Upload ALD-E-ImageMiner ImageFolder dataset package

README.md

@@ -118,12 +118,12 @@ The source `dev` split is exposed as the Hugging Face `validation` split.
 | `subset` | Lowercase kebab-case subset chosen from the first source classification label. |
 | `split` | Hugging Face split: train, validation, or test. |
 | `classification` | All source panel classifications as a list of panel_id/label structs. |
+| `summarization, data_extraction, vqa` | Source task annotations preserved as JSON strings containing stable panel lists. Empty lists indicate that the source record did not provide that task annotation. |
+| `bbox` | Source bounding boxes converted to a stable panel list. |
 | `source` | Relative path to the source image in this repository. |
 | `provenance` | Relative annotation/content/PDF paths and source navigation fields. |
 | `width, height, image_format, image_sha256` | Image properties computed from the copied source image. |
 | `metadata_license, image_license, image_reuse_status` | Conservative reuse and licensing metadata. |
-| `bbox` | Source bounding boxes converted to a stable panel list. |
-| `summarization, data_extraction, vqa` | Source task annotations preserved as JSON strings containing stable panel lists. Empty lists indicate that the source record did not provide that task annotation. |
 | `schema_version` | Metadata schema version, currently 1.0.0. |
 
 ## Loading

subsets/3d-scatter-plot/test/metadata.jsonl

@@ -1,3 +1,3 @@
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce.jpg","caption":"(a) Data set for Step A generated from the microscopic model.","id":"test/atomic-layer-etching/simulation-usecase/11/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce","sample_id":"atomic-layer-etching/simulation-usecase/11/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce","subset":"3d-scatter-plot","split":"test","classification":[{"panel_id":"a","label":"3d scatter plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":505,"height":442,"image_format":"jpeg","image_sha256":"48c2269aed17d81ab39416e1f6060ff4cf04d71a93225707e3a9e4a550352fd1","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":8,"width":499,"height":433}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"3D scatter plot of kMC simulation results for Step A (HF fluorination) showing half-cycle time as a function of temperature (533-598 K) and pressure (10-100 Pa). The data shows strong temperature dependence but near-complete pressure independence, with half-cycle times ranging from approximately 1 s at high temperatures to 12 s at low temperatures. The horizontal bands at each temperature confirm that HF adsorption kinetics are not significantly affected by pressure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Pressure (Pa) | Time (s) |\\n|----------------|---------------|----------|\\n| 520           | 20            | 19       |\\n| 540           | 20            | 18       |\\n| 560           | 20            | 17       |\\n| 580           | 20            | 16       |\\n| 600           | 20            | 15       |\\n| 520           | 40            | 14       |\\n| 540           | 40            | 13       |\\n| 560           | 40            | 12       |\\n| 580           | 40            | 11       |\\n| 600           | 40            | 10       |\\n| 520           | 60            | 9        |\\n| 540           | 60            | 8        |\\n| 560           | 60            | 7        |\\n| 580           | 60            | 6        |\\n| 600           | 60            | 5        |\\n| 520           | 80            | 4        |\\n| 540           | 80            | 3        |\\n| 560           | 80            | 2        |\\n| 580           | 80            | 1        |\\n| 600           | 80            | 0        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Step A shows virtually no pressure dependence, as evidenced by the horizontal bands of data points at each temperature. In contrast, Step B exhibits clear pressure dependence with half-cycle times decreasing at higher pressures. This difference arises from the reaction mechanisms: HF dissociatively adsorbs readily onto the Al2O3 surface regardless of pressure, while TMA adsorption in Step B is more kinetically limited and benefits from higher partial pressures to drive secondary adsorption events.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Half-cycle times range from approximately 1 s at high temperatures (~598 K) to about 12 s at low temperatures (~533 K).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Temperature governs the rate constants through the Arrhenius equation, so higher temperatures accelerate surface reactions like hydrogen diffusion and water desorption that determine the fluorination half-cycle completion time.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Process engineers have flexibility in selecting HF pressure without significantly affecting Step A half-cycle time, allowing pressure optimization based on other factors such as precursor utilization efficiency or reactor design constraints.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig_6.jpg","caption":"Fig. 6 - Mass change over time during 3 cycles with purge time of  $30s$ . (b) Data set for Step B generated from the microscopic model. Fig. 7 - Data points collected for Step A and Step B at various operating conditions.","id":"test/atomic-layer-etching/simulation-usecase/11/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/11/fig_6","subset":"3d-scatter-plot","split":"test","classification":[{"panel_id":"a","label":"3d scatter plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_6.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":505,"height":442,"image_format":"jpeg","image_sha256":"603a875cd5ef217aaada71e57748f1221ca1a573bfc898020fb0c95fdec8441f","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":5,"width":502,"height":435}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"3D scatter plot of kMC simulation results for Step B (TMA etching) showing half-cycle time as a function of temperature (533-598 K) and pressure (10-100 Pa). The data reveals strong dependencies on both parameters: half-cycle times range from approximately 2 s at high temperature/pressure to over 20 s at low temperature. This dataset was used to train the FNN model for Step B.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Pressure (Pa) | Time (s) |\\n|----------------|---------------|----------|\\n| 520           | 20            | 38       |\\n| 540           | 20            | 36       |\\n| 560           | 20            | 34       |\\n| 580           | 20            | 32       |\\n| 600           | 20            | 30       |\\n| 520           | 40            | 28       |\\n| 540           | 40            | 26       |\\n| 560           | 40            | 24       |\\n| 580           | 40            | 22       |\\n| 600           | 40            | 20       |\\n| 520           | 60            | 18       |\\n| 540           | 60            | 16       |\\n| 560           | 60            | 14       |\\n| 580           | 60            | 12       |\\n| 600           | 60            | 10       |\\n| 520           | 80            | 8        |\\n| 540           | 80            | 6        |\\n| 560           | 80            | 4        |\\n| 580           | 80            | 2        |\\n| 600           | 80            | 0        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both parameters significantly influence Step B half-cycle time. Temperature has the dominant effect: at low temperatures (~533 K), times exceed 20 s, while at high temperatures (~598 K), times drop to around 2-3 s, reflecting Arrhenius kinetics. Pressure also matters for Step B because TMA adsorption kinetics depend on partial pressure. Higher pressures accelerate the secondary TMA adsorption step, reducing half-cycle time. This dual dependence contrasts with Step A, which is largely pressure-independent.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Data points were generated by running kMC simulations across temperature (533-598 K in 5 K intervals) and pressure (10-100 Pa in 5 Pa intervals), averaging 10 simulations per condition to reduce stochastic variation.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Each kMC simulation takes approximately 20 minutes on average, and multiple simulations are averaged per operating condition to account for the inherent randomness of the Monte Carlo method.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Temperature in Kelvin (x-axis), Pressure in Pascals (y-axis), and half-cycle time in seconds (z-axis).\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig_8.jpg","caption":"Fig. 8 - Comparison of the kMC simulation data and of the predicted data calculated by the PIN for Steps A and B. Scatter (a) and line (b) plots of Step A represent a mean squared error of  $0.0856\\%$ , which were calculated from  $20\\%$  of the data points of the full kMC data set for Step A. Scatter (c) and line (d) plots of Step B represent a mean squared error of  $0.175\\%$ , which were calculated from  $20\\%$  of the data points of the full kMC data set for Step B.","id":"test/atomic-layer-etching/simulation-usecase/11/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/11/fig_8","subset":"3d-scatter-plot","split":"test","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"3d scatter plot"},{"panel_id":"d","label":"scatter plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_8.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":1017,"height":809,"image_format":"jpeg","image_sha256":"81a19a6274ca23cb24bdcdaae12cd8251b223c75aa6718f116ecc8a243dc968b","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":12,"y":0,"width":458,"height":384},{"panel_id":"b","x":548,"y":0,"width":468,"height":387},{"panel_id":"c","x":12,"y":426,"width":460,"height":380},{"panel_id":"d","x":536,"y":433,"width":480,"height":373}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Scatter plot comparing kMC simulation data (red) with FNN-predicted data (blue) for Step A (HF fluorination) across different operating conditions. The close overlap of red and blue points demonstrates excellent model agreement.\"},{\"panel_id\":\"b\",\"text\":\"Parity plot for Step A showing FNN-predicted half-cycle time versus kMC simulation time. Points cluster tightly along the diagonal y=x line, indicating a mean squared error of only 0.0856% on the 20% test dataset.\"},{\"panel_id\":\"c\",\"text\":\"Scatter plot comparing kMC simulation data (red) with FNN-predicted data (blue) for Step B (TMA etching). The agreement is similarly strong, though with slightly more scatter at higher normalized times.\"},{\"panel_id\":\"d\",\"text\":\"Parity plot for Step B showing FNN-predicted versus simulated half-cycle times. The data closely follows the diagonal reference line with a mean squared error of 0.175%, slightly higher than Step A but still excellent.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Iteration Index | Top Band (Normalized Time, s) | Bottom Band (Normalized Time, s) |\\n|---|---|---|\\n| 0 | 2.40 | 0.60 |\\n| 25 | 2.40 | -0.80 |\\n| 50 | 2.42 | -0.90 |\\n| 75 | 2.45 | -0.90 |\\n| 100 | 2.50 | -0.90 |\\n| 150 | 2.60 | -0.90 |\\n| 200 | 2.70 | -0.90 |\\n| 250 | 2.80 | -0.90 |\"},{\"panel_id\":\"b\",\"text\":\"| Simulation Time (s) | Predicted Time (s) |\\n|---|---|\\n| -0.9 | -0.9 |\\n| 0.0 | 0.0 |\\n| 1.0 | 1.0 |\\n| 2.0 | 2.0 |\\n| 2.7 | 2.7 |\"},{\"panel_id\":\"c\",\"text\":\"| Iteration Index | Top Band (Normalized Time, s) | Bottom Band (Normalized Time, s) |\\n|---|---|---|\\n| 0 | 3.10 | 0.80 |\\n| 25 | 3.00 | -0.30 |\\n| 50 | 2.90 | -0.70 |\\n| 75 | 2.80 | -0.85 |\\n| 100 | 2.65 | -0.95 |\\n| 150 | 2.40 | -1.00 |\\n| 200 | 2.10 | -1.00 |\\n| 250 | 1.80 | -1.00 |\"},{\"panel_id\":\"d\",\"text\":\"| Simulation Time (s) | Predicted Time (s) |\\n|---|---|\\n| -1.0 | -1.0 |\\n| 0.0 | 0.0 |\\n| 1.0 | 1.0 |\\n| 2.0 | 2.0 |\\n| 3.0 | 3.