning-based\nmultiscale method for SFRC modeling.\n", "date": "2023-01-06", "categories": []}
{"id": "2301.03372", "title": "Machine-Learning Prediction of the Computed Band Gaps of Double\n  Perovskite Materials", "abstract": "  Prediction of the electronic structure of functional materials is essential\nfor the engineering of new devices. Conventional electronic structure\nprediction methods based on density functional theory (DFT) suffer from not\nonly high computational cost, but also limited accuracy arising from the\napproximations of the exchange-correlation functional. Surrogate methods based\non machine learning have garnered much attention as a viable alternative to\nbypass these limitations, especially in the prediction of solid-state band\ngaps, which motivated this research study. Herein, we construct a random forest\nregression model for band gaps of double perovskite materials, using a dataset\nof 1306 band gaps computed with the GLLBSC (Gritsenko, van Leeuwen, van Lenthe,\nand Baerends solid correlation) functional. Among the 20 physical features\nemployed, we find that the bulk modulus, superconductivity temperature, and\ncation electronegativity exhibit the highest importance scores, consistent with\nthe physics of the underlying electronic structure. Using the top 10 features,\na model accuracy of 85.6% with a root mean square error of 0.64 eV is obtained,\ncomparable to previous studies. Our results are significant in the sense that\nthey attest to the potential of machine learning regressions for the rapid\nscreening of promising candidate functional materials.\n", "date": "2023-01-04", "categories": []}
{"id": "2212.12047", "title": "Deep learning for size-agnostic inverse design of random-network 3D\n  printed mechanical metamaterials", "abstract": "  Practical applications of mechanical metamaterials often involve solving\ninverse problems where the objective is to find the (multiple)\nmicroarchitectures that give rise to a given set of properties. The limited\nresolution of additive manufacturing techniques often requires solving such\ninverse problems for specific sizes. One should, therefore, find multiple\nmicroarchitectural designs that exhibit the desired properties for a specimen\nwith given dimensions. Moreover, the candidate microarchitectures should be\nresistant to fatigue and fracture, meaning that peak stresses should be\nminimized as well. Such a multi-objective inverse design problem is formidably\ndifficult to solve but its solution is the key to real-world applications of\nmechanical metamaterials. Here, we propose a modular approach titled\n'Deep-DRAM' that combines four decoupled models, including two deep learning\nmodels (DLM), a deep generative model (DGM) based on conditional variational\nautoencoders (CVAE), and direct finite element (FE) simulations. Deep-DRAM\n(deep learning for the design of random-network metamaterials) integrates these\nmodels into a unified framework capable of finding many solutions to the\nmulti-objective inverse design problem posed here. The integrated framework\nfirst introduces the desired elastic properties to the DGM, which returns a set\nof candidate designs. The candidate designs, together with the target specimen\ndimensions are then passed to the DLM which predicts their actual elastic\nproperties considering the specimen size. After a filtering step based on the\ncloseness of the actual properties to the desired ones, the last step uses\ndirect FE simulations to identify the designs with the minimum peak stresses.\n", "date": "2022-12-22", "categories": []}
{"id": "2209.00055", "title": "Computational design of antimicrobial active surfaces via automated\n  Bayesian optimization", "abstract": "  Biofilms pose significant problems for engineers in diverse fields, such as\nmarine science, bioenergy, and biomedicine, where effective biofilm control is\na long-term goal. The adhesion and surface mechanics of biofilms play crucial\nroles in generating and removing biofilm. Designing customized nano-surfaces\nwith different surface topologies can alter the adhesive properties to remove\nbiofilms more easily and greatly improve long-term biofilm control. To rapidly\ndesign such topologies, we employ individual-based modeling and Bayesian\noptimization to automate the design process and generate different active\nsurfaces for effective biofilm removal. Our framework successfully generated\nideal nano-surfaces for biofilm removal through applied shear and vibration.\nDensely distributed short pillar topography is the optimal geometry to prevent\nbiofilm formation. Under fluidic shearing, the optimal topography is to\nsparsely distribute tall, slim, pillar-like structures. When subjected to\neither vertical or lateral vibrations, thick trapezoidal cones are found to be\noptimal. Optimizing the vibrational loading indicates a small vibration\nmagnitude with relatively low frequencies is more efficient in removing\nbiofilm. Our results provide insights into various engineering fields that\nrequire surface-mediated biofilm control. Our framework can also be applied to\nmore general materials design and optimization.\n", "date": "2022-08-31", "categories": []}
{"id": "2212.01136", "title": "Robustness in Fatigue Strength Estimation", "abstract": "  Fatigue strength estimation is a costly manual material characterization\nprocess in which state-of-the-art approaches follow a standardized experiment\nand analysis procedure. In this paper, we examine a modular, Machine\nLearning-based approach for fatigue strength estimation that is likely to\nreduce the number of experiments and, thus, the overall experimental costs.\nDespite its high potential, deployment of a new approach in a real-life lab\nrequires more than the theoretical definition and simulation. Therefore, we\nstudy the robustness of the approach against misspecification of the prior and\ndiscretization of the specified loads. We identify its applicability and its\nadvantageous behavior over the state-of-the-art methods, potentially reducing\nthe number of costly experiments.\n", "date": "2022-12-02", "categories": []}
{"id": "2211.16684", "title": "Capturing long-range interaction with reciprocal space neural network", "abstract": "  Machine Learning (ML) interatomic models and potentials have been widely\nemployed in simulations of materials. Long-range interactions often dominate in\nsome ionic systems whose dynamics behavior is significantly influenced.\nHowever, the long-range effect such as Coulomb and Van der Wales potential is\nnot considered in most ML interatomic potentials. To address this issue, we put\nforward a method that can take long-range effects into account for most ML\nlocal interatomic models with the reciprocal space neural network. The\nstructure information in real space is firstly transformed into reciprocal\nspace and then encoded into a reciprocal space potential or a global descriptor\nwith full atomic interactions. The reciprocal space potential and descriptor\nkeep full invariance of Euclidean symmetry and choice of the cell. Benefiting\nfrom the reciprocal-space information, ML interatomic models can be extended to\ndescribe the long-range potential including not only Coulomb but any other\nlong-range interaction. A model NaCl system considering Coulomb interaction and\nthe GaxNy system with defects are applied to illustrate the advantage of our\napproach. At the same time, our approach helps to improve the prediction\naccuracy of some global properties such as the band gap where the full atomic\ninteraction beyond local atomic environments plays a very important role. In\nsummary, our work has expanded the ability of current ML interatomic models and\npotentials when dealing with the long-range effect, hence paving a new way for\naccurate prediction of global properties and large-scale dynamic simulations of\nsystems with defects.\n", "date": "2022-11-30", "categories": []}
{"id": "2211.15895", "title": "Composition based oxidation state prediction of materials using deep\n  learning", "abstract": "  Oxidation states are the charges of atoms after their ionic approximation of\ntheir bonds, which have been widely used in charge-neutrality verification,\ncrystal structure determination, and reaction estimation. Currently only\nheuristic rules exist for guessing the oxidation states of a given compound\nwith many exceptions. Recent work has developed machine learning models based\non heuristic structural features for predicting the oxidation states of metal\nions. However, composition based oxidation state prediction still remains\nelusive so far, which is more important in new material discovery for which the\nstructures are not even available. This work proposes a novel deep learning\nbased BERT transformer language model BERTOS for predicting the oxidation\nstates of all elements of inorganic compounds given only their chemical\ncomposition. Our model achieves 96.82\\% accuracy for all-element oxidation\nstates prediction benchmarked on the cleaned ICSD dataset and achieves 97.61\\%\naccuracy for oxide materials. We also demonstrate how it can be used to conduct\nlarge-scale screening of hypothetical material compositions for materials\ndiscovery.\n", "date": "2022-11-29", "categories": []}
{"id": "2211.16443", "title": "Synthetic data enable experiments in atomistic machine learning", "abstract": "  Machine-learning models are increasingly used to predict properties of atoms\nin chemical systems. There have been major advances in developing descriptors\nand regression frameworks for this task, typically starting from (relatively)\nsmall sets of quantum-mechanical reference data. Larger datasets of this kind\nare becoming available, but remain expensive to generate. Here we demonstrate\nthe use of a large dataset that we have \"synthetically\" labelled with per-atom\nenergies from an existing ML potential model. The cheapness of this process,\ncompared to the quantum-mechanical ground truth, allows us to generate millions\nof datapoints, in turn enabling rapid experimentation with atomistic ML models\nfrom the small- to the large-data regime. This approach allows us here to\ncompare regression frameworks in depth, and to explore visualisation based on\nlearned representations. We also show that learning synthetic data labels can\nbe a useful pre-training task for subsequent fine-tuning on small datasets. In\nthe future, we expect that our open-sourced dataset, and similar ones, will be\nuseful in rapidly exploring deep-learning models in the limit of abundant\nchemical data.\n", "date": "2022-11-29", "categories": []}
{"id": "2211.12773", "title": "Learning Regularized Positional Encoding for Molecular Prediction", "abstract": "  Machine learning has become a promising approach for molecular modeling.\nPositional quantities, such as interatomic distances and bond angles, play a\ncrucial role in molecule physics. The existing works rely on careful manual\ndesign of their representation. To model the complex nonlinearity in predicting\nmolecular properties in an more end-to-end approach, we propose to encode the\npositional quantities with a learnable embedding that is continuous and\ndifferentiable. A regularization technique is employed to encourage embedding\nsmoothness along the physical dimension. We experiment with a variety of\nmolecular property and force field prediction tasks. Improved performance is\nobserved for three different model architectures after plugging in the proposed\npositional encoding method. In addition, the learned positional encoding allows\neasier physics-based interpretation. We observe that tasks of similar physics\nhave the similar learned positional encoding.\n", "date": "2022-11-23", "categories": []}
