shanto268's picture
Card: paper title 'QCap: a physics-corrected graph neural network for the capacitance matrix of any planar superconducting device layout'
7319bc5 verified
|
Raw History Blame Contribute Delete
5.2 kB
metadata
license: cc-by-4.0
pretty_name: SQuADDS_PlanarCapacitance-Synthetic
private: true
task_categories:
  - tabular-regression
  - graph-ml
tags:
  - superconducting-qubits
  - capacitance
  - physics-prior
  - synthetic
size_categories:
  - 100K<n<1M
configs:
  - config_name: surrogate
    data_files:
      - split: gen_v1
        path: surrogate/gen_v1/*.parquet
      - split: gen_t1
        path: surrogate/gen_t1/*.parquet
  - config_name: generic
    data_files:
      - split: gc5
        path: generic/gc5/*.parquet
      - split: gc7_nest
        path: generic/gc7_nest/*.parquet
      - split: gc8_wide
        path: generic/gc8_wide/*.parquet
  - config_name: generic_release_convention_p6
    data_files:
      - split: all
        path: generic/release_convention_p6/*.parquet
SQuADDS logo

SQuADDS_PlanarCapacitance-Synthetic: generated layouts solved by the physics prior alone

Generated planar layouts whose capacitance comes only from our thin-film method of moments, with no Q3D simulation. We used them to train and study the learned prior of the fast model in our paper QCap: a physics-corrected graph neural network for the capacitance matrix of any planar superconducting device layout.

What's inside

set config layouts targets
surrogate set surrogate 385,593 (346,993 train, 38,600 validation) mutual capacitances from two coarse solves
generic set generic and generic_release_convention_p6 286,134 the coarse prior, and the extrapolated prior on 48,075 of them

The generic_release_convention_p6 config holds those 48,075 layouts recomputed with the prior settings of the surrogate set and of our released models. Use it, rather than the C_R column of generic, when you train on both sets. Join it to generic on lid for the geometry.

Conventions

  • Capacitance is in fF. C_* columns are full Maxwell matrices flattened row by row, and mutual_* columns hold the positive mutual capacitances of the pairs i < j.
  • Ground is conductor 0. Nets are never merged, so N counts the ground.
  • The prior assumes a 350 µm silicon substrate (eps_r 11.45). Lengths are in µm.

Reading

import pandas as pd

df = pd.read_parquet("hf://datasets/SQuADDS/SQuADDS_PlanarCapacitance-Synthetic/surrogate/gen_t1/part-00000.parquet")

Generic rows store their geometry (geometry_wkb). Surrogate rows are regenerated from (seed, index) with the generator code in extras/code/generators.tar.zst, which also writes GDS files.

Test geometry

No generated layout matches a test layout of the Q3D dataset exactly (0 matches of the canonical geometry hash). A looser shape-signature check flags 336 generic layouts with one or two signal conductors, and generic/test_signature_matches.parquet lists them for anyone who wants to drop them.

Contributors

Sadman Ahmed Shanto wrote the generators and computed the priors.

Citation

If you use these data or models, please cite our paper and SQuADDS.

@misc{Shanto2026planarcapacitance,
  title   = {{QCap}: a physics-corrected graph neural network for the capacitance matrix of any planar superconducting device layout},
  author  = {Shanto, Sadman Ahmed and Das, Saikat and Levenson-Falk, Eli},
  year    = {2026},
  url     = {https://lfl-lab.github.io/SQuADDS/ml-paper/paper/}
}

@article{Shanto2024squaddsvalidated,
  doi = {10.22331/q-2024-09-09-1465},
  url = {https://doi.org/10.22331/q-2024-09-09-1465},
  title = {{SQ}u{ADDS}: {A} validated design database and simulation workflow for superconducting qubit design},
  author = {Shanto, Sadman and Kuo, Andre and Miyamoto, Clark and Zhang, Haimeng and Maurya, Vivek and Vlachos, Evangelos and Hecht, Malida and Shum, Chung Wa and Levenson-Falk, Eli},
  journal = {{Quantum}},
  issn = {2521-327X},
  publisher = {{Verein zur F{\"{o}}rderung des Open Access Publizierens in den Quantenwissenschaften}},
  volume = {8},
  pages = {1465},
  month = sep,
  year = {2024}
}

Licence

CC-BY-4.0.