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The FNN model dramatically reduces computation time while maintaining high accuracy. While kMC simulations require approximately 20 minutes per data point on average, the trained FNN predicts half-cycle times in less than one second. This speed enables real-time operational parameter calculations, which is essential for semiconductor manufacturing where batch half-cycles typically complete in under 5 seconds. The FNN also allows rapid screening of operating conditions without costly experiments.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"20% of the kMC data points were used for testing and validation, while the remaining 80% were used for training the FNN models.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The FNN model achieves slightly better accuracy for Step A than Step B. Step A (HF fluorination) shows a mean squared error of 0.0856%, while Step B (TMA etching) has a MSE of 0.175%. This difference likely reflects the greater complexity of Step B's reaction mechanism, which involves more reaction pathways including ligand exchange and DMAF desorption. Nevertheless, both models demonstrate excellent predictive capability with sub-1% errors.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A point on the diagonal indicates perfect agreement between the FNN-predicted half-cycle time and the kMC-simulated time.\"}]}]"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce.jpg","caption":"(a) Data set for Step A generated from the microscopic model.","id":"test/atomic-layer-etching/simulation-usecase/11/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce","sample_id":"atomic-layer-etching/simulation-usecase/11/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce","subset":"3d-scatter-plot","split":"test","classification":[{"panel_id":"a","label":"3d scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"3D scatter plot of kMC simulation results for Step A (HF fluorination) showing half-cycle time as a function of temperature (533-598 K) and pressure (10-100 Pa). The data shows strong temperature dependence but near-complete pressure independence, with half-cycle times ranging from approximately 1 s at high temperatures to 12 s at low temperatures. The horizontal bands at each temperature confirm that HF adsorption kinetics are not significantly affected by pressure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Pressure (Pa) | Time (s) |\\n|----------------|---------------|----------|\\n| 520           | 20            | 19       |\\n| 540           | 20            | 18       |\\n| 560           | 20            | 17       |\\n| 580           | 20            | 16       |\\n| 600           | 20            | 15       |\\n| 520           | 40            | 14       |\\n| 540           | 40            | 13       |\\n| 560           | 40            | 12       |\\n| 580           | 40            | 11       |\\n| 600           | 40            | 10       |\\n| 520           | 60            | 9        |\\n| 540           | 60            | 8        |\\n| 560           | 60            | 7        |\\n| 580           | 60            | 6        |\\n| 600           | 60            | 5        |\\n| 520           | 80            | 4        |\\n| 540           | 80            | 3        |\\n| 560           | 80            | 2        |\\n| 580           | 80            | 1        |\\n| 600           | 80            | 0        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Step A shows virtually no pressure dependence, as evidenced by the horizontal bands of data points at each temperature. In contrast, Step B exhibits clear pressure dependence with half-cycle times decreasing at higher pressures. This difference arises from the reaction mechanisms: HF dissociatively adsorbs readily onto the Al2O3 surface regardless of pressure, while TMA adsorption in Step B is more kinetically limited and benefits from higher partial pressures to drive secondary adsorption events.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Half-cycle times range from approximately 1 s at high temperatures (~598 K) to about 12 s at low temperatures (~533 K).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Temperature governs the rate constants through the Arrhenius equation, so higher temperatures accelerate surface reactions like hydrogen diffusion and water desorption that determine the fluorination half-cycle completion time.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Process engineers have flexibility in selecting HF pressure without significantly affecting Step A half-cycle time, allowing pressure optimization based on other factors such as precursor utilization efficiency or reactor design constraints.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":8,"width":499,"height":433}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/f15f209f2d385c0b325b181192c48ae716e73fcf75a8c14d4bd8538703ffdfce.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":505,"height":442,"image_format":"jpeg","image_sha256":"48c2269aed17d81ab39416e1f6060ff4cf04d71a93225707e3a9e4a550352fd1","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig_6.jpg","caption":"Fig. 6 - Mass change over time during 3 cycles with purge time of  $30s$ . (b) Data set for Step B generated from the microscopic model. Fig. 7 - Data points collected for Step A and Step B at various operating conditions.","id":"test/atomic-layer-etching/simulation-usecase/11/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/11/fig_6","subset":"3d-scatter-plot","split":"test","classification":[{"panel_id":"a","label":"3d scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"3D scatter plot of kMC simulation results for Step B (TMA etching) showing half-cycle time as a function of temperature (533-598 K) and pressure (10-100 Pa). The data reveals strong dependencies on both parameters: half-cycle times range from approximately 2 s at high temperature/pressure to over 20 s at low temperature. This dataset was used to train the FNN model for Step B.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Pressure (Pa) | Time (s) |\\n|----------------|---------------|----------|\\n| 520           | 20            | 38       |\\n| 540           | 20            | 36       |\\n| 560           | 20            | 34       |\\n| 580           | 20            | 32       |\\n| 600           | 20            | 30       |\\n| 520           | 40            | 28       |\\n| 540           | 40            | 26       |\\n| 560           | 40            | 24       |\\n| 580           | 40            | 22       |\\n| 600           | 40            | 20       |\\n| 520           | 60            | 18       |\\n| 540           | 60            | 16       |\\n| 560           | 60            | 14       |\\n| 580           | 60            | 12       |\\n| 600           | 60            | 10       |\\n| 520           | 80            | 8        |\\n| 540           | 80            | 6        |\\n| 560           | 80            | 4        |\\n| 580           | 80            | 2        |\\n| 600           | 80            | 0        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both parameters significantly influence Step B half-cycle time. Temperature has the dominant effect: at low temperatures (~533 K), times exceed 20 s, while at high temperatures (~598 K), times drop to around 2-3 s, reflecting Arrhenius kinetics. Pressure also matters for Step B because TMA adsorption kinetics depend on partial pressure. Higher pressures accelerate the secondary TMA adsorption step, reducing half-cycle time. This dual dependence contrasts with Step A, which is largely pressure-independent.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Data points were generated by running kMC simulations across temperature (533-598 K in 5 K intervals) and pressure (10-100 Pa in 5 Pa intervals), averaging 10 simulations per condition to reduce stochastic variation.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Each kMC simulation takes approximately 20 minutes on average, and multiple simulations are averaged per operating condition to account for the inherent randomness of the Monte Carlo method.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Temperature in Kelvin (x-axis), Pressure in Pascals (y-axis), and half-cycle time in seconds (z-axis).\"}]}]","bbox":[{"panel_id":"a","x":3,"y":5,"width":502,"height":435}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_6.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":505,"height":442,"image_format":"jpeg","image_sha256":"603a875cd5ef217aaada71e57748f1221ca1a573bfc898020fb0c95fdec8441f","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig_8.jpg","caption":"Fig. 8 - Comparison of the kMC simulation data and of the predicted data calculated by the PIN for Steps A and B. Scatter (a) and line (b) plots of Step A represent a mean squared error of  $0.0856\\%$ , which were calculated from  $20\\%$  of the data points of the full kMC data set for Step A. Scatter (c) and line (d) plots of Step B represent a mean squared error of  $0.175\\%$ , which were calculated from  $20\\%$  of the data points of the full kMC data set for Step B.","id":"test/atomic-layer-etching/simulation-usecase/11/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/11/fig_8","subset":"3d-scatter-plot","split":"test","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"3d scatter plot"},{"panel_id":"d","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Scatter plot comparing kMC simulation data (red) with FNN-predicted data (blue) for Step A (HF fluorination) across different operating conditions. The close overlap of red and blue points demonstrates excellent model agreement.\"},{\"panel_id\":\"b\",\"text\":\"Parity plot for Step A showing FNN-predicted half-cycle time versus kMC simulation time. Points cluster tightly along the diagonal y=x line, indicating a mean squared error of only 0.0856% on the 20% test dataset.\"},{\"panel_id\":\"c\",\"text\":\"Scatter plot comparing kMC simulation data (red) with FNN-predicted data (blue) for Step B (TMA etching). The agreement is similarly strong, though with slightly more scatter at higher normalized times.\"},{\"panel_id\":\"d\",\"text\":\"Parity plot for Step B showing FNN-predicted versus simulated half-cycle times. The data closely follows the diagonal reference line with a mean squared error of 0.175%, slightly higher than Step A but still excellent.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Iteration Index | Top Band (Normalized Time, s) | Bottom Band (Normalized Time, s) |\\n|---|---|---|\\n| 0 | 2.40 | 0.60 |\\n| 25 | 2.40 | -0.80 |\\n| 50 | 2.42 | -0.90 |\\n| 75 | 2.45 | -0.90 |\\n| 100 | 2.50 | -0.90 |\\n| 150 | 2.60 | -0.90 |\\n| 200 | 2.70 | -0.90 |\\n| 250 | 2.80 | -0.90 |\"},{\"panel_id\":\"b\",\"text\":\"| Simulation Time (s) | Predicted Time (s) |\\n|---|---|\\n| -0.9 | -0.9 |\\n| 0.0 | 0.0 |\\n| 1.0 | 1.0 |\\n| 2.0 | 2.0 |\\n| 2.7 | 2.7 |\"},{\"panel_id\":\"c\",\"text\":\"| Iteration Index | Top Band (Normalized Time, s) | Bottom Band (Normalized Time, s) |\\n|---|---|---|\\n| 0 | 3.10 | 0.80 |\\n| 25 | 3.00 | -0.30 |\\n| 50 | 2.90 | -0.70 |\\n| 75 | 2.80 | -0.85 |\\n| 100 | 2.65 | -0.95 |\\n| 150 | 2.40 | -1.00 |\\n| 200 | 2.10 | -1.00 |\\n| 250 | 1.80 | -1.00 |\"},{\"panel_id\":\"d\",\"text\":\"| Simulation Time (s) | Predicted Time (s) |\\n|---|---|\\n| -1.0 | -1.0 |\\n| 0.0 | 0.0 |\\n| 1.0 | 1.0 |\\n| 2.0 | 2.0 |\\n| 3.0 | 3.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The FNN model dramatically reduces computation time while maintaining high accuracy. While kMC simulations require approximately 20 minutes per data point on average, the trained FNN predicts half-cycle times in less than one second. This speed enables real-time operational parameter calculations, which is essential for semiconductor manufacturing where batch half-cycles typically complete in under 5 seconds. The FNN also allows rapid screening of operating conditions without costly experiments.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"20% of the kMC data points were used for testing and validation, while the remaining 80% were used for training the FNN models.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The FNN model achieves slightly better accuracy for Step A than Step B. Step A (HF fluorination) shows a mean squared error of 0.0856%, while Step B (TMA etching) has a MSE of 0.175%. This difference likely reflects the greater complexity of Step B's reaction mechanism, which involves more reaction pathways including ligand exchange and DMAF desorption. Nevertheless, both models demonstrate excellent predictive capability with sub-1% errors.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A point on the diagonal indicates perfect agreement between the FNN-predicted half-cycle time and the kMC-simulated time.\"}]}]","bbox":[{"panel_id":"a","x":12,"y":0,"width":458,"height":384},{"panel_id":"b","x":548,"y":0,"width":468,"height":387},{"panel_id":"c","x":12,"y":426,"width":460,"height":380},{"panel_id":"d","x":536,"y":433,"width":480,"height":373}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_8.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":1017,"height":809,"image_format":"jpeg","image_sha256":"81a19a6274ca23cb24bdcdaae12cd8251b223c75aa6718f116ecc8a243dc968b","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/3d-scatter-plot/train/metadata.jsonl