{"id": "2211.07587", "title": "Artificial neural networks for predicting the viscosity of\n  lead-containing glasses", "abstract": "  The viscosity of lead-containing glasses is of fundamental importance for the\nmanufacturing process, and can be predicted by algorithms such as artificial\nneural networks. The SciGlass database was used to provide training, validation\nand test data of chemical composition, temperature and viscosity for the\nconstruction of artificial neural networks with node variation in the hidden\nlayer. The best model built with training data and validation data was compared\nwith 7 other models from the literature, demonstrating better statistical\nevaluations of mean absolute error and coefficient of determination to the test\ndata, with subsequent sensitivity analysis in agreement with the literature.\nSkewness and kurtosis were calculated and there is a good correlation between\nthe values predicted by the best neural network built with the test data.\n", "date": "2022-11-11", "categories": []}
{"id": "2211.11543", "title": "Leveraging Orbital Information and Atomic Feature in Deep Learning Model", "abstract": "  Predicting material properties base on micro structure of materials has long\nbeen a challenging problem. Recently many deep learning methods have been\ndeveloped for material property prediction. In this study, we propose a crystal\nrepresentation learning framework, Orbital CrystalNet, OCrystalNet, which\nconsists of two parts: atomic descriptor generation and graph representation\nlearning. In OCrystalNet, we first incorporate orbital field matrix (OFM) and\natomic features to construct OFM-feature atomic descriptor, and then the atomic\ndescriptor is used as atom embedding in the atom-bond message passing module\nwhich takes advantage of the topological structure of crystal graphs to learn\ncrystal representation. To demonstrate the capabilities of OCrystalNet we\nperformed a number of prediction tasks on Material Project dataset and JARVIS\ndataset and compared our model with other baselines and state of art methods.\nTo further present the effectiveness of OCrystalNet, we conducted ablation\nstudy and case study of our model. The results show that our model have various\nadvantages over other state of art models.\n", "date": "2022-10-29", "categories": []}
{"id": "2206.12159", "title": "Data-driven discovery of novel 2D materials by deep generative models", "abstract": "  Efficient algorithms to generate candidate crystal structures with good\nstability properties can play a key role in data-driven materials discovery.\nHere we show that a crystal diffusion variational autoencoder (CDVAE) is\ncapable of generating two-dimensional (2D) materials of high chemical and\nstructural diversity and formation energies mirroring the training structures.\nSpecifically, we train the CDVAE on 2615 2D materials with energy above the\nconvex hull $\\Delta H_{\\mathrm{hull}}< 0.3$ eV/atom, and generate 5003\nmaterials that we relax using density functional theory (DFT). We also generate\n14192 new crystals by systematic element substitution of the training\nstructures. We find that the generative model and lattice decoration approach\nare complementary and yield materials with similar stability properties but\nvery different crystal structures and chemical compositions. In total we find\n11630 predicted new 2D materials, where 8599 of these have $\\Delta\nH_{\\mathrm{hull}}< 0.3$ eV/atom as the seed structures, while 2004 are within\n50 meV of the convex hull and could potentially be synthesized. The relaxed\natomic structures of all the materials are available in the open Computational\n2D Materials Database (C2DB). Our work establishes the CDVAE as an efficient\nand reliable crystal generation machine, and significantly expands the space of\n2D materials.\n", "date": "2022-06-24", "categories": []}
{"id": "2207.12118", "title": "Machine-learning accelerated identification of exfoliable\n  two-dimensional materials", "abstract": "  Two-dimensional (2D) materials have been a central focus of recent research\nbecause they host a variety of properties, making them attractive both for\nfundamental science and for applications. It is thus crucial to be able to\nidentify accurately and efficiently if bulk three-dimensional (3D) materials\nare formed by layers held together by a weak binding energy that, thus, can be\npotentially exfoliated into 2D materials. In this work, we develop a\nmachine-learning (ML) approach that, combined with a fast preliminary\ngeometrical screening, is able to efficiently identify potentially exfoliable\nmaterials. Starting from a combination of descriptors for crystal structures,\nwe work out a subset of them that are crucial for accurate predictions. Our\nfinal ML model, based on a random forest classifier, has a very high recall of\n98\\%. Using a SHapely Additive exPlanations (SHAP) analysis, we also provide an\nintuitive explanation of the five most important variables of the model.\nFinally, we compare the performance of our best ML model with a deep neural\nnetwork architecture using the same descriptors. To make our algorithms and\nmodels easily accessible, we publish an online tool on the Materials Cloud\nportal that only requires a bulk 3D crystal structure as input. Our tool thus\nprovides a practical yet straightforward approach to assess whether any 3D\ncompound can be exfoliated into 2D layers.\n", "date": "2022-07-18", "categories": []}
{"id": "2211.09727", "title": "A Survey on Evaluation Metrics for Synthetic Material Micro-Structure\n  Images from Generative Models", "abstract": "  The evaluation of synthetic micro-structure images is an emerging problem as\nmachine learning and materials science research have evolved together. Typical\nstate of the art methods in evaluating synthetic images from generative models\nhave relied on the Fr\\'echet Inception Distance. However, this and other\nsimilar methods, are limited in the materials domain due to both the unique\nfeatures that characterize physically accurate micro-structures and limited\ndataset sizes. In this study we evaluate a variety of methods on scanning\nelectron microscope (SEM) images of graphene-reinforced polyurethane foams. The\nprimary objective of this paper is to report our findings with regards to the\nshortcomings of existing methods so as to encourage the machine learning\ncommunity to consider enhancements in metrics for assessing quality of\nsynthetic images in the material science domain.\n", "date": "2022-11-03", "categories": []}
{"id": "2211.08506", "title": "ParticleGrid: Enabling Deep Learning using 3D Representation of\n  Materials", "abstract": "  From AlexNet to Inception, autoencoders to diffusion models, the development\nof novel and powerful deep learning models and learning algorithms has\nproceeded at breakneck speeds. In part, we believe that rapid iteration of\nmodel architecture and learning techniques by a large community of researchers\nover a common representation of the underlying entities has resulted in\ntransferable deep learning knowledge. As a result, model scale, accuracy,\nfidelity, and compute performance have dramatically increased in computer\nvision and natural language processing. On the other hand, the lack of a common\nrepresentation for chemical structure has hampered similar progress. To enable\ntransferable deep learning, we identify the need for a robust 3-dimensional\nrepresentation of materials such as molecules and crystals. The goal is to\nenable both materials property prediction and materials generation with 3D\nstructures. While computationally costly, such representations can model a\nlarge set of chemical structures. We propose $\\textit{ParticleGrid}$, a\nSIMD-optimized library for 3D structures, that is designed for deep learning\napplications and to seamlessly integrate with deep learning frameworks. Our\nhighly optimized grid generation allows for generating grids on the fly on the\nCPU, reducing storage and GPU compute and memory requirements. We show the\nefficacy of 3D grids generated via $\\textit{ParticleGrid}$ and accurately\npredict molecular energy properties using a 3D convolutional neural network.\nOur model is able to get 0.006 mean square error and nearly match the values\ncalculated using computationally costly density functional theory at a fraction\nof the time.\n", "date": "2022-11-15", "categories": []}
{"id": "2211.02235", "title": "Materials Property Prediction with Uncertainty Quantification: A\n  Benchmark Study", "abstract": "  Uncertainty quantification (UQ) has increasing importance in building robust\nhigh-performance and generalizable materials property prediction models. It can\nalso be used in active learning to train better models by focusing on getting\nnew training data from uncertain regions. There are several categories of UQ\nmethods each considering different types of uncertainty sources. Here we\nconduct a comprehensive evaluation on the UQ methods for graph neural network\nbased materials property prediction and evaluate how they truly reflect the\nuncertainty that we want in error bound estimation or active learning. Our\nexperimental results over four crystal materials datasets (including formation\nenergy, adsorption energy, total energy, and band gap properties) show that the\npopular ensemble methods for uncertainty estimation is NOT the best choice for\nUQ in materials property prediction. For the convenience of the community, all\nthe source code and data sets can be accessed freely at\n\\url{https://github.com/usccolumbia/materialsUQ}.\n", "date": "2022-11-04", "categories": []}
{"id": "2210.01120", "title": "Predicting CO$_2$ Absorption in Ionic Liquids with Molecular Descriptors\n  and Explainable Graph Neural Networks", "abstract": "  Ionic Liquids (ILs) provide a promising solution for CO$_2$ capture and\nstorage to mitigate global warming. However, identifying and designing the\nhigh-capacity IL from the giant chemical space requires expensive, and\nexhaustive simulations and experiments. Machine learning (ML) can accelerate\nthe process of searching for desirable ionic molecules through accurate and\nefficient property predictions in a data-driven manner. But existing\ndescriptors and ML models for the ionic molecule suffer from the inefficient\nadaptation of molecular graph structure. Besides, few works have investigated\nthe explainability of ML models to help understand the learned features that\ncan guide the design of efficient ionic molecules. In this work, we develop\nboth fingerprint-based ML models and Graph Neural Networks (GNNs) to predict\nthe CO$_2$ absorption in ILs. Fingerprint works on graph structure at the\nfeature extraction stage, while GNNs directly handle molecule structure in both\nthe feature extraction and model prediction stage. We show that our method\noutperforms previous ML models by reaching a high accuracy (MAE of 0.0137,\n$R^2$ of 0.9884). Furthermore, we take the advantage of GNNs feature\nrepresentation and develop a substructure-based explanation method that\nprovides insight into how each chemical fragments within IL molecules\ncontribute to the CO$_2$ absorption prediction of ML models. We also show that\nour explanation result agrees with some ground truth from the theoretical\nreaction mechanism of CO$_2$ absorption in ILs, which can advise on the design\nof novel and efficient functional ILs in the future.\n", "date": "2022-09-29", "categories": []}