@@ -1,5 +1,5 @@
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_29_figure_10.jpg","caption":"Figure 10. AFM image of a  $350\\mathrm{\\AA}$  thick tungsten nitride film deposited at  $600~\\mathrm{K}$  after  $140\\mathrm{AB}$  cycles on a  $125\\mathrm{\\AA}$  thick  $\\mathrm{SiO_2}$  film on  $\\mathrm{Si}(100)$ . The reactant exposures of  $12,000\\mathrm{L}\\mathrm{NH_3}$  and  $4300\\mathrm{L}\\mathrm{WE_2}$  were sufficient for complete half-reactions during each AB cycle. The light-to-dark range is  $25\\mathrm{\\AA}$ .","id":"train/atomic-layer-deposition/experimental-usecase/29/figure_10","sample_id":"atomic-layer-deposition/experimental-usecase/29/figure_10","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/J. W. Klaus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:image_caption"},"width":683,"height":519,"image_format":"jpeg","image_sha256":"4d15477c997e8bf84ea41df2d602463ffed2a5179d1360a9f0f4457315188444","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":1,"y":6,"width":676,"height":513}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_63_fig_4.jpg","caption":"FIG. 4. (Color online) 3D surface morphologies of (a) as-deposited  $(250^{\\circ}\\mathrm{C})$  and (b) annealed  $\\sim 26 \\mathrm{nm}$  thick  $\\mathrm{Ga}_2\\mathrm{O}_3$  thin films.","id":"train/atomic-layer-deposition/experimental-usecase/63/fig_4","sample_id":"atomic-layer-deposition/experimental-usecase/63/fig_4","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"3d scatter plot"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/Donmez et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"63","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:image_caption"},"width":642,"height":786,"image_format":"jpeg","image_sha256":"41c2b92e624d68da87e702841311b2ef087c704c734547bc3e5b40f78b293411","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":0,"width":635,"height":388},{"panel_id":"b","x":6,"y":402,"width":632,"height":381}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"3D AFM surface topography of the as deposited Ga2O3 film (about 26 nm, deposited at 250 C), shown over a 1 µm by 1 µm scan area with height contrast in nm. The reported rms roughness is 0.16 nm.\"},{\"panel_id\":\"b\",\"text\":\"3D AFM surface topography of the annealed Ga2O3 film (about 26 nm), shown over a 1 µm by 1 µm scan area with height contrast in nm. The reported rms roughness increases to 0.37 nm after annealing, consistent with a rougher, more grain like surface.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| x-axis | y-axis | z-axis |\\n|--------|--------|--------|\\n| 0.0    | 0.0    | 0.0    |\\n| 250.0  | 250.0  | 0.0  |\\n| 500.0  | 500.0  | 0.0  |\\n| 750.0  | 750.0  | 0.0  |\\n| 1000.0 | 1000.0 | 0.0  |\"},{\"panel_id\":\"b\",\"text\":\"| x-axis | y-axis | z-axis |\\n|--------|--------|--------|\\n| 0.0    | 0.0    | 0.0    |\\n| 250.0  | 250.0  | 0.0  |\\n| 500.0  | 500.0  | 0.0  |\\n| 750.0  | 750.0  | 0.0  |\\n| 1000.0 | 1000.0 | 0.0  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The color scale represents surface height in nm across the scanned area.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The rms roughness increases from 0.16 nm (as deposited) to 0.37 nm (annealed).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The paper attributes the roughness increase to grain formation during crystallization upon annealing. As grains form and grow, the surface develops larger height variations compared with the amorphous as deposited film. This interpretation links the morphology change directly to the structural transition induced by annealing.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increased roughness can affect interface quality when Ga2O3 is integrated with metals or semiconductors, because local thickness and contact area vary across the surface. Rougher surfaces can also increase scattering and raise leakage risk in thin insulating layers, depending on the device stack. In exchange, annealing can deliver the crystalline phase and associated property improvements, so roughness becomes a tradeoff to manage.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_11.jpg","caption":"Figure 11. Atomic force microscopy (AFM) images of the surfaces etched using (a) ALE and (b) digital etching and (c) the as-grown surface before etching.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_11","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"3d scatter plot"},{"panel_id":"c","label":"3d scatter plot"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/Low-Damage and Self-Limiting (Al)GaN Etching Process.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"13","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":1186,"height":798,"image_format":"jpeg","image_sha256":"5c5b1963b56929ede4c66d4fe42024c5680165bbee269dd74653a26c2b219aa3","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":30,"width":563,"height":315},{"panel_id":"b","x":626,"y":11,"width":560,"height":312},{"panel_id":"c","x":296,"y":436,"width":542,"height":362}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_6.jpg","caption":"FIG. 6. (Color online) Surface morphology of the trench bottom for  $\\Gamma_{i} / \\Gamma_{n} = (a)0.01$  b  $10^{-3}$  and (c)  $10^{-4}$  in the passivation step after 25 pulses (87.5 s etch time). The color bar indicates the profile height, with dark representing deeper etching and light representing higher features.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_6","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"3d scatter plot"},{"panel_id":"c","label":"3d scatter plot"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":578,"height":1114,"image_format":"jpeg","image_sha256":"c73ead0e71dd35a25264c29156f3c541e48d52a4a2aa74129e15ddc0434c2e35","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":8,"width":574,"height":385},{"panel_id":"b","x":0,"y":388,"width":579,"height":341},{"panel_id":"c","x":1,"y":723,"width":576,"height":388}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image shows the surface morphology of the trench bottom for Γi/Γn  = 0.01 in the passivation step after 25 pulses (87.5 s etch time).\"},{\"panel_id\":\"b\",\"text\":\"The image shows the surface morphology of the trench bottom for Γi/Γn  = 10^-3 in the passivation step after 25 pulses (87.5 s etch time).\"},{\"panel_id\":\"c\",\"text\":\"The image shows the surface morphology of the trench bottom Γi/Γn  = 10^-5i in the passivation step after 25 pulses (87.5 s etch time).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Image | Γᵢ/Γₙ Ratio | Surface Width (nm) | Height Scale |\\n|-------|-------------|----------------------|--------------|\\n| (a)   | 0.01        | 25                   | Low → High (Black → Red → Yellow) |\\n| (b)   | 10⁻³        | 25                   | Low → High (Black → Red → Yellow) |\\n| (c)   | 10⁻⁴        | 25                   | Low → High (Black → Red → Yellow) |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface for Γᵢ/Γₙ =10^-4 has a scattering of small divots caused by individual ion strikes and larger divots where several ion strikes stochastically occurred in close proximity during the same passivation step. The surface for Γᵢ/Γₙ=10^-3 has a similar pattern of larger divots, but with significantly more uniform roughening as well. The surface for Γᵢ/Γ=0.01 has few large divots with the surface being dominated by random roughness; however, the divots\\nwhich persist are larger and deeper.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"There is a competition between a roughening process, produced by stochastic ion impacts during the passivation phase, and a smoothing process during the otherwise ideal ALE cycle\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The overall roughness increases as the ion to neutral flux ratio increases\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_8.jpg","caption":"FIG. 8. (Color online) Surface morphology of the trench bottom with  $100\\mathrm{ppm}\\mathrm{Cl}_2$  in the ion bombardment step after 25 pulses (32 s etch time). The color bar indicates the profile height, with dark representing deeper etching and light representing higher features.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_8","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":575,"height":447,"image_format":"jpeg","image_sha256":"10b781237ceb0fb2535052d1efb7ff2f8575938c6450449feb4987e9724b7b3d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":3,"width":571,"height":441}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Surface morphology of the trench bottom with 100 ppm Cl2 in the ion bombardment step after 32 s etch time. The color bar indicates the profile height,\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Parameter        | Value      |\\n|------------------|-----------|\\n| Cl₂ concentration | 100 ppm   |\\n| Surface width     | 27 nm     |\\n| Height scale      | Low → High (color gradient: black → red → yellow) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface has the characteristic divots, but the surface is also slightly concave due to a higher etch rate in the center of the feature. This higher etch rate is due to a higher flux of Cl atoms incident onto the center of the feature. The concavity becomes more pronounced with longer etch times (more pulses).