{"id": "2211.03075", "title": "Prediction of superconducting properties of materials based on machine\n  learning models", "abstract": "  The application of superconducting materials is becoming more and more\nwidespread. Traditionally, the discovery of new superconducting materials\nrelies on the experience of experts and a large number of \"trial and error\"\nexperiments, which not only increases the cost of experiments but also prolongs\nthe period of discovering new superconducting materials. In recent years,\nmachine learning has been increasingly applied to materials science. Based on\nthis, this manuscript proposes the use of XGBoost model to identify\nsuperconductors; the first application of deep forest model to predict the\ncritical temperature of superconductors; the first application of deep forest\nto predict the band gap of materials; and application of a new sub-network\nmodel to predict the Fermi energy level of materials. Compared with our known\nsimilar literature, all the above algorithms reach state-of-the-art. Finally,\nthis manuscript uses the above models to search the COD public dataset and\nidentify 50 candidate superconducting materials with possible critical\ntemperature greater than 90 K.\n", "date": "2022-11-06", "categories": []}
{"id": "2211.01583", "title": "Data-based Polymer-Unit Fingerprint (PUFp): A Newly Accessible\n  Expression of Polymer Organic Semiconductors for Machine Learning", "abstract": "  In the process of finding high-performance organic semiconductors (OSCs), it\nis of paramount importance in material development to identify important\nfunctional units that play key roles in material performance and subsequently\nestablish substructure-property relationships. Herein, we describe a\npolymer-unit fingerprint (PUFp) generation framework. Machine learning (ML)\nmodels can be used to determine structure-mobility relationships by using PUFp\ninformation as structural input with 678 pieces of collected OSC data. A\npolymer-unit library consisting of 445 units is constructed, and the key\npolymer units for the mobility of OSCs are identified. By investigating the\ncombinations of polymer units with mobility performance, a scheme for designing\npolymer OSC materials by combining ML approaches and PUFp information is\nproposed to not only passively predict OSC mobility but also actively provide\nstructural guidance for new high-mobility OSC material design. The proposed\nscheme demonstrates the ability to screen new materials through pre-evaluation\nand classification ML steps and is an alternative methodology for applying ML\nin new high-mobility OSC discovery.\n", "date": "2022-11-03", "categories": []}
{"id": "2210.16994", "title": "Comparison of two artificial neural networks trained for the surrogate\n  modeling of stress in materially heterogeneous elastoplastic solids", "abstract": "  The purpose of this work is the systematic comparison of the application of\ntwo artificial neural networks (ANNs) to the surrogate modeling of the stress\nfield in materially heterogeneous periodic polycrystalline microstructures. The\nfirst ANN is a UNet-based convolutional neural network (CNN) for periodic data,\nand the second is based on Fourier neural operators (FNO). Both of these were\ntrained, validated, and tested with results from the numerical solution of the\nboundary-value problem (BVP) for quasi-static mechanical equilibrium in\nperiodic grain microstructures with square domains. More specifically, these\nANNs were trained to correlate the spatial distribution of material properties\nwith the equilibrium stress field under uniaxial tensile loading. The resulting\ntrained ANNs (tANNs) calculate the stress field for a given microstructure on\nthe order of 1000 (UNet) to 2500 (FNO) times faster than the numerical solution\nof the corresponding BVP.\n  For microstructures in the test dataset, the FNO-based tANN, or simply FNO,\nis more accurate than its UNet-based counterpart; the normalized mean absolute\nerror of different stress components for the former is 0.25-0.40% as compared\nto 1.41-2.15% for the latter. Errors in FNO are restricted to grain boundary\nregions, whereas the error in U-Net also comes from within the grain. In\ncomparison to U-Net, errors in FNO are more robust to large variations in\nspatial resolution as well as small variations in grain density. On other hand,\nerrors in U-Net are robust to variations in boundary box aspect ratio, whereas\nerrors in FNO increase as the domain becomes rectangular. Both tANNs are\nhowever unable to reproduce strong stress gradients, especially around regions\nof stress concentration.\n", "date": "2022-10-31", "categories": []}
{"id": "2210.13453", "title": "A multi-category inverse design neural network and its application to\n  diblock copolymers", "abstract": "  In this work, we design a multi-category inverse design neural network to map\nordered periodic structure to physical parameters. The neural network model\nconsists of two parts, a classifier and Structure-Parameter-Mapping (SPM)\nsubnets. The classifier is used to identify structure, and the SPM subnets are\nused to predict physical parameters for desired structures. We also present an\nextensible reciprocal-space data augmentation method to guarantee the rotation\nand translation invariant of periodic structures. We apply the proposed network\nmodel and data augmentation method to two-dimensional diblock copolymers based\non the Landau-Brazovskii model. Results show that the multi-category inverse\ndesign neural network is high accuracy in predicting physical parameters for\ndesired structures. Moreover, the idea of multi-categorization can also be\nextended to other inverse design problems.\n", "date": "2022-10-12", "categories": []}
{"id": "2210.14188", "title": "MOFormer: Self-Supervised Transformer model for Metal-Organic Framework\n  Property Prediction", "abstract": "  Metal-Organic Frameworks (MOFs) are materials with a high degree of porosity\nthat can be used for applications in energy storage, water desalination, gas\nstorage, and gas separation. However, the chemical space of MOFs is close to an\ninfinite size due to the large variety of possible combinations of building\nblocks and topology. Discovering the optimal MOFs for specific applications\nrequires an efficient and accurate search over an enormous number of potential\ncandidates. Previous high-throughput screening methods using computational\nsimulations like DFT can be time-consuming. Such methods also require\noptimizing 3D atomic structure of MOFs, which adds one extra step when\nevaluating hypothetical MOFs. In this work, we propose a structure-agnostic\ndeep learning method based on the Transformer model, named as MOFormer, for\nproperty predictions of MOFs. The MOFormer takes a text string representation\nof MOF (MOFid) as input, thus circumventing the need of obtaining the 3D\nstructure of hypothetical MOF and accelerating the screening process.\nFurthermore, we introduce a self-supervised learning framework that pretrains\nthe MOFormer via maximizing the cross-correlation between its\nstructure-agnostic representations and structure-based representations of\ncrystal graph convolutional neural network (CGCNN) on >400k publicly available\nMOF data. Using self-supervised learning allows the MOFormer to intrinsically\nlearn 3D structural information though it is not included in the input.\nExperiments show that pretraining improved the prediction accuracy of both\nmodels on various downstream prediction tasks. Furthermore, we revealed that\nMOFormer can be more data-efficient on quantum-chemical property prediction\nthan structure-based CGCNN when training data is limited. Overall, MOFormer\nprovides a novel perspective on efficient MOF design using deep learning.\n", "date": "2022-10-25", "categories": []}
{"id": "2210.14191", "title": "A Database of Ultrastable MOFs Reassembled from Stable Fragments with\n  Machine Learning Models", "abstract": "  High-throughput screening of large hypothetical databases of metal-organic\nframeworks (MOFs) can uncover new materials, but their stability in real-world\napplications is often unknown. We leverage community knowledge and machine\nlearning (ML) models to identify MOFs that are thermally stable and stable upon\nactivation. We separate these MOFs into their building blocks and recombine\nthem to make a new hypothetical MOF database of over 50,000 structures that\nsamples orders of magnitude more connectivity nets and inorganic building\nblocks than prior databases. This database shows an order of magnitude\nenrichment of ultrastable MOF structures that are stable upon activation and\nmore than one standard deviation more thermally stable than the average\nexperimentally characterized MOF. For the nearly 10,000 ultrastable MOFs, we\ncompute bulk elastic moduli to confirm these materials have good mechanical\nstability, and we report methane deliverable capacities. Our work identifies\nprivileged metal nodes in ultrastable MOFs that optimize gas storage and\nmechanical stability simultaneously.\n", "date": "2022-10-25", "categories": []}
{"id": "2210.11931", "title": "Deep Reinforcement Learning for Inverse Inorganic Materials Design", "abstract": "  A major obstacle to the realization of novel inorganic materials with\ndesirable properties is the inability to perform efficient optimization across\nboth materials properties and synthesis of those materials. In this work, we\npropose a reinforcement learning (RL) approach to inverse inorganic materials\ndesign, which can identify promising compounds with specified properties and\nsynthesizability constraints. Our model learns chemical guidelines such as\ncharge and electronegativity neutrality while maintaining chemical diversity\nand uniqueness. We demonstrate a multi-objective RL approach, which can\ngenerate novel compounds with targeted materials properties including formation\nenergy and bulk/shear modulus alongside a lower sintering temperature synthesis\nobjectives. Using this approach, the model can predict promising compounds of\ninterest, while suggesting an optimized chemical design space for inorganic\nmaterials discovery.\n", "date": "2022-10-21", "categories": []}
{"id": "2210.09783", "title": "Machine-Learning-Optimized Perovskite Nanoplatelet Synthesis", "abstract": "  With the demand for renewable energy and efficient devices rapidly\nincreasing, a need arises to find and optimize novel (nano)materials. This can\nbe an extremely tedious process, often relying significantly on trial and\nerror. Machine learning has emerged recently as a powerful alternative;\nhowever, most approaches require a substantial amount of data points, i.e.,\nsyntheses. Here, we merge three machine-learning models with Bayesian\nOptimization and are able to dramatically improve the quality of CsPbBr3\nnanoplatelets (NPLs) using only approximately 200 total syntheses. The\nalgorithm can predict the resulting PL emission maxima of the NPL dispersions\nbased on the precursor ratios, which lead to previously unobtainable 7 and 8 ML\nNPLs. Aided by heuristic knowledge, the algorithm should be easily applicable\nto other nanocrystal syntheses and significantly help to identify interesting\ncompositions and rapidly improve their quality.\n", "date": "2022-10-18", "categories": []}
{"id": "2210.08047", "title": "Injecting Domain Knowledge from Empirical Interatomic Potentials to\n  Neural Networks for Predicting Material Properties", "abstract": "  For decades, atomistic modeling has played a crucial role in predicting the\nbehavior of materials in numerous fields ranging from nanotechnology to drug\ndiscovery. The most accurate methods in this domain are rooted in\nfirst-principles quantum mechanical calculations such as density functional\ntheory (DFT). Because these methods have remained computationally prohibitive,\npractitioners have traditionally focused on defining physically motivated\nclosed-form expressions known as empirical interatomic potentials (EIPs) that\napproximately model the interactions between atoms in materials. In recent\nyears, neural network (NN)-based potentials trained on quantum mechanical\n(DFT-labeled) data have emerged as a more accurate alternative to conventional\nEIPs. However, the generalizability of these models relies heavily on the\namount of labeled training data, which is often still insufficient to generate\nmodels suitable for general-purpose applications. In this paper, we propose two\ngeneric strategies that take advantage of unlabeled training instances to\ninject domain knowledge from conventional EIPs to NNs in order to increase\ntheir generalizability. The first strategy, based on weakly supervised\nlearning, trains an auxiliary classifier on EIPs and selects the\nbest-performing EIP to generate energies to supplement the ground-truth DFT\nenergies in training the NN. The second strategy, based on transfer learning,\nfirst pretrains the NN on a large set of easily obtainable EIP energies, and\nthen fine-tunes it on ground-truth DFT energies. Experimental results on three\nbenchmark datasets demonstrate that the first strategy improves baseline NN\nperformance by 5% to 51% while the second improves baseline performance by up\nto 55%. Combining them further boosts performance.\n", "date": "2022-10-14", "categories": []}
{"id": "2206.14331", "title": "Spherical Channels for Modeling Atomic Interactions", "abstract": "  Modeling the energy and forces of atomic systems is a fundamental problem in\ncomputational chemistry with the potential to help address many of the world's\nmost pressing problems, including those related to energy scarcity and climate\nchange. These calculations are traditionally performed using Density Functional\nTheory, which is computationally very expensive. Machine learning has the\npotential to dramatically improve the efficiency of these calculations from\ndays or hours to seconds. We propose the Spherical Channel Network (SCN) to\nmodel atomic energies and forces. The SCN is a graph neural network where nodes\nrepresent atoms and edges their neighboring atoms. The atom embeddings are a\nset of spherical functions, called spherical channels, represented using\nspherical harmonics. We demonstrate, that by rotating the embeddings based on\nthe 3D edge orientation, more information may be utilized while maintaining the\nrotational equivariance of the messages. While equivariance is a desirable\nproperty, we find that by relaxing this constraint in both message passing and\naggregation, improved accuracy may be achieved. We demonstrate state-of-the-art\nresults on the large-scale Open Catalyst dataset in both energy and force\nprediction for numerous tasks and metrics.\n", "date": "2022-06-29", "categories": []}
{"id": "2210.06541", "title": "MicroLib: A library of 3D microstructures generated from 2D micrographs\n  using SliceGAN", "abstract": "  3D microstructural datasets are commonly used to define the geometrical\ndomains used in finite element modelling. This has proven a useful tool for\nunderstanding how complex material systems behave under applied stresses,\ntemperatures and chemical conditions. However, 3D imaging of materials is\nchallenging for a number of reasons, including limited field of view, low\nresolution and difficult sample preparation. Recently, a machine learning\nmethod, SliceGAN, was developed to statistically generate 3D microstructural\ndatasets of arbitrary size using a single 2D input slice as training data. In\nthis paper, we present the results from applying SliceGAN to 87 different\nmicrostructures, ranging from biological materials to high-strength steels. To\ndemonstrate the accuracy of the synthetic volumes created by SliceGAN, we\ncompare three microstructural properties between the 2D training data and 3D\ngenerations, which show good agreement. This new microstructure library both\nprovides valuable 3D microstructures that can be used in models, and also\ndemonstrates the broad applicability of the SliceGAN algorithm.\n", "date": "2022-10-12", "categories": []}
{"id": "2210.06622", "title": "When does deep learning fail and how to tackle it? A critical analysis\n  on polymer sequence-property surrogate models", "abstract": "  Deep learning models are gaining popularity and potency in predicting polymer\nproperties. These models can be built using pre-existing data and are useful\nfor the rapid prediction of polymer properties. However, the performance of a\ndeep learning model is intricately connected to its topology and the volume of\ntraining data. There is no facile protocol available to select a deep learning\narchitecture, and there is a lack of a large volume of homogeneous\nsequence-property data of polymers. These two factors are the primary\nbottleneck for the efficient development of deep learning models. Here we\nassess the severity of these factors and propose new algorithms to address\nthem. We show that a linear layer-by-layer expansion of a neural network can\nhelp in identifying the best neural network topology for a given problem.\nMoreover, we map the discrete sequence space of a polymer to a continuous\none-dimensional latent space using a machine learning pipeline to identify\nminimal data points for building a universal deep learning model. We implement\nthese approaches for three representative cases of building sequence-property\nsurrogate models, viz., the single-molecule radius of gyration of a copolymer,\nadhesive free energy of a copolymer, and copolymer compatibilizer,\ndemonstrating the generality of the proposed strategies. This work establishes\nefficient methods for building universal deep learning models with minimal data\nand hyperparameters for predicting sequence-defined properties of polymers.\n", "date": "2022-10-12", "categories": []}
{"id": "2210.00579", "title": "Large-scale machine-learning-assisted exploration of the whole materials\n  space", "abstract": "  Crystal-graph attention networks have emerged recently as remarkable tools\nfor the prediction of thermodynamic stability and materials properties from\nunrelaxed crystal structures. Previous networks trained on two million\nmaterials exhibited, however, strong biases originating from underrepresented\nchemical elements and structural prototypes in the available data. We tackled\nthis issue computing additional data to provide better balance across both\nchemical and crystal-symmetry space. Crystal-graph networks trained with this\nnew data show unprecedented generalization accuracy, and allow for reliable,\naccelerated exploration of the whole space of inorganic compounds. We applied\nthis universal network to perform machine-learning assisted high-throughput\nmaterials searches including 2500 binary and ternary structure prototypes and\nspanning about 1 billion compounds. After validation using density-functional\ntheory, we uncover in total 19512 additional materials on the convex hull of\nthermodynamic stability and ~150000 compounds with a distance of less than 50\nmeV/atom from the hull. Combining again machine learning and ab-initio methods,\nwe finally evaluate the discovered materials for applications as\nsuperconductors, superhard materials, and we look for candidates with large gap\ndeformation potentials, finding several compounds with extreme values of these\nproperties.\n", "date": "2022-10-02", "categories": []}
{"id": "2210.00842", "title": "A micromechanics-based recurrent neural networks model for\n  path-dependent cyclic deformation of short fiber composites", "abstract": "  The macroscopic response of short fiber reinforced composites is dependent on\nan extensive range of microstructural parameters. Thus, micromechanical\nmodeling of these materials is challenging and in some cases, computationally\nexpensive. This is particularly important when path-dependent plastic behavior\nis needed to be predicted. A solution to this challenge is to enhance\nmicromechanical solutions with machine learning techniques such as artificial\nneural networks. In this work, a recurrent deep neural network model is trained\nto predict the path-dependent elasto-plastic stress response of short fiber\nreinforced composites, given the microstructural parameters and the strain\npath. Micromechanical meanfield simulations are conducted to create a data base\nfor training the validating the model. The model gives very accurate\npredictions in a computationally efficient manner when compared with\nindependent micromechanical simulations.\n", "date": "2022-09-27", "categories": []}
{"id": "2209.12948", "title": "Developing Machine-Learned Potentials for Coarse-Grained Molecular\n  Simulations: Challenges and Pitfalls", "abstract": "  Coarse graining (CG) enables the investigation of molecular properties for\nlarger systems and at longer timescales than the ones attainable at the\natomistic resolution. Machine learning techniques have been recently proposed\nto learn CG particle interactions, i.e. develop CG force fields. Graph\nrepresentations of molecules and supervised training of a graph convolutional\nneural network architecture are used to learn the potential of mean force\nthrough a force matching scheme. In this work, the force acting on each CG\nparticle is correlated to a learned representation of its local environment\nthat goes under the name of SchNet, constructed via continuous filter\nconvolutions. We explore the application of SchNet models to obtain a CG\npotential for liquid benzene, investigating the effect of model architecture\nand hyperparameters on the thermodynamic, dynamical, and structural properties\nof the simulated CG systems, reporting and discussing challenges encountered\nand future directions envisioned.\n", "date": "2022-09-26", "categories": []}