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface morphology generated by having a Cl flux during the ion bombardment phase differs slightly from when having an ion flux during the passivation phase. The height of the etched surface at the bottom of the AR=2 trench after 25 ALE pulses (32 s) shown in the image for the 100 ppm Cl2 case\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"27 nm\"}]}]"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_29_figure_10.jpg","caption":"Figure 10. AFM image of a  $350\\mathrm{\\AA}$  thick tungsten nitride film deposited at  $600~\\mathrm{K}$  after  $140\\mathrm{AB}$  cycles on a  $125\\mathrm{\\AA}$  thick  $\\mathrm{SiO_2}$  film on  $\\mathrm{Si}(100)$ . The reactant exposures of  $12,000\\mathrm{L}\\mathrm{NH_3}$  and  $4300\\mathrm{L}\\mathrm{WE_2}$  were sufficient for complete half-reactions during each AB cycle. The light-to-dark range is  $25\\mathrm{\\AA}$ .","id":"train/atomic-layer-deposition/experimental-usecase/29/figure_10","sample_id":"atomic-layer-deposition/experimental-usecase/29/figure_10","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":1,"y":6,"width":676,"height":513}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/J. W. Klaus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:image_caption"},"width":683,"height":519,"image_format":"jpeg","image_sha256":"4d15477c997e8bf84ea41df2d602463ffed2a5179d1360a9f0f4457315188444","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_63_fig_4.jpg","caption":"FIG. 4. (Color online) 3D surface morphologies of (a) as-deposited  $(250^{\\circ}\\mathrm{C})$  and (b) annealed  $\\sim 26 \\mathrm{nm}$  thick  $\\mathrm{Ga}_2\\mathrm{O}_3$  thin films.","id":"train/atomic-layer-deposition/experimental-usecase/63/fig_4","sample_id":"atomic-layer-deposition/experimental-usecase/63/fig_4","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"3d scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"3D AFM surface topography of the as deposited Ga2O3 film (about 26 nm, deposited at 250 C), shown over a 1 µm by 1 µm scan area with height contrast in nm. The reported rms roughness is 0.16 nm.\"},{\"panel_id\":\"b\",\"text\":\"3D AFM surface topography of the annealed Ga2O3 film (about 26 nm), shown over a 1 µm by 1 µm scan area with height contrast in nm. The reported rms roughness increases to 0.37 nm after annealing, consistent with a rougher, more grain like surface.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| x-axis | y-axis | z-axis |\\n|--------|--------|--------|\\n| 0.0    | 0.0    | 0.0    |\\n| 250.0  | 250.0  | 0.0  |\\n| 500.0  | 500.0  | 0.0  |\\n| 750.0  | 750.0  | 0.0  |\\n| 1000.0 | 1000.0 | 0.0  |\"},{\"panel_id\":\"b\",\"text\":\"| x-axis | y-axis | z-axis |\\n|--------|--------|--------|\\n| 0.0    | 0.0    | 0.0    |\\n| 250.0  | 250.0  | 0.0  |\\n| 500.0  | 500.0  | 0.0  |\\n| 750.0  | 750.0  | 0.0  |\\n| 1000.0 | 1000.0 | 0.0  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The color scale represents surface height in nm across the scanned area.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The rms roughness increases from 0.16 nm (as deposited) to 0.37 nm (annealed).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The paper attributes the roughness increase to grain formation during crystallization upon annealing. As grains form and grow, the surface develops larger height variations compared with the amorphous as deposited film. This interpretation links the morphology change directly to the structural transition induced by annealing.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increased roughness can affect interface quality when Ga2O3 is integrated with metals or semiconductors, because local thickness and contact area vary across the surface. Rougher surfaces can also increase scattering and raise leakage risk in thin insulating layers, depending on the device stack. In exchange, annealing can deliver the crystalline phase and associated property improvements, so roughness becomes a tradeoff to manage.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":0,"width":635,"height":388},{"panel_id":"b","x":6,"y":402,"width":632,"height":381}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/63/Donmez et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"63","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:image_caption"},"width":642,"height":786,"image_format":"jpeg","image_sha256":"41c2b92e624d68da87e702841311b2ef087c704c734547bc3e5b40f78b293411","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_11.jpg","caption":"Figure 11. Atomic force microscopy (AFM) images of the surfaces etched using (a) ALE and (b) digital etching and (c) the as-grown surface before etching.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_11","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"3d scatter plot"},{"panel_id":"c","label":"3d scatter plot"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":30,"width":563,"height":315},{"panel_id":"b","x":626,"y":11,"width":560,"height":312},{"panel_id":"c","x":296,"y":436,"width":542,"height":362}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/Low-Damage and Self-Limiting (Al)GaN Etching Process.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"13","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":1186,"height":798,"image_format":"jpeg","image_sha256":"5c5b1963b56929ede4c66d4fe42024c5680165bbee269dd74653a26c2b219aa3","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_6.jpg","caption":"FIG. 6. (Color online) Surface morphology of the trench bottom for  $\\Gamma_{i} / \\Gamma_{n} = (a)0.01$  b  $10^{-3}$  and (c)  $10^{-4}$  in the passivation step after 25 pulses (87.5 s etch time). The color bar indicates the profile height, with dark representing deeper etching and light representing higher features.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_6","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"},{"panel_id":"b","label":"3d scatter plot"},{"panel_id":"c","label":"3d scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image shows the surface morphology of the trench bottom for Γi/Γn  = 0.01 in the passivation step after 25 pulses (87.5 s etch time).\"},{\"panel_id\":\"b\",\"text\":\"The image shows the surface morphology of the trench bottom for Γi/Γn  = 10^-3 in the passivation step after 25 pulses (87.5 s etch time).\"},{\"panel_id\":\"c\",\"text\":\"The image shows the surface morphology of the trench bottom Γi/Γn  = 10^-5i in the passivation step after 25 pulses (87.5 s etch time).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Image | Γᵢ/Γₙ Ratio | Surface Width (nm) | Height Scale |\\n|-------|-------------|----------------------|--------------|\\n| (a)   | 0.01        | 25                   | Low → High (Black → Red → Yellow) |\\n| (b)   | 10⁻³        | 25                   | Low → High (Black → Red → Yellow) |\\n| (c)   | 10⁻⁴        | 25                   | Low → High (Black → Red → Yellow) |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface for Γᵢ/Γₙ =10^-4 has a scattering of small divots caused by individual ion strikes and larger divots where several ion strikes stochastically occurred in close proximity during the same passivation step. The surface for Γᵢ/Γₙ=10^-3 has a similar pattern of larger divots, but with significantly more uniform roughening as well. The surface for Γᵢ/Γ=0.01 has few large divots with the surface being dominated by random roughness; however, the divots\\nwhich persist are larger and deeper.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"There is a competition between a roughening process, produced by stochastic ion impacts during the passivation phase, and a smoothing process during the otherwise ideal ALE cycle\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The overall roughness increases as the ion to neutral flux ratio increases\"}]}]","bbox":[{"panel_id":"a","x":3,"y":8,"width":574,"height":385},{"panel_id":"b","x":0,"y":388,"width":579,"height":341},{"panel_id":"c","x":1,"y":723,"width":576,"height":388}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":578,"height":1114,"image_format":"jpeg","image_sha256":"c73ead0e71dd35a25264c29156f3c541e48d52a4a2aa74129e15ddc0434c2e35","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_8.jpg","caption":"FIG. 8. (Color online) Surface morphology of the trench bottom with  $100\\mathrm{ppm}\\mathrm{Cl}_2$  in the ion bombardment step after 25 pulses (32 s etch time). The color bar indicates the profile height, with dark representing deeper etching and light representing higher features.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_8","subset":"3d-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"3d scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Surface morphology of the trench bottom with 100 ppm Cl2 in the ion bombardment step after 32 s etch time. The color bar indicates the profile height,\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Parameter        | Value      |\\n|------------------|-----------|\\n| Cl₂ concentration | 100 ppm   |\\n| Surface width     | 27 nm     |\\n| Height scale      | Low → High (color gradient: black → red → yellow) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface has the characteristic divots, but the surface is also slightly concave due to a higher etch rate in the center of the feature. This higher etch rate is due to a higher flux of Cl atoms incident onto the center of the feature. The concavity becomes more pronounced with longer etch times (more pulses).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface morphology generated by having a Cl flux during the ion bombardment phase differs slightly from when having an ion flux during the passivation phase. The height of the etched surface at the bottom of the AR=2 trench after 25 ALE pulses (32 s) shown in the image for the 100 ppm Cl2 case\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"27 nm\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":571,"height":441}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"3d scatter plot","caption_source":"content.json:img_caption"},"width":575,"height":447,"image_format":"jpeg","image_sha256":"10b781237ceb0fb2535052d1efb7ff2f8575938c6450449feb4987e9724b7b3d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/area-chart/train/metadata.jsonl

@@ -1 +1 @@
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_12.jpg","caption":"FIG. 12. (Color online) Trend of etch depth per cycle (EPC) as a function of ion bombardment time  $(T_{I})$  and passivation time  $(T_{P})$ . The solid line represents an EPC of 1, and the shaded region is the window from  $0.9 < \\mathrm{EPC} < 1.1$ .","id":"train/atomic-layer-etching/simulation-usecase/3/fig_12","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_12","subset":"area-chart","split":"train","classification":[{"panel_id":"a","label":"area chart"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"area chart","caption_source":"content.json:img_caption"},"width":575,"height":492,"image_format":"jpeg","image_sha256":"1b13cf1e19b524f8184a8bd08fb27db7ee2d703f2aa83cec9e8280b65dda37a2","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":2,"width":571,"height":484}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The area chart illustrates the relationship between etch per cycle (EPC) with respect to passivation time (\\\\(T_p\\\\)) and ion bombardment time (\\\\(T_I\\\\)), with shaded regions indicating specific ranges of EPC.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Passivation Time (Tp, ms) | Ion Bombardment Time (Ti, s) AR=2 | AR=4 | AR=6 |\\n|---------------------------|------------------------------------|------|------|\\n| 5                         | 1.30                               | 1.30 | 1.30 |\\n| 25                        | 0.70                               | 0.90 | 1.10 |\\n| 45                        | 0.40                               | 0.60 | 0.90 |\\n| 64                        | 0.30                               | 0.45 | 0.70 |\\n| 84                        | 0.25                               | 0.35 | 0.55 |\\n| 104                       | 0.20                               | 0.30 | 0.45 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"To investigate the pulse time parameter space, simulations were performed for TP from 5 to 105 ms, and TI from 0.1 to 1.4 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The solid lines are for conditions that produce EPC=1 ML, and the shaded areas indicate the range of 0.9<EPC<1.1 ML. Results are shown for aspect ratios of 2, 4 and 6. There are regions where EPC is near 1 ML and which are nearly independent of TI (nearly vertical in the image), depending dominantly\\non TP. These regions might be described as radical-starved for a continuous etch.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Other regions where EPC is near 1ML are almost independent of TP (nearly horizontal), depending dominantly on TI. These regions might be\\ndescribed as ion-starved for a continuous etch\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The width of the window for which 0.9<EPC<1.1 ML is largest at the transition between these two limiting regimes, indicating more tolerance to process variation\"}]}]"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_12.jpg","caption":"FIG. 12. (Color online) Trend of etch depth per cycle (EPC) as a function of ion bombardment time  $(T_{I})$  and passivation time  $(T_{P})$ . The solid line represents an EPC of 1, and the shaded region is the window from  $0.9 < \\mathrm{EPC} < 1.1$ .","id":"train/atomic-layer-etching/simulation-usecase/3/fig_12","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_12","subset":"area-chart","split":"train","classification":[{"panel_id":"a","label":"area chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The area chart illustrates the relationship between etch per cycle (EPC) with respect to passivation time (\\\\(T_p\\\\)) and ion bombardment time (\\\\(T_I\\\\)), with shaded regions indicating specific ranges of EPC.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Passivation Time (Tp, ms) | Ion Bombardment Time (Ti, s) AR=2 | AR=4 | AR=6 |\\n|---------------------------|------------------------------------|------|------|\\n| 5                         | 1.30                               | 1.30 | 1.30 |\\n| 25                        | 0.70                               | 0.90 | 1.10 |\\n| 45                        | 0.40                               | 0.60 | 0.90 |\\n| 64                        | 0.30                               | 0.45 | 0.70 |\\n| 84                        | 0.25                               | 0.35 | 0.55 |\\n| 104                       | 0.20                               | 0.30 | 0.45 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"To investigate the pulse time parameter space, simulations were performed for TP from 5 to 105 ms, and TI from 0.1 to 1.4 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The solid lines are for conditions that produce EPC=1 ML, and the shaded areas indicate the range of 0.9<EPC<1.1 ML. Results are shown for aspect ratios of 2, 4 and 6. There are regions where EPC is near 1 ML and which are nearly independent of TI (nearly vertical in the image), depending dominantly\\non TP. These regions might be described as radical-starved for a continuous etch.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Other regions where EPC is near 1ML are almost independent of TP (nearly horizontal), depending dominantly on TI. These regions might be\\ndescribed as ion-starved for a continuous etch\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The width of the window for which 0.9<EPC<1.1 ML is largest at the transition between these two limiting regimes, indicating more tolerance to process variation\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":571,"height":484}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"area chart","caption_source":"content.json:img_caption"},"width":575,"height":492,"image_format":"jpeg","image_sha256":"1b13cf1e19b524f8184a8bd08fb27db7ee2d703f2aa83cec9e8280b65dda37a2","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/band-diagram/test/metadata.jsonl

@@ -1,2 +1,2 @@
-{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_51_fig_9.jpg","caption":"Fig. 9 Sketch illustrating differences in  $\\mathrm{MoO}_x$  induced dark band bending above: high work function  $\\mathrm{MoO}_x$ , below: low work function  $\\mathrm{MoO}_x$","id":"test/atomic-layer-deposition/experimental-usecase/51/fig_9","sample_id":"atomic-layer-deposition/experimental-usecase/51/fig_9","subset":"band-diagram","split":"test","classification":[{"panel_id":"a","label":"band diagram"},{"panel_id":"b","label":"band diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/Ziegler et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"51","first_classification_panel_id":"a","first_classification_label":"band diagram","caption_source":"content.json:image_caption"},"width":656,"height":992,"image_format":"jpeg","image_sha256":"3f816a6645804e3ea4def7641d7ba8fbc9da5107835f98897a12f9e220f1ef82","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":4,"width":655,"height":499},{"panel_id":"b","x":2,"y":574,"width":650,"height":414}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_34_figure_11.jpg","caption":"Figure 11. DFT/vdW-DF2 level orbital-projected band structures along the M-Γ-K-M high-symmetry path of (left) free-standing and (right) Pt(111)-top-contacted a) pristine graphene (PG), b) graphene oxide (GO), and c) hydrogenated graphene (SSHG). The corresponding partial DOS per atom are also plotted. The energy levels (eV) are shifted by the Fermi energy  $(E_{\\mathrm{f}}$  dashed horizontal lines) and only energies within  $3\\mathrm{eV}$  of  $E_{\\mathrm{f}}$  are depicted. The DOS and band diagrams are LM-decomposed to show individual orbital contributions from selected atom types. Specific orbital contributions are depicted by colored beads (s: red;  $\\mathbb{P}_{\\mathrm{xy}}$ : green;  $\\mathbb{P}_{\\mathrm{z}}$ : blue; others: gray). The size of the beads indicates the weight of corresponding Bloch eigenstate of a given orbital. For the Pt-contacted graphene samples, contributions from Pt atoms are given by pale solid lines (for clarity) and labeled as “others.” The model systems used for band diagrams can be found in the insets of Figure 10.","id":"test/atomic-layer-deposition/simulation-usecase/34/figure_11","sample_id":"atomic-layer-deposition/simulation-usecase/34/figure_11","subset":"band-diagram","split":"test","classification":[{"panel_id":"a","label":"band diagram"},{"panel_id":"b","label":"band diagram"},{"panel_id":"c","label":"band diagram"},{"panel_id":"d","label":"band diagram"},{"panel_id":"e","label":"band diagram"},{"panel_id":"f","label":"band diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/images/figure_11.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/René H. J. Vervuurt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"band diagram","caption_source":"content.json:img_caption"},"width":1333,"height":1472,"image_format":"jpeg","image_sha256":"a4deb38edf97ee6f18e3aed514f27efae0df71fa6fd0dbc6815a69c81eac554a","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":1,"width":680,"height":521},{"panel_id":"b","x":707,"y":3,"width":609,"height":518},{"panel_id":"c","x":10,"y":535,"width":672,"height":448},{"panel_id":"d","x":703,"y":531,"width":618,"height":455},{"panel_id":"e","x":4,"y":995,"width":672,"height":472},{"panel_id":"f","x":718,"y":999,"width":596,"height":465}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
+{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_51_fig_9.jpg","caption":"Fig. 9 Sketch illustrating differences in  $\\mathrm{MoO}_x$  induced dark band bending above: high work function  $\\mathrm{MoO}_x$ , below: low work function  $\\mathrm{MoO}_x$","id":"test/atomic-layer-deposition/experimental-usecase/51/fig_9","sample_id":"atomic-layer-deposition/experimental-usecase/51/fig_9","subset":"band-diagram","split":"test","classification":[{"panel_id":"a","label":"band diagram"},{"panel_id":"b","label":"band diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":4,"width":655,"height":499},{"panel_id":"b","x":2,"y":574,"width":650,"height":414}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/51/Ziegler et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"51","first_classification_panel_id":"a","first_classification_label":"band diagram","caption_source":"content.json:image_caption"},"width":656,"height":992,"image_format":"jpeg","image_sha256":"3f816a6645804e3ea4def7641d7ba8fbc9da5107835f98897a12f9e220f1ef82","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_34_figure_11.jpg","caption":"Figure 11. DFT/vdW-DF2 level orbital-projected band structures along the M-Γ-K-M high-symmetry path of (left) free-standing and (right) Pt(111)-top-contacted a) pristine graphene (PG), b) graphene oxide (GO), and c) hydrogenated graphene (SSHG). The corresponding partial DOS per atom are also plotted. The energy levels (eV) are shifted by the Fermi energy  $(E_{\\mathrm{f}}$  dashed horizontal lines) and only energies within  $3\\mathrm{eV}$  of  $E_{\\mathrm{f}}$  are depicted. The DOS and band diagrams are LM-decomposed to show individual orbital contributions from selected atom types. Specific orbital contributions are depicted by colored beads (s: red;  $\\mathbb{P}_{\\mathrm{xy}}$ : green;  $\\mathbb{P}_{\\mathrm{z}}$ : blue; others: gray). The size of the beads indicates the weight of corresponding Bloch eigenstate of a given orbital. For the Pt-contacted graphene samples, contributions from Pt atoms are given by pale solid lines (for clarity) and labeled as “others.” The model systems used for band diagrams can be found in the insets of Figure 10.","id":"test/atomic-layer-deposition/simulation-usecase/34/figure_11","sample_id":"atomic-layer-deposition/simulation-usecase/34/figure_11","subset":"band-diagram","split":"test","classification":[{"panel_id":"a","label":"band diagram"},{"panel_id":"b","label":"band diagram"},{"panel_id":"c","label":"band diagram"},{"panel_id":"d","label":"band diagram"},{"panel_id":"e","label":"band diagram"},{"panel_id":"f","label":"band diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":2,"y":1,"width":680,"height":521},{"panel_id":"b","x":707,"y":3,"width":609,"height":518},{"panel_id":"c","x":10,"y":535,"width":672,"height":448},{"panel_id":"d","x":703,"y":531,"width":618,"height":455},{"panel_id":"e","x":4,"y":995,"width":672,"height":472},{"panel_id":"f","x":718,"y":999,"width":596,"height":465}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/images/figure_11.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/René H. J. Vervuurt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"band diagram","caption_source":"content.json:img_caption"},"width":1333,"height":1472,"image_format":"jpeg","image_sha256":"a4deb38edf97ee6f18e3aed514f27efae0df71fa6fd0dbc6815a69c81eac554a","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/band-diagram/train/metadata.jsonl

@@ -1 +1 @@