{"id": "2209.11807", "title": "Periodic Graph Transformers for Crystal Material Property Prediction", "abstract": "  We consider representation learning on periodic graphs encoding crystal\nmaterials. Different from regular graphs, periodic graphs consist of a minimum\nunit cell repeating itself on a regular lattice in 3D space. How to effectively\nencode these periodic structures poses unique challenges not present in regular\ngraph representation learning. In addition to being E(3) invariant, periodic\ngraph representations need to be periodic invariant. That is, the learned\nrepresentations should be invariant to shifts of cell boundaries as they are\nartificially imposed. Furthermore, the periodic repeating patterns need to be\ncaptured explicitly as lattices of different sizes and orientations may\ncorrespond to different materials. In this work, we propose a transformer\narchitecture, known as Matformer, for periodic graph representation learning.\nOur Matformer is designed to be invariant to periodicity and can capture\nrepeating patterns explicitly. In particular, Matformer encodes periodic\npatterns by efficient use of geometric distances between the same atoms in\nneighboring cells. Experimental results on multiple common benchmark datasets\nshow that our Matformer outperforms baseline methods consistently. In addition,\nour results demonstrate the importance of periodic invariance and explicit\nrepeating pattern encoding for crystal representation learning.\n", "date": "2022-09-23", "categories": []}
{"id": "2209.12466", "title": "Learned Force Fields Are Ready For Ground State Catalyst Discovery", "abstract": "  We present evidence that learned density functional theory (``DFT'') force\nfields are ready for ground state catalyst discovery. Our key finding is that\nrelaxation using forces from a learned potential yields structures with similar\nor lower energy to those relaxed using the RPBE functional in over 50\\% of\nevaluated systems, despite the fact that the predicted forces differ\nsignificantly from the ground truth. This has the surprising implication that\nlearned potentials may be ready for replacing DFT in challenging catalytic\nsystems such as those found in the Open Catalyst 2020 dataset. Furthermore, we\nshow that a force field trained on a locally harmonic energy surface with the\nsame minima as a target DFT energy is also able to find lower or similar energy\nstructures in over 50\\% of cases. This ``Easy Potential'' converges in fewer\nsteps than a standard model trained on true energies and forces, which further\naccelerates calculations. Its success illustrates a key point: learned\npotentials can locate energy minima even when the model has high force errors.\nThe main requirement for structure optimisation is simply that the learned\npotential has the correct minima. Since learned potentials are fast and scale\nlinearly with system size, our results open the possibility of quickly finding\nground states for large systems.\n", "date": "2022-09-26", "categories": []}
{"id": "2209.11259", "title": "Computational Discovery of Energy-Efficient Heat Treatment for\n  Microstructure Design using Deep Reinforcement Learning", "abstract": "  Deep Reinforcement Learning (DRL) is employed to develop autonomously\noptimized and custom-designed heat-treatment processes that are both,\nmicrostructure-sensitive and energy efficient. Different from conventional\nsupervised machine learning, DRL does not rely on static neural network\ntraining from data alone, but a learning agent autonomously develops optimal\nsolutions, based on reward and penalty elements, with reduced or no\nsupervision. In our approach, a temperature-dependent Allen-Cahn model for\nphase transformation is used as the environment for the DRL agent, serving as\nthe model world in which it gains experience and takes autonomous decisions.\nThe agent of the DRL algorithm is controlling the temperature of the system, as\na model furnace for heat-treatment of alloys. Microstructure goals are defined\nfor the agent based on the desired microstructure of the phases. After\ntraining, the agent can generate temperature-time profiles for a variety of\ninitial microstructure states to reach the final desired microstructure state.\nThe agent's performance and the physical meaning of the heat-treatment profiles\ngenerated are investigated in detail. In particular, the agent is capable of\ncontrolling the temperature to reach the desired microstructure starting from a\nvariety of initial conditions. This capability of the agent in handling a\nvariety of conditions paves the way for using such an approach also for\nrecycling-oriented heat treatment process design where the initial composition\ncan vary from batch to batch, due to impurity intrusion, and also for the\ndesign of energy-efficient heat treatments. For testing this hypothesis, an\nagent without penalty on the total consumed energy is compared with one that\nconsiders energy costs. The energy cost penalty is imposed as an additional\ncriterion on the agent for finding the optimal temperature-time profile.\n", "date": "2022-09-22", "categories": []}
{"id": "2209.07567", "title": "Prediction of $\\textrm{CO}_2$ Adsorption in Nano-Pores with Graph Neural\n  Networks", "abstract": "  We investigate the graph-based convolutional neural network approach for\npredicting and ranking gas adsorption properties of crystalline Metal-Organic\nFramework (MOF) adsorbents for application in post-combustion capture of\n$\\textrm{CO}_2$. Our model is based solely on standard structural input files\ncontaining atomistic descriptions of the adsorbent material candidates. We\nconstruct novel methodological extensions to match the prediction accuracy of\nclassical machine learning models that were built with hundreds of features at\nmuch higher computational cost. Our approach can be more broadly applied to\noptimize gas capture processes at industrial scale.\n", "date": "2022-08-22", "categories": []}
{"id": "2209.07300", "title": "Multi-Task Mixture Density Graph Neural Networks for Predicting Cu-based\n  Single-Atom Alloy Catalysts for CO2 Reduction Reaction", "abstract": "  Graph neural networks (GNNs) have drawn more and more attention from material\nscientists and demonstrated a high capacity to establish connections between\nthe structure and properties. However, with only unrelaxed structures provided\nas input, few GNN models can predict the thermodynamic properties of relaxed\nconfigurations with an acceptable level of error. In this work, we develop a\nmulti-task (MT) architecture based on DimeNet++ and mixture density networks to\nimprove the performance of such task. Taking CO adsorption on Cu-based\nsingle-atom alloy catalysts as an illustration, we show that our method can\nreliably estimate CO adsorption energy with a mean absolute error of 0.087 eV\nfrom the initial CO adsorption structures without costly first-principles\ncalculations. Further, compared to other state-of-the-art GNN methods, our\nmodel exhibits improved generalization ability when predicting catalytic\nperformance of out-of-domain configurations, built with either unseen substrate\nsurfaces or doping species. We show that the proposed MT GNN strategy can\nfacilitate catalyst discovery.\n", "date": "2022-09-15", "categories": []}
{"id": "2209.04416", "title": "Deep autoencoders for physics-constrained data-driven nonlinear\n  materials modeling", "abstract": "  Physics-constrained data-driven computing is an emerging computational\nparadigm that allows simulation of complex materials directly based on material\ndatabase and bypass the classical constitutive model construction. However, it\nremains difficult to deal with high-dimensional applications and extrapolative\ngeneralization. This paper introduces deep learning techniques under the\ndata-driven framework to address these fundamental issues in nonlinear\nmaterials modeling. To this end, an autoencoder neural network architecture is\nintroduced to learn the underlying low-dimensional representation (embedding)\nof the given material database. The offline trained autoencoder and the\ndiscovered embedding space are then incorporated in the online data-driven\ncomputation such that the search of optimal material state from database can be\nperformed on a low-dimensional space, aiming to enhance the robustness and\npredictability with projected material data. To ensure numerical stability and\nrepresentative constitutive manifold, a convexity-preserving interpolation\nscheme tailored to the proposed autoencoder-based data-driven solver is\nproposed for constructing the material state. In this study, the applicability\nof the proposed approach is demonstrated by modeling nonlinear biological\ntissues. A parametric study on data noise, data size and sparsity, training\ninitialization, and model architectures, is also conducted to examine the\nrobustness and convergence property of the proposed approach.\n", "date": "2022-09-03", "categories": []}
{"id": "2207.12858", "title": "Transition1x -- a Dataset for Building Generalizable Reactive Machine\n  Learning Potentials", "abstract": "  Machine Learning (ML) models have, in contrast to their usefulness in\nmolecular dynamics studies, had limited success as surrogate potentials for\nreaction barrier search. It is due to the scarcity of training data in relevant\ntransition state regions of chemical space. Currently, available datasets for\ntraining ML models on small molecular systems almost exclusively contain\nconfigurations at or near equilibrium. In this work, we present the dataset\nTransition1x containing 9.6 million Density Functional Theory (DFT)\ncalculations of forces and energies of molecular configurations on and around\nreaction pathways at the wB97x/6-31G(d) level of theory. The data was generated\nby running Nudged Elastic Band (NEB) calculations with DFT on 10k reactions\nwhile saving intermediate calculations. We train state-of-the-art equivariant\ngraph message-passing neural network models on Transition1x and cross-validate\non the popular ANI1x and QM9 datasets. We show that ML models cannot learn\nfeatures in transition-state regions solely by training on hitherto popular\nbenchmark datasets. Transition1x is a new challenging benchmark that will\nprovide an important step towards developing next-generation ML force fields\nthat also work far away from equilibrium configurations and reactive systems.\n", "date": "2022-07-25", "categories": []}
{"id": "2208.13742", "title": "Machine Learning guided high-throughput search of non-oxide garnets", "abstract": "  Garnets, known since the early stages of human civilization, have found\nimportant applications in modern technologies including magnetorestriction,\nspintronics, lithium batteries, etc. The overwhelming majority of\nexperimentally known garnets are oxides, while explorations (experimental or\ntheoretical) for the rest of the chemical space have been limited in scope. A\nkey issue is that the garnet structure has a large primitive unit cell,\nrequiring an enormous amount of computational resources. To perform a\ncomprehensive search of the complete chemical space for new garnets,we combine\nrecent progress in graph neural networks with high-throughput calculations. We\napply the machine learning model to identify the potential (meta-)stable garnet\nsystems before systematic density-functional calculations to validate the\npredictions. In this way, we discover more than 600 ternary garnets with\ndistances to the convex hull below 100~meV/atom with a variety of physical and\nchemical properties. This includes sulfide, nitride and halide garnets. For\nthese, we analyze the electronic structure and discuss the connection between\nthe value of the electronic band gap and charge balance.\n", "date": "2022-08-29", "categories": []}