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_25_figure_1.jpg","caption":"Figure 1. Estimated range of  $E_{c}$  and  $E_{\\mathrm{v}}$  positions as a function of the cation ratio in ZGO (blue) and TGO (red) films based on previous findings for the  $E_{c}$  and  $E_{\\mathrm{v}}$  positions of  $\\mathrm{ZnO}$ ,  $\\mathrm{SnO_y}$ , and  $\\mathrm{GeO_y}$  21-23","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_1","subset":"band-diagram","split":"train","classification":[{"panel_id":"a","label":"band diagram"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/Adam Hultqvist et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"band diagram","caption_source":"content.json:image_caption"},"width":400,"height":305,"image_format":"jpeg","image_sha256":"bb7ac9fe9d78f0fc26186bb70b6f610720cf6561331071f4b8d0bb7dc2296eee","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":1,"width":397,"height":303}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The band diagram shows the energy levels of ZnO, SnOy, Zn1-xGexOy (ZGO), and Sn1-xGexOy (TGO) materials plotted against the cation ratio, x.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Material                | Conduction Band (E_c) [eV] | Valence Band (E_v) [eV] |\\n|-------------------------|-----------------------------|---------------------------|\\n| ZnO                    | -4.5                        | -7.5                      |\\n| SnO_y                  | -4.8                        | -8.0                      |\\n| ZGO (Zn₁₋ₓGeₓO_y)      | -4.5 → -4.0 (as x → 1)      | -7.5 → -8.5 (as x → 1)    |\\n| TGO (Sn₁₋ₓGeₓO_y)      | -4.8 → -4.0 (as x → 1)      | -8.0 → -8.8 (as x → 1)    |\\n| GeO_y                  | -4.0                        | -8.8                      |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"To make ESL compounds with high and adjustable Ec through an ALD supercycle scheme by first selecting GeOy due to its high Ec placement and 4−6 eV\\nenergy band gap (Eg). In order to adjust the Ec of the resulting ternary compound, we mixed the GeOy process with either the ZnO or SnOy process, both having lower Ec placements.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Ideally, the resulting ternary Zn1−xGexOy (ZGO) and Sn1−xGexOy (TGO) compounds should have Ec positions that shift to higher energies as the Ge content of the compounds increases\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ESL Ec placements for the solar cell performance were identified for each of the ternary compounds\\n1. Zn(O,S),\\n2. (Zn,Mg)O\\n3.  Zn1−xSnxOy (ZTO)\\n4. ZnInxSy\\n5. ZnTiO\\n6. SnGaOx (TGAO)\\n7. In2(S,O)3\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Certain placement of the energy conduction band level (Ec) in the ESL was desired for optimal charge extraction from the ACIGS absorber and minimized interface recombination.\"}]}]"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_25_figure_1.jpg","caption":"Figure 1. Estimated range of  $E_{c}$  and  $E_{\\mathrm{v}}$  positions as a function of the cation ratio in ZGO (blue) and TGO (red) films based on previous findings for the  $E_{c}$  and  $E_{\\mathrm{v}}$  positions of  $\\mathrm{ZnO}$ ,  $\\mathrm{SnO_y}$ , and  $\\mathrm{GeO_y}$  21-23","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_1","subset":"band-diagram","split":"train","classification":[{"panel_id":"a","label":"band diagram"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The band diagram shows the energy levels of ZnO, SnOy, Zn1-xGexOy (ZGO), and Sn1-xGexOy (TGO) materials plotted against the cation ratio, x.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Material                | Conduction Band (E_c) [eV] | Valence Band (E_v) [eV] |\\n|-------------------------|-----------------------------|---------------------------|\\n| ZnO                    | -4.5                        | -7.5                      |\\n| SnO_y                  | -4.8                        | -8.0                      |\\n| ZGO (Zn₁₋ₓGeₓO_y)      | -4.5 → -4.0 (as x → 1)      | -7.5 → -8.5 (as x → 1)    |\\n| TGO (Sn₁₋ₓGeₓO_y)      | -4.8 → -4.0 (as x → 1)      | -8.0 → -8.8 (as x → 1)    |\\n| GeO_y                  | -4.0                        | -8.8                      |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"To make ESL compounds with high and adjustable Ec through an ALD supercycle scheme by first selecting GeOy due to its high Ec placement and 4−6 eV\\nenergy band gap (Eg). In order to adjust the Ec of the resulting ternary compound, we mixed the GeOy process with either the ZnO or SnOy process, both having lower Ec placements.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Ideally, the resulting ternary Zn1−xGexOy (ZGO) and Sn1−xGexOy (TGO) compounds should have Ec positions that shift to higher energies as the Ge content of the compounds increases\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ESL Ec placements for the solar cell performance were identified for each of the ternary compounds\\n1. Zn(O,S),\\n2. (Zn,Mg)O\\n3.  Zn1−xSnxOy (ZTO)\\n4. ZnInxSy\\n5. ZnTiO\\n6. SnGaOx (TGAO)\\n7. In2(S,O)3\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Certain placement of the energy conduction band level (Ec) in the ESL was desired for optimal charge extraction from the ACIGS absorber and minimized interface recombination.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":397,"height":303}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/Adam Hultqvist et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"band diagram","caption_source":"content.json:image_caption"},"width":400,"height":305,"image_format":"jpeg","image_sha256":"bb7ac9fe9d78f0fc26186bb70b6f610720cf6561331071f4b8d0bb7dc2296eee","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/bar-chart/test/metadata.jsonl

@@ -1,6 +1,6 @@
-{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_10_fig_2.jpg","caption":"FIG. 2. (Color online) Number of publications per year on the subject of plasma-assisted ALD, between 1991 and 2011 (status May 31, 2011). The search was run in published abstracts using Web of Science (Ref. 23). The search terms included \"plasma-assisted ALD,\" \"plasma-enhanced ALD,\" \"radical enhanced ALD,\" \"remote plasma ALD,\" \"direct plasma ALD,\" and \"plasma ALD.\" The first report of a plasma-assisted ALD process by De Keijser and Van Opdorp (Philips Research Laboratories, Eindhoven), published in 1991, is also included.","id":"test/atomic-layer-deposition/experimental-usecase/10/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/10/fig_2","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/H. B. Profijt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":684,"height":525,"image_format":"jpeg","image_sha256":"cc44f53361d1a94a999978171e546bb4aa7c06c0565a85a5cc3e08c92cde45bf","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":684,"height":525}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart presents the number of publications per year referencing plasma-assisted ALD and related terms. It shows minimal activity before 2000, followed by rapid growth after 2005, indicating a strong rise in research interest and adoption of plasma-based ALD methods.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Year | Publications per Year |\\n|------|-----------------------|\\n| 1990 | 1                     |\\n| 1995 | 1                     |\\n| 2000 | 1                     |\\n| 2005 | 10                    |\\n| 2010 | 55                    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The number of publications remained low through the 1990s but began increasing sharply after 2000, reaching a significant rise after 2005, reflecting growing research and industrial interest in plasma-assisted ALD.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Each bar represents the total number of publications per year that include plasma-assisted ALD and related keywords.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Plasma-assisted, Plasma-enhanced, Radical-enhanced, Remote plasma, Direct plasma ALD\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
-{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_39_fig_12.jpg","caption":"FIG. 12. (Color online) Bar graph mass spectrum during the  $\\mathrm{O}_2$  plasma step as composed from the time-resolved quadrupole mass spectrometry (QMS) measurements carried out in the Oxford Instruments FlexAL reactor. The spectrum has been corrected for (background) signals during regular  $\\mathrm{O}_2$  plasma operation and shows therefore only species that are related to the ALD surface reactions taking place during the  $\\mathrm{O}_2$  plasma step. The most likely ions contributing to the signals at the selected mass-to-charge ratios  $(m / z)$  are indicated.","id":"test/atomic-layer-deposition/experimental-usecase/39/fig_12","sample_id":"atomic-layer-deposition/experimental-usecase/39/fig_12","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/images/fig_12.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/Heil et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"39","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:image_caption"},"width":673,"height":528,"image_format":"jpeg","image_sha256":"95fdb71c7e563f9ee8a0c2e4e2623bd5e404b80a8c4116f2760d1339e73d3b58","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":673,"height":528}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart displays the peak ion current (A) of various ions (CH<sub>x</sub><sup>+</sup>, CO<sup>+</sup>, C<sub>2</sub>H<sub>x</sub><sup>+</sup>, CO<sub>2</sub><sup>+</sup>) against mass (amu). The y-axis represents the peak ion current on a logarithmic scale, while the x-axis shows the mass in atomic mass units.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Mass (amu) | QMS: Peak ion current (A) |\\n|---|---|\\n| 12 | 2*10^-10 |\\n| 13 | 8*10^-11 |\\n| 14 | 1*10^-11 |\\n| 15 | 2*10^-10 |\\n| 24 | 2*10^-13 |\\n| 25 | 4*10^-13 |\\n| 26 | 2*10^-12 |\\n| 27 | 2*10^-12 |\\n| 28 | 3*10^-11 |\\n| 29 | 3*10^-12 |\\n| 30 | 3*10^-12 |\\n| 44 | 7*10^-12 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 12 - 15 amu.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"28 amu.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 24 - 30 amu.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"44 amu.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_18_fig_2.jpg","caption":"FIG. 2. Distribution of coordination numbers for the bulk  $\\mathsf{aHfO_2}$  model used in this study that was prepared using classical molecular dynamics.","id":"test/atomic-layer-etching/simulation-usecase/18/fig_2","sample_id":"atomic-layer-etching/simulation-usecase/18/fig_2","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/images/fig_2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/Origin of enhanced thermal atomic layer etching of amorphous HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":669,"height":447,"image_format":"jpeg","image_sha256":"620c91904954add8d0c747ee04122658ae6170a60afbcd76565446d5d6ee77be","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":4,"y":3,"width":661,"height":443}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows a bar chart of the number of atoms at various coordination numbers for oxygen (O) and hafnium (Hf). Oxygen has a higher count at coordination numbers 3, while hafnium shows a peak at coordination number 7.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Coordination Number | O | Hf |\\n|---|---|---|\\n| 2 | 3 | - |\\n| 3 | 98 | - |\\n| 4 | 42 | - |\\n| 5 | - | 2 |\\n| 6 | - | 32 |\\n| 7 | - | 34 |\\n| 8 | - | 4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3 or 4 for oxygen, and 7 for Hf.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of oxygen atoms increases for 2 to 3 as a coordination number. Then it decreases for 4.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen, the oxygen atoms has a higher number of atoms at a lower coordination number.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Theoretical studies require knowledge of the coordination number. For this reason, it is important to determine the amount that dominates the bulk model.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc.jpg","caption":"","id":"test/atomic-layer-etching/simulation-usecase/34/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc","sample_id":"atomic-layer-etching/simulation-usecase/34/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"not_found"},"width":509,"height":436,"image_format":"jpeg","image_sha256":"814910c13b589a6d68b429060e588444be116c5e66859e1936dea6b4909a50bc","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":509,"height":436}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart shows the distribution of coverage fraction values relative to the mean. Most measurements fall between −σ and +σ, with the highest frequencies near the mean. Very few values occur at the extreme ±3σ bins, indicating a narrow spread around μ.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Coverage Fraction | Frequency |\\n|-------------------|-----------|\\n| -3σ              | 1         |\\n| -2σ              | 4         |\\n| -σ               | 20        |\\n| μ                 | 19        |\\n| +σ                | 16        |\\n| +2σ              | 10        |\\n| +3σ              | 0         |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The distribution is strongly centered around the mean. The highest frequencies occur near μ and within ±σ, while the number of observations drops off as the coverage fraction moves toward ±2σ and ±3σ. This indicates that most values cluster close to the mean with relatively few outliers.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"μ = 0.9814.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The −σ bin, the mean (μ) bin, and the +σ bin.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9.jpg","caption":"","id":"test/atomic-layer-etching/simulation-usecase/34/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9","sample_id":"atomic-layer-etching/simulation-usecase/34/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9","subset":"bar-chart","split":"test","classification":[{"panel_id":"b","label":"bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"b","first_classification_label":"bar chart","caption_source":"not_found"},"width":503,"height":436,"image_format":"jpeg","image_sha256":"bea52bc8f608d6e1fd4f5134a702eb921207fb49f3ad1217d880ec9f3ecacbbb","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"b","x":0,"y":0,"width":503,"height":436}],"summarization":"[{\"panel_id\":\"b\",\"text\":\"The bar chart shows the distribution of etching fraction values relative to the mean. Most measurements cluster around μ = 0.7777, with the highest frequency near the center. The number of observations decreases toward the ±3σ bins, indicating a relatively narrow spread with few extreme values.\"}]","data_extraction":"[{\"panel_id\":\"b\",\"text\":\"| Etching Fraction | Frequency |\\n|------------------|-----------|\\n|-3σ              | 1         |\\n|-2σ              | 4         |\\n|-σ               | 20        |\\n|μ                 | 25        |\\n|+σ                | 12        |\\n|+2σ               | 5         |\\n|+3σ               | 2         |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The distribution is centered around the mean, with most values occurring within ±σ. Frequencies drop off as the etching fraction moves toward ±2σ and ±3σ, indicating that extreme values are relatively uncommon.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"μ = 0.7777.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The μ bin, the −σ bin, and the +σ bin.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_fig_11.jpg","caption":"Fig. 11. Histograms depicting the distribution of the coverage fraction (a), etching fraction (b), and etching per cycle (c) for a sample size of 50 multiscale CFD data points for 20 sccm HF flow rate, 40 sccm TMA flow rate, and  $80~\\mathrm{mm / s}$  substrate velocity without the influence of disturbances. Each bar represents a standard deviation,  $\\sigma$  from the mean,  $\\mu$  of the data set.","id":"test/atomic-layer-etching/simulation-usecase/34/fig_11","sample_id":"atomic-layer-etching/simulation-usecase/34/fig_11","subset":"bar-chart","split":"test","classification":[{"panel_id":"c","label":"bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_11.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"c","first_classification_label":"bar chart","caption_source":"content.json:image_caption"},"width":512,"height":445,"image_format":"jpeg","image_sha256":"60baea328e7fdfc8611054a5828426acc4bdacf4cd0722ce6fea8a444e8ee57e","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"c","x":0,"y":0,"width":512,"height":445}],"summarization":"[{\"panel_id\":\"c\",\"text\":\"The bar chart shows the distribution of etching per cycle values. Most measurements cluster around the mean (μ = 0.3511 Å/cycle), with the highest frequencies near μ and −σ. Frequencies decrease toward the ±3σ bins, indicating relatively few extreme values.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Etching per Cycle (Å/cycle) | Frequency |\\n|-----------------------------|-----------|\\n| -3σ                         | 1         |\\n| -2σ                         | 4         |\\n| -σ                          | 23        |\\n| μ                           | 15        |\\n| +σ                          | 14        |\\n| +2σ                         | 5         |\\n| +3σ                         | 2         |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The distribution is centered around the mean, with most values occurring near μ and within ±σ. Frequencies decrease toward the outer bins. This shows that extreme etching rates are relatively uncommon.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"μ = 0.3511 Å/cycle.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"−σ bin, μ bin.\"}]}]"}
+{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_10_fig_2.jpg","caption":"FIG. 2. (Color online) Number of publications per year on the subject of plasma-assisted ALD, between 1991 and 2011 (status May 31, 2011). The search was run in published abstracts using Web of Science (Ref. 23). The search terms included \"plasma-assisted ALD,\" \"plasma-enhanced ALD,\" \"radical enhanced ALD,\" \"remote plasma ALD,\" \"direct plasma ALD,\" and \"plasma ALD.\" The first report of a plasma-assisted ALD process by De Keijser and Van Opdorp (Philips Research Laboratories, Eindhoven), published in 1991, is also included.","id":"test/atomic-layer-deposition/experimental-usecase/10/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/10/fig_2","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart presents the number of publications per year referencing plasma-assisted ALD and related terms. It shows minimal activity before 2000, followed by rapid growth after 2005, indicating a strong rise in research interest and adoption of plasma-based ALD methods.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Year | Publications per Year |\\n|------|-----------------------|\\n| 1990 | 1                     |\\n| 1995 | 1                     |\\n| 2000 | 1                     |\\n| 2005 | 10                    |\\n| 2010 | 55                    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The number of publications remained low through the 1990s but began increasing sharply after 2000, reaching a significant rise after 2005, reflecting growing research and industrial interest in plasma-assisted ALD.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Each bar represents the total number of publications per year that include plasma-assisted ALD and related keywords.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Plasma-assisted, Plasma-enhanced, Radical-enhanced, Remote plasma, Direct plasma ALD\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":684,"height":525}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/H. B. Profijt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":684,"height":525,"image_format":"jpeg","image_sha256":"cc44f53361d1a94a999978171e546bb4aa7c06c0565a85a5cc3e08c92cde45bf","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_39_fig_12.jpg","caption":"FIG. 12. (Color online) Bar graph mass spectrum during the  $\\mathrm{O}_2$  plasma step as composed from the time-resolved quadrupole mass spectrometry (QMS) measurements carried out in the Oxford Instruments FlexAL reactor. The spectrum has been corrected for (background) signals during regular  $\\mathrm{O}_2$  plasma operation and shows therefore only species that are related to the ALD surface reactions taking place during the  $\\mathrm{O}_2$  plasma step. The most likely ions contributing to the signals at the selected mass-to-charge ratios  $(m / z)$  are indicated.","id":"test/atomic-layer-deposition/experimental-usecase/39/fig_12","sample_id":"atomic-layer-deposition/experimental-usecase/39/fig_12","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart displays the peak ion current (A) of various ions (CH<sub>x</sub><sup>+</sup>, CO<sup>+</sup>, C<sub>2</sub>H<sub>x</sub><sup>+</sup>, CO<sub>2</sub><sup>+</sup>) against mass (amu). The y-axis represents the peak ion current on a logarithmic scale, while the x-axis shows the mass in atomic mass units.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Mass (amu) | QMS: Peak ion current (A) |\\n|---|---|\\n| 12 | 2*10^-10 |\\n| 13 | 8*10^-11 |\\n| 14 | 1*10^-11 |\\n| 15 | 2*10^-10 |\\n| 24 | 2*10^-13 |\\n| 25 | 4*10^-13 |\\n| 26 | 2*10^-12 |\\n| 27 | 2*10^-12 |\\n| 28 | 3*10^-11 |\\n| 29 | 3*10^-12 |\\n| 30 | 3*10^-12 |\\n| 44 | 7*10^-12 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 12 - 15 amu.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"28 amu.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 24 - 30 amu.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"44 amu.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":673,"height":528}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/images/fig_12.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/39/Heil et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"39","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:image_caption"},"width":673,"height":528,"image_format":"jpeg","image_sha256":"95fdb71c7e563f9ee8a0c2e4e2623bd5e404b80a8c4116f2760d1339e73d3b58","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_18_fig_2.jpg","caption":"FIG. 2. Distribution of coordination numbers for the bulk  $\\mathsf{aHfO_2}$  model used in this study that was prepared using classical molecular dynamics.","id":"test/atomic-layer-etching/simulation-usecase/18/fig_2","sample_id":"atomic-layer-etching/simulation-usecase/18/fig_2","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows a bar chart of the number of atoms at various coordination numbers for oxygen (O) and hafnium (Hf). Oxygen has a higher count at coordination numbers 3, while hafnium shows a peak at coordination number 7.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Coordination Number | O | Hf |\\n|---|---|---|\\n| 2 | 3 | - |\\n| 3 | 98 | - |\\n| 4 | 42 | - |\\n| 5 | - | 2 |\\n| 6 | - | 32 |\\n| 7 | - | 34 |\\n| 8 | - | 4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3 or 4 for oxygen, and 7 for Hf.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of oxygen atoms increases for 2 to 3 as a coordination number. Then it decreases for 4.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen, the oxygen atoms has a higher number of atoms at a lower coordination number.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Theoretical studies require knowledge of the coordination number. For this reason, it is important to determine the amount that dominates the bulk model.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":3,"width":661,"height":443}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/images/fig_2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/Origin of enhanced thermal atomic layer etching of amorphous HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":669,"height":447,"image_format":"jpeg","image_sha256":"620c91904954add8d0c747ee04122658ae6170a60afbcd76565446d5d6ee77be","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc.jpg","caption":"","id":"test/atomic-layer-etching/simulation-usecase/34/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc","sample_id":"atomic-layer-etching/simulation-usecase/34/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc","subset":"bar-chart","split":"test","classification":[{"panel_id":"a","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart shows the distribution of coverage fraction values relative to the mean. Most measurements fall between −σ and +σ, with the highest frequencies near the mean. Very few values occur at the extreme ±3σ bins, indicating a narrow spread around μ.