{"id": "2208.12705", "title": "A machine learning approach to predict the structural and magnetic\n  properties of Heusler alloy families", "abstract": "  Random forest (RF) regression model is used to predict the lattice constant,\nmagnetic moment and formation energies of full Heusler alloys, half Heusler\nalloys, inverse Heusler alloys and quaternary Heusler alloys based on existing\nas well as indigenously prepared databases. Prior analysis was carried out to\ncheck the distribution of the data points of the response variables and found\nthat in most of the cases, the data is not normally distributed. The outcome of\nthe RF model performance is sufficiently accurate to predict the response\nvariables on the test data and also shows its robustness against overfitting,\noutliers, multicollinearity and distribution of data points. The parity plots\nbetween the machine learning predicted values against the computed values using\ndensity functional theory (DFT) shows linear behavior with adjusted R2 values\nlying in the range of 0.80 to 0.94 for all the predicted properties for\ndifferent types of Heusler alloys. Feature importance analysis shows that the\nvalence electron numbers plays an important feature role in the prediction for\nmost of the predicted outcomes. Case studies with one full Heusler alloy and\none quaternary Heusler alloy were also mentioned comparing the machine learning\npredicted results with our earlier theoretical calculated values and\nexperimentally measured results, suggesting high accuracy of the model\npredicted results.\n", "date": "2022-08-07", "categories": []}
{"id": "2208.01692", "title": "A cloud platform for automating and sharing analysis of raw simulation\n  data from high throughput polymer molecular dynamics simulations", "abstract": "  Open material databases storing hundreds of thousands of material structures\nand their corresponding properties have become the cornerstone of modern\ncomputational materials science. Yet, the raw outputs of the simulations, such\nas the trajectories from molecular dynamics simulations and charge densities\nfrom density functional theory calculations, are generally not shared due to\ntheir huge size. In this work, we describe a cloud-based platform to facilitate\nthe sharing of raw data and enable the fast post-processing in the cloud to\nextract new properties defined by the user. As an initial demonstration, our\ndatabase currently includes 6286 molecular dynamics trajectories for amorphous\npolymer electrolytes and 5.7 terabytes of data. We create a public analysis\nlibrary at https://github.com/TRI-AMDD/htp_md to extract multiple properties\nfrom the raw data, using both expert designed functions and machine learning\nmodels. The analysis is run automatically with computation in the cloud, and\nresults then populate a database that can be accessed publicly. Our platform\nencourages users to contribute both new trajectory data and analysis functions\nvia public interfaces. Newly analyzed properties will be incorporated into the\ndatabase. Finally, we create a front-end user interface at\nhttps://www.htpmd.matr.io for browsing and visualization of our data. We\nenvision the platform to be a new way of sharing raw data and new insights for\nthe computational materials science community.\n", "date": "2022-08-02", "categories": []}
{"id": "2207.13227", "title": "Atomic structure generation from reconstructing structural fingerprints", "abstract": "  Data-driven machine learning methods have the potential to dramatically\naccelerate the rate of materials design over conventional human-guided\napproaches. These methods would help identify or, in the case of generative\nmodels, even create novel crystal structures of materials with a set of\nspecified functional properties to then be synthesized or isolated in the\nlaboratory. For crystal structure generation, a key bottleneck lies in\ndeveloping suitable atomic structure fingerprints or representations for the\nmachine learning model, analogous to the graph-based or SMILES representations\nused in molecular generation. However, finding data-efficient representations\nthat are invariant to translations, rotations, and permutations, while\nremaining invertible to the Cartesian atomic coordinates remains an ongoing\nchallenge. Here, we propose an alternative approach to this problem by taking\nexisting non-invertible representations with the desired invariances and\ndeveloping an algorithm to reconstruct the atomic coordinates through\ngradient-based optimization using automatic differentiation. This can then be\ncoupled to a generative machine learning model which generates new materials\nwithin the representation space, rather than in the data-inefficient Cartesian\nspace. In this work, we implement this end-to-end structure generation approach\nusing atom-centered symmetry functions as the representation and conditional\nvariational autoencoders as the generative model. We are able to successfully\ngenerate novel and valid atomic structures of sub-nanometer Pt nanoparticles as\na proof of concept. Furthermore, this method can be readily extended to any\nsuitable structural representation, thereby providing a powerful, generalizable\nframework towards structure-based generation.\n", "date": "2022-07-27", "categories": []}
{"id": "2207.06512", "title": "A Generalized Framework for Microstructural Optimization using Neural\n  Networks", "abstract": "  Microstructures, i.e., architected materials, are designed today, typically,\nby maximizing an objective, such as bulk modulus, subject to a volume\nconstraint. However, in many applications, it is often more appropriate to\nimpose constraints on other physical quantities of interest. In this paper, we\nconsider such generalized microstructural optimization problems where any of\nthe microstructural quantities, namely, bulk, shear, Poisson ratio, or volume,\ncan serve as the objective, while the remaining can serve as constraints. In\nparticular, we propose here a neural-network (NN) framework to solve such\nproblems. The framework relies on the classic density formulation of\nmicrostructural optimization, but the density field is represented through the\nNN's weights and biases. The main characteristics of the proposed NN framework\nare: (1) it supports automatic differentiation, eliminating the need for manual\nsensitivity derivations, (2) smoothing filters are not required due to implicit\nfiltering, (3) the framework can be easily extended to multiple-materials, and\n(4) a high-resolution microstructural topology can be recovered through a\nsimple post-processing step. The framework is illustrated through a variety of\nmicrostructural optimization problems.\n", "date": "2022-07-13", "categories": []}
{"id": "2206.08011", "title": "Hardness prediction of age-hardening aluminum alloy based on ensemble\n  learning", "abstract": "  With the rapid development of artificial intelligence, the combination of\nmaterial database and machine learning has driven the progress of material\ninformatics. Because aluminum alloy is widely used in many fields, so it is\nsignificant to predict the properties of aluminum alloy. In this thesis, the\ndata of Al-Cu-Mg-X (X: Zn, Zr, etc.) alloy are used to input the composition,\naging conditions (time and temperature) and predict its hardness. An ensemble\nlearning solution based on automatic machine learning and an attention\nmechanism introduced into the secondary learner of deep neural network are\nproposed respectively. The experimental results show that selecting the correct\nsecondary learner can further improve the prediction accuracy of the model.\nThis manuscript introduces the attention mechanism to improve the secondary\nlearner based on deep neural network, and obtains a fusion model with better\nperformance. The R-Square of the best model is 0.9697 and the MAE is 3.4518HV.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.13578", "title": "Materials Transformers Language Models for Generative Materials Design:\n  a benchmark study", "abstract": "  Pre-trained transformer language models on large unlabeled corpus have\nproduced state-of-the-art results in natural language processing, organic\nmolecule design, and protein sequence generation. However, no such models have\nbeen applied to learn the composition patterns of inorganic materials. Here we\ntrain a series of seven modern transformer language models (GPT, GPT-2,\nGPT-Neo, GPT-J, BLMM, BART, and RoBERTa) using the expanded formulas from\nmaterial deposited in the ICSD, OQMD, and Materials Projects databases. Six\ndifferent datasets with/out non-charge-neutral or balanced electronegativity\nsamples are used to benchmark the performances and uncover the generation\nbiases of modern transformer models for the generative design of materials\ncompositions. Our extensive experiments showed that the causal language models\nbased materials transformers can generate chemically valid materials\ncompositions with as high as 97.54\\% to be charge neutral and 91.40\\% to be\nelectronegativity balanced, which has more than 6 times higher enrichment\ncompared to a baseline pseudo-random sampling algorithm. These models also\ndemonstrate high novelty and their potential in new materials discovery has\nbeen proved by their capability to recover the leave-out materials. We also\nfind that the properties of the generated samples can be tailored by training\nthe models with selected training sets such as high-bandgap materials. Our\nexperiments also showed that different models each have their own preference in\nterms of the properties of the generated samples and their running time\ncomplexity varies a lot. We have applied our materials transformer models to\ndiscover a set of new materials as validated using DFT calculations.\n", "date": "2022-06-27", "categories": []}
{"id": "2206.09002", "title": "Cluster Generation via Deep Energy-Based Model", "abstract": "  We present a new approach for the generation of stable structures of\nnanoclusters using deep learning methods. Our method consists in constructing\nan artificial potential energy surface, with local minima corresponding to the\nmost stable structures and which is much smoother than \"real\" potential in the\nintermediate regions of the configuration space. To build the surface, graph\nconvolutional networks are used. The method can extrapolates the potential\nsurface to cases of structures with larger number of atoms than was used in\ntraining. Thus, having a sufficient number of low-energy structures in the\ntraining set, the method allows to generate new candidates for the ground-state\nstructures, including ones with larger number of atoms. We applied the approach\nto silica clusters $(SiO_2)_n$ and for the first time found the stable\nstructures with n=28...51. The method is universal and does not depend on the\natomic composition and number of atoms.\n", "date": "2022-06-17", "categories": []}
{"id": "2206.08048", "title": "Automated analysis of continuum fields from atomistic simulations using\n  statistical machine learning", "abstract": "  Atomistic simulations of the molecular dynamics/statics kind are regularly\nused to study small scale plasticity. Contemporary simulations are performed\nwith tens to hundreds of millions of atoms, with snapshots of these\nconfigurations written out at regular intervals for further analysis. Continuum\nscale constitutive models for material behavior can benefit from information on\nthe atomic scale, in particular in terms of the deformation mechanisms, the\naccommodation of the total strain and partitioning of stress and strain fields\nin individual grains. In this work we develop a methodology using statistical\ndata mining and machine learning algorithms to automate the analysis of\ncontinuum field variables in atomistic simulations. We focus on three important\nfield variables: total strain, elastic strain and microrotation. Our results\nshow that the elastic strain in individual grains exhibits a unimodal\nlog-normal distribution, whilst the total strain and microrotation fields\nevidence a multimodal distribution. The peaks in the distribution of total\nstrain are identified with a Gaussian mixture model and methods to circumvent\noverfitting problems are presented. Subsequently, we evaluate the identified\npeaks in terms of deformation mechanisms in a grain, which e.g., helps to\nquantify the strain for which individual deformation mechanisms are\nresponsible. The overall statistics of the distributions over all grains are an\nimportant input for higher scale models, which ultimately also helps to be able\nto quantitatively discuss the implications for information transfer to\nphenomenological models.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.02040", "title": "MetaNOR: A Meta-Learnt Nonlocal Operator Regression Approach for\n  Metamaterial Modeling", "abstract": "  We propose MetaNOR, a meta-learnt approach for transfer-learning operators\nbased on the nonlocal operator regression. The overall goal is to efficiently\nprovide surrogate models for new and unknown material-learning tasks with\ndifferent microstructures. The algorithm consists of two phases: (1) learning a\ncommon nonlocal kernel representation from existing tasks; (2) transferring the\nlearned knowledge and rapidly learning surrogate operators for unseen tasks\nwith a different material, where only a few test samples are required. We apply\nMetaNOR to model the wave propagation within 1D metamaterials, showing\nsubstantial improvements on the sampling efficiency for new materials.\n", "date": "2022-06-04", "categories": []}
{"id": "2606.23725", "title": "Computational references are not experiments: pre-registered validation of machine-learned sodium-cathode voltages", "abstract": "Machine-learning screens for battery materials are trained and judged almost entirely against computed reference voltages, and those references carry their own systematic errors. We report a case in which this matters quantitatively: our own screening stack (a graph-network voltage screen, a prior-art triage layer, and a local PBE+U bench) fails pre-registered validation against experiment-anchored literature values. Verdict thresholds, failure modes, and the primary metric were committed before analysis. On an operator-audited set of known Na-ion cathodes (n = 6 after one documented exclusion; verdict unchanged at n = 7), the raw held-out mean absolute error was 0.67 V, the pre-registered conservative metric, the upper 95% confidence bound of the cross-validated bias-corrected error, was 1.09 V, and the residual was strongly voltage-dependent (r = -0.94), so no additive calibration is valid. On the two compounds where prediction, database reference, and experiment could all be compared, the Materials Project PBE+U reference sat about 0.54 V below measurement: the reference, not the model, dominated the error. A prior-art screen found at least 70% of the targeted Na substitution space already published. We retire the screen, bound what \"verified\" means for our DFT ledger, and pre-register a calibration audit of it against four benchmark Li couples.", "date": "2026-06-19", "categories": []}
{"id": "2606.22425", "title": "SVGym (SciVerseGym): An Environment for Reinforcement Learning and Bayesian Optimization in Crystal Discovery", "abstract": "Machine-learned interatomic potentials now enable efficient atomistic evaluation for interactive materials discovery, yet closed-loop crystal search methods remain fragmented across bespoke pipelines for editing, relaxation, scoring, constraints, and bookkeeping. We introduce SciVerseGym, a Gymnasium-compatible environment for sequential crystal discovery that frames crystal design as a Markov decision process. Agents observe an atomistic structure, apply chemically meaningful edits, and receive feedback from a configurable evaluator. SciVerseGym supports local and global actions, including elemental substitution, lattice perturbation, atomic displacement, vacancy creation, and atom insertion, along with configurable chemical spaces, structure pools, atomistic and graph-based observations, custom rewards, optional relaxation, and stability or phonon-related diagnostics. Each step applies an edit, evaluates the candidate using a machine-learned interatomic potential or any ASE-compatible calculator, and returns the standard (obs, reward, terminated, truncated, info) tuple. By decoupling agent logic from materials infrastructure, SciVerseGym provides an open, reproducible, and extensible testbed for reinforcement learning, Bayesian optimization, evolutionary search, and language-agent workflows in closed-loop crystal discovery. Code is available at: https://github.com/Bin-Cao/SciVerseGym.", "date": "2026-06-21", "categories": []}
{"id": "2606.22566", "title": "Deep material network for homogenization of piezoelectric composites", "abstract": "Piezoelectric composites are widely used in sensors, actuators, transducers, and energy-harvesting devices because their effective electromechanical performance can be tailored by combining constituent phases and microstructural architecture. However, conventional computational homogenization based on direct numerical simulation (DNS) is computationally expensive, particularly for multiscale simulations and material design tasks that require repeated homogenization analyses. To address this limitation, this work proposes a piezoelectric deep material network (PDMN) to efficiently homogenize two-phase piezoelectric composites. The proposed framework embeds the governing electromechanical homogenization relations directly into the network architecture, yielding a physics-informed, semi-analytical surrogate that explicitly captures the two-way coupling between the mechanical and electrical fields across constituent phases. The network is trained offline on linear electroelastic datasets and, through a fully coupled Newton--Raphson solution with a consistent electromechanical tangent, subsequently used for efficient online prediction under broader constitutive settings, including nonlinear electroelasticity and history-dependent responses. The framework is validated on two-phase composites of polyvinylidene fluoride (PVDF) and lithium niobate (LiNbO$_3$) with reversed phase arrangements under nonlinear electroelastic loading, and on a viscoelastic--piezoelectric composite exhibiting coupled stress relaxation. Numerical examples show that the proposed PDMN achieves high predictive accuracy while reducing the computational cost by more than three orders of magnitude compared with DNS. The proposed framework, therefore, provides an efficient and reliable surrogate for the multiscale analysis and design of piezoelectric composites.", "date": "2026-06-21", "categories": []}
{"id": "2606.22752", "title": "One-Step Flow Matching for Generative Modeling of Path-Dependent Physical Fields", "abstract": "Physical simulations for intricate geometries with path-dependent constitutive models face difficulties due to the enormous computational cost they require. Recently, the emergence of generative AI models, which succeed in image and video synthesis tasks, has provided a promise to further improve simulations. Although U-Net-based denoising diffusion probabilistic models (DDPMs) have been adopted for elastic stress field generation, they typically require hundreds of sampling steps, and applications of generative models to path-dependent, e.g. plastic, stress fields remain very limited. In this work, we propose a novel flow matching (FM) model based on a transformer backbone for high-resolution path-dependent stress field generation with stochastic loading-unloading paths and geometry. The proposed model operates within the latent space of a variational autoencoder (VAE) and formulates the simulation of plastic fields as a video synthesis task, directly generating the stress fields across all time steps. Meanwhile, we design a non-Gaussian source distribution for flow matching, such that crossings among conditional transport paths are reduced during training. This enables our model to generate satisfactory samples in one step without relying on distillation. In addition, we introduce token-level loading embeddings and two auxiliary networks to further enhance the model performance in path-dependent simulation. The results demonstrate that, even with a limited training dataset, our model can accurately generate high-resolution path-dependent fields. It is much more computationally efficient than finite element analysis, providing a speedup of 6 to 7 times over FEM on CPUs and approximately two orders of magnitude speedup on consumer-grade GPUs.", "date": "2026-06-22", "categories": []}
{"id": "2606.22866", "title": "Discovering Crystal Structure Prediction Algorithms with an AI Co-Scientist", "abstract": "We introduce Human-AI Co-discovery system (HACO) for scientific algorithm discovery through cross-domain search and sparse human steering. Starting from the goal of generating crystal structures from chemical compositions, HACO searched across generative modeling methodologies from multiple fields and identified MaskGIT, a masked generative model from vision, as a promising framework for crystal structure prediction (CSP). HACO instantiated this masked formulation as a discrete token model of crystal structure; guided by sparse high-level human objectives, it then added crystallographic symmetry tokens, space group stratified sampling for polymorph coverage, and sub-bin coordinate refinement, yielding the Masked Generative Crystal Transformer (MaskGXT). On the MP-20 polymorph split, MaskGXT reaches 79.06% match-everyone-to-reference (METRe) accuracy, compared with 70.87% for the strongest evaluated baseline. MaskGXT also attains the best match rate on standard MP-20 and MPTS-52 CSP benchmarks. These results provide evidence that, in domains offering cheap, fast, and well-aligned validation, transfer-guided interactive AI co-scientists can contribute to scientific algorithm discovery by identifying transferable modeling principles and combining them with targeted human domain guidance.", "date": "2026-06-22", "categories": []}
{"id": "2606.23166", "title": "Substitution-Based Analysis of Structural Novelty for Generative Models of Materials", "abstract": "There has been rapid progress in generative artificial intelligence (AI) models for inorganic crystal design, which can efficiently generate large numbers of candidate compounds after being trained on databases of known crystals. However, it remains unclear whether they genuinely expand the accessible materials search space beyond conventional strategies such as elemental substitution within known structure types. We address this question by developing a workflow to assess whether AI-generated crystals are duplicates of training structures, reproducible by elemental substitution, or unmatched by either criterion. Applying this workflow to representative generative models reveals that 81-92% of chemically valid and metastable generated crystals are either training duplicates or substitution-derived structures. This tendency is particularly strong in high-symmetry crystal systems, even though many possible structural prototypes remain unexplored. Further analysis of the underlying structural fingerprints shows that low-symmetry structures beyond duplication or substitution can be interpreted as interpolation in training-data-rich regions, while high-symmetry duplicates appear to result from memorisation in training-sparse regions. Our findings highlight a limitation in the current generation of models that exhibit a bias towards known structural prototypes in the high symmetry regions, but enable wider exploration of the low-symmetry structural space.", "date": "2026-06-22", "categories": []}
{"id": "2606.28578", "title": "Surrogate-Gated Generation and Foundation-Model Embeddings for Bayesian Materials Design", "abstract": "Closed-loop materials discovery iterates between proposing candidate structures and evaluating their properties, and property evaluation dominates the cost. In the generative variant, a learned prior proposes candidate crystals and a property oracle scores them; we ask whether a cheap probabilistic surrogate can triage the generator's output, and what such a surrogate must do well. Across three architecturally distinct pretrained diffusion priors (MatterGen, CrystalFlow, ADiT) and two targets (room-temperature heat capacity and bulk modulus), we insert a Gaussian process acquisition gate between structure generation and the oracle in an RL-steered generative workflow. The gate matches or exceeds ungated fine-tuning of the generative model while capping oracle calls at a fixed per-cycle budget. Budget-matched ablations isolate the mechanism. At an identical four-call budget, ranking-based selection outperforms arbitrary selection, confirming that the gain comes from the surrogate's choice; the gate comes within $\\sim$9\\% of exhaustive oracle spending at roughly one-fifth of the calls. A density-functional-theory check of the bulk-modulus discoveries confirms the learned oracle to within 2.5\\% on average and the surrogate's ranking of the generated structures at Spearman $\\rho = 0.94$. A cross-factorial benchmark of surrogate performance spanning mechanical, electronic, and vibrational properties identifies pretrained ORB embeddings with a Gaussian process as the most reliable combination, which we adopt as the building blocks of the proposed workflow. The complete pipeline is released as open-source software.", "date": "2026-06-26", "categories": []}
{"id": "2606.28996", "title": "On Surrogate Modeling of Static Response of AM Short-Fiber Thermoplastics Using Graph Neural Networks", "abstract": "Short-fiber thermoplastic (SFT) composites are increasingly employed in lightweight aerospace and automotive structures owing to their favorable strength-to-weight ratio, high production rates, and recyclability. Unlike continuous-fiber systems, the mechanical response of SFTs is governed by mesoscale interactions among fiber orientation, spatial clustering, and manufacturing-induced porosity. These features exhibit significant spatial variability in manufactured components and influence stiffness, damage initiation, and nonlinear deformation. Although mesoscale finite element (FE) models can resolve such heterogeneity, their application to realistic three-dimensional microstructures remains computationally intractable.\n  A data-driven surrogate framework is proposed to predict the mechanical behavior of additively manufactured, compression-molded (AM-CM) SFTs. Microstructures reconstructed from micro-computed tomography data were discretized into Voronoi-based cells representing distinct fiber-interaction neighborhoods. Each cell was homogenized via nonlinear FE simulations incorporating matrix damage, and the resulting stress-strain responses trained a hybrid Graph Neural Network-Long Short-Term Memory (GNN-LSTM) architecture encoding microstructural topology and history-dependent mechanical evolution.\n  The surrogate accurately predicts stiffness and stress-strain behavior of unseen microstructures, achieving $R^2\\approx 0.98$ relative to high-fidelity FE simulations with over two orders-of-magnitude reduction in computational cost. Coupling the framework with experimentally calibrated damage laws demonstrates that fiber orientation, clustering, and porosity collectively govern local effective stiffness. The approach provides a physics-informed, data-efficient pathway to identify mechanically weak microstructural cells and accelerate digital-twin development for SFT components.", "date": "2026-06-27", "categories": []}
{"id": "2606.29459", "title": "Interpretable Inverse Design of Metal-Organic Frameworks with Large Language Model Agents", "abstract": "Inverse design of metal-organic frameworks (MOFs) requires searching a combinatorially vast space where property labels are expensive and most machine-learning models reveal little about why a structure succeeds. We introduce LLM4MOF, a closed-loop framework in which language-model agents reason about chemistry, build candidate MOFs, and test them in simulation, refining hypotheses over ten autonomous iterations. One agent proposes interpretable design hypotheses over metal nodes, linkers, pore geometry, and functional chemistry, and a second translates them into constraints that select candidate MOFs, each made of a metal node, organic linker, and matching topology. Each hypothesis is tested through four diagnostic beams that apply different subsets of its constraints, so comparing them shows whether geometry, chemistry, or metal choice drives performance. Even when blind to the global property landscape of databases, LLM4MOF concentrates its search on top-performing structures across six adsorption, separation, and electronic-structure tasks within 400 property evaluations. The same loop also generates new MOFs de novo and validates them in live simulation, where it adapts the geometry to each requested condition, outperforming random search and a genetic algorithm at roughly $1 per campaign. LLM4MOF shows that language-model agents can run interpretable, simulation-grounded inverse design without training a model per objective.", "date": "2026-06-28", "categories": []}
{"id": "2606.29584", "title": "Geometric Algebra Meets Cartesian Tensors: Higher-Order Equivariance for Interatomic Potentials", "abstract": "$\\mathrm{Cl}(3,0)$ interatomic potentials, despite their algebraic elegance, predict force magnitudes accurately but force directions poorly. Across ten rMD17 molecules, every $L \\leq 1$ baseline in our twelve-model study attains aggregate force-cosine similarity below $0.25$. The cause is structural. The geometric product of two vectors in $\\mathbb{R}^3$ realises only the $L=0$ and $L=1$ components of its irreducible representation content, leaving the symmetric-traceless rank-2 component absent from the per-edge bilinear that drives each message-passing layer. We address this with CliffordSTF, which couples the Clifford multivector to closed-form symmetric-traceless tensor tracks at ranks two and three through bilinear cross-track contractions, using a single learned bilinear and no Clebsch--Gordan tables, Wigner-$D$ matrices, or e3nn calls. On rMD17, CliffordSTF raises aggregate force-cosine similarity from $0.055$ (base Clifford) to $0.551$, an order-of-magnitude relative directional gain, alongside improved magnitude accuracy (force MAE $15.8\\%$ lower; energy MAE $10.9\\%$ lower). It outperforms all CG-free or body-ordered baselines in our study (all $\\leq 0.17$). On catalysis benchmarks, CliffordSTF achieves the best out-of-distribution S2EF energy MAE on OC22 in our experiments, and the best in-distribution energy MAE among $L \\geq 2$ methods on OC22 IS2RE. An eleven-variant ablation shows the two tracks are complementary: neither alone matches the combined model.", "date": "2026-06-28", "categories": []}
{"id": "2606.29717", "title": "Optimizing Expert-Designed Crystal Graph Networks for Band-Gap Prediction with an Autonomous LLM Research Loop", "abstract": "Predicting a material's properties from its structure is a central, fast-advancing problem in computational materials science. A decade of work has produced standard public benchmarks and many published machine-learning models for the task (Dunn et al., 2020). The task's fixed metric and these baselines make it a natural setting for autonomous agent research (Karpathy, 2026). On the MatBench band-gap benchmark ($>$100k crystals), a general-purpose coding agent autonomously built the most accurate model trained without external pretraining, ahead of all seventeen expert-designed models reported for the task. A closer analysis shows it reached this by implementing known methods: either already standard in crystal neural-network models, or borrowed from other areas of machine learning. The contributing implementations include element-pair features on each message-passing edge and a crystal space-group embedding. The work not only demonstrates that LLM-agent autonomous research can optimize an expert-designed machine learning model for material property prediction, but also investigates the limitations of such autonomous research.", "date": "2026-06-29", "categories": []}
{"id": "2606.31068", "title": "Hybrid Unet-Transformer Model for Generating Stress and Strain Fields from Composite Geometrics", "abstract": "Accurate prediction of stress and strain fields in hierarchical composite microstructures is critical for physics-informed material design, yet conventional finite element method (FEM) simulations are computationally prohibitive at scale, requiring minutes to days per evaluation. In this work, we propose a hybrid UNet-Transformer architecture that predicts complex mechanical field distributions directly from composite microstructure geometry images, serving as an efficient surrogate for FEM across ten distinct stress and strain field types spanning diverse two-phase composite configurations including square, hexagonal, and triangular tessellations, multiple boundary conditions, and high-resolution geometries. Results demonstrate that the proposed architecture achieves strong predictive performance across the majority of subdatasets, with peak accuracy on periodic tessellation geometries reaching R2=0.9991, SSIM=0.9936, and MAE=0.0050 on the boundary condition subdataset and the triangular tessellation subdataset respectively. Across six of the eight evaluated subdatasets, MAE remains below 0.05 on the normalized [0,1] pixel scale. Encoder attention analysis via Grad-CAM and Grad-CAM++ confirms that the model develops physically meaningful internal representations, localizing attention at mechanically critical regions including phase boundaries, ligament junctions, and indenter contact zones without explicit structural supervision. Performance degrades on irregular square-grid geometries with sparse soft-phase inclusions, with the S11 normal stress subdataset yielding R2=0.7735 and SSIM=0.7126, consistent with the known limitation of smooth-loss image translation models in reproducing sharp stress discontinuities.", "date": "2026-06-30", "categories": []}