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Coverage Fraction | Frequency |\\n|-------------------|-----------|\\n| -3σ              | 1         |\\n| -2σ              | 4         |\\n| -σ               | 20        |\\n| μ                 | 19        |\\n| +σ                | 16        |\\n| +2σ              | 10        |\\n| +3σ              | 0         |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The distribution is strongly centered around the mean. The highest frequencies occur near μ and within ±σ, while the number of observations drops off as the coverage fraction moves toward ±2σ and ±3σ. This indicates that most values cluster close to the mean with relatively few outliers.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"μ = 0.9814.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The −σ bin, the mean (μ) bin, and the +σ bin.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":509,"height":436}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/04961dd67c8a719f41a4bdbc44ee1c49b3357525fdfa15b93e9df870f3b868cc.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"not_found"},"width":509,"height":436,"image_format":"jpeg","image_sha256":"814910c13b589a6d68b429060e588444be116c5e66859e1936dea6b4909a50bc","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9.jpg","caption":"","id":"test/atomic-layer-etching/simulation-usecase/34/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9","sample_id":"atomic-layer-etching/simulation-usecase/34/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9","subset":"bar-chart","split":"test","classification":[{"panel_id":"b","label":"bar chart"}],"summarization":"[{\"panel_id\":\"b\",\"text\":\"The bar chart shows the distribution of etching fraction values relative to the mean. Most measurements cluster around μ = 0.7777, with the highest frequency near the center. The number of observations decreases toward the ±3σ bins, indicating a relatively narrow spread with few extreme values.\"}]","data_extraction":"[{\"panel_id\":\"b\",\"text\":\"| Etching Fraction | Frequency |\\n|------------------|-----------|\\n|-3σ              | 1         |\\n|-2σ              | 4         |\\n|-σ               | 20        |\\n|μ                 | 25        |\\n|+σ                | 12        |\\n|+2σ               | 5         |\\n|+3σ               | 2         |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The distribution is centered around the mean, with most values occurring within ±σ. Frequencies drop off as the etching fraction moves toward ±2σ and ±3σ, indicating that extreme values are relatively uncommon.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"μ = 0.7777.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The μ bin, the −σ bin, and the +σ bin.\"}]}]","bbox":[{"panel_id":"b","x":0,"y":0,"width":503,"height":436}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/2d6ab73da07b7b6bbf9a4fc1ae8384de54332a287fa24f422745afeb7863d9c9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"b","first_classification_label":"bar chart","caption_source":"not_found"},"width":503,"height":436,"image_format":"jpeg","image_sha256":"bea52bc8f608d6e1fd4f5134a702eb921207fb49f3ad1217d880ec9f3ecacbbb","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_fig_11.jpg","caption":"Fig. 11. Histograms depicting the distribution of the coverage fraction (a), etching fraction (b), and etching per cycle (c) for a sample size of 50 multiscale CFD data points for 20 sccm HF flow rate, 40 sccm TMA flow rate, and  $80~\\mathrm{mm / s}$  substrate velocity without the influence of disturbances. Each bar represents a standard deviation,  $\\sigma$  from the mean,  $\\mu$  of the data set.","id":"test/atomic-layer-etching/simulation-usecase/34/fig_11","sample_id":"atomic-layer-etching/simulation-usecase/34/fig_11","subset":"bar-chart","split":"test","classification":[{"panel_id":"c","label":"bar chart"}],"summarization":"[{\"panel_id\":\"c\",\"text\":\"The bar chart shows the distribution of etching per cycle values. Most measurements cluster around the mean (μ = 0.3511 Å/cycle), with the highest frequencies near μ and −σ. Frequencies decrease toward the ±3σ bins, indicating relatively few extreme values.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Etching per Cycle (Å/cycle) | Frequency |\\n|-----------------------------|-----------|\\n| -3σ                         | 1         |\\n| -2σ                         | 4         |\\n| -σ                          | 23        |\\n| μ                           | 15        |\\n| +σ                          | 14        |\\n| +2σ                         | 5         |\\n| +3σ                         | 2         |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The distribution is centered around the mean, with most values occurring near μ and within ±σ. Frequencies decrease toward the outer bins. This shows that extreme etching rates are relatively uncommon.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"μ = 0.3511 Å/cycle.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"−σ bin, μ bin.\"}]}]","bbox":[{"panel_id":"c","x":0,"y":0,"width":512,"height":445}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_11.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"c","first_classification_label":"bar chart","caption_source":"content.json:image_caption"},"width":512,"height":445,"image_format":"jpeg","image_sha256":"60baea328e7fdfc8611054a5828426acc4bdacf4cd0722ce6fea8a444e8ee57e","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/bar-chart/validation/metadata.jsonl

@@ -1,2 +1,2 @@
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_15_fig_3.jpg","caption":"FIG. 3. Number of ALE publications per year from 1988 to 2014. The first wave of publications occurred in the 1990s. A resurgence of ALE activity is now occurring.","id":"validation/atomic-layer-etching/experimental-usecase/15/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/15/fig_3","subset":"bar-chart","split":"validation","classification":[{"panel_id":"a","label":"bar chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/images/fig_3.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/Overview of atomic layer etching in the semiconductor industry.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"15","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":648,"height":569,"image_format":"jpeg","image_sha256":"5e2792e29a49b6314f82fdef2bbf0a10850174c5574e35603835696ede8998de","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":645,"height":569}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart illustrates the number of ALE reports per year from 1988 to 2014, showing a significant increase during the 'Resurgence' period (approximately after 2006).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Year | Number of ALE Reports Per Year |\\n|---|---|\\n| 1988 | 2 |\\n| 1990 | 4 |\\n| 1992 | 15 |\\n| 1994 | 16 |\\n| 1996 | 8 |\\n| 1998 | 7 |\\n| 2000 | 8 |\\n| 2002 | 8 |\\n| 2004 | 5 |\\n| 2006 | 17 |\\n| 2008 | 22 |\\n| 2010 | 5 |\\n| 2012 | 28 |\\n| 2014 | 55 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 1990 and 1998.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of ALE reports have been increased from 1988 to 2014, with a slight decrease in 2001.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2014.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Based on the behavior shown in the figure and the additional publications that have been published over the year, it is reasonable to expect an increase in the number of ALE reports after 2014.\"}]}]"}
-{"file_name":"images/validation_atomic-layer-etching_simulation-usecase_22_fig_3.jpg","caption":"FIG. 3. Number of ALE publications per year from 1988 to 2014. The first wave of publications occurred in the 1990s. A resurgence of ALE activity is now occurring.","id":"validation/atomic-layer-etching/simulation-usecase/22/fig_3","sample_id":"atomic-layer-etching/simulation-usecase/22/fig_3","subset":"bar-chart","split":"validation","classification":[{"panel_id":"a","label":"bar chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/images/fig_3.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/Review Paper -- Overview of atomic layer etching in the semiconductor industry.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":648,"height":569,"image_format":"jpeg","image_sha256":"5e2792e29a49b6314f82fdef2bbf0a10850174c5574e35603835696ede8998de","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":1,"width":643,"height":564}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart illustrates the number of ALE reports per year from 1988 to 2014, showing a significant increase during the 'Resurgence' period (between 2006 and 2014).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Year | Number of ALE Reports Per Year |\\n|---|---|\\n| 1988 | 2 |\\n| 1990 | 4 |\\n| 1992 | 15 |\\n| 1994 | 16 |\\n| 1996 | 8 |\\n| 1998 | 7 |\\n| 2000 | 8 |\\n| 2002 | 8 |\\n| 2004 | 5 |\\n| 2006 | 17 |\\n| 2008 | 22 |\\n| 2010 | 5 |\\n| 2012 | 28 |\\n| 2014 | 55 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 1990 and 1998.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of ALE reports have been increased from 1998 to 2014, with a slight decrease in 2001.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2014.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Based on the behavior shown in the figure and the additional publications that have been published over the year, it is reasonable to expect an increase in the number of ALE reports after 2014.\"}]}]"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_15_fig_3.jpg","caption":"FIG. 3. Number of ALE publications per year from 1988 to 2014. The first wave of publications occurred in the 1990s. A resurgence of ALE activity is now occurring.","id":"validation/atomic-layer-etching/experimental-usecase/15/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/15/fig_3","subset":"bar-chart","split":"validation","classification":[{"panel_id":"a","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart illustrates the number of ALE reports per year from 1988 to 2014, showing a significant increase during the 'Resurgence' period (approximately after 2006).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Year | Number of ALE Reports Per Year |\\n|---|---|\\n| 1988 | 2 |\\n| 1990 | 4 |\\n| 1992 | 15 |\\n| 1994 | 16 |\\n| 1996 | 8 |\\n| 1998 | 7 |\\n| 2000 | 8 |\\n| 2002 | 8 |\\n| 2004 | 5 |\\n| 2006 | 17 |\\n| 2008 | 22 |\\n| 2010 | 5 |\\n| 2012 | 28 |\\n| 2014 | 55 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 1990 and 1998.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of ALE reports have been increased from 1988 to 2014, with a slight decrease in 2001.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2014.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Based on the behavior shown in the figure and the additional publications that have been published over the year, it is reasonable to expect an increase in the number of ALE reports after 2014.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":645,"height":569}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/images/fig_3.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/15/Overview of atomic layer etching in the semiconductor industry.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"15","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":648,"height":569,"image_format":"jpeg","image_sha256":"5e2792e29a49b6314f82fdef2bbf0a10850174c5574e35603835696ede8998de","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/validation_atomic-layer-etching_simulation-usecase_22_fig_3.jpg","caption":"FIG. 3. Number of ALE publications per year from 1988 to 2014. The first wave of publications occurred in the 1990s. A resurgence of ALE activity is now occurring.","id":"validation/atomic-layer-etching/simulation-usecase/22/fig_3","sample_id":"atomic-layer-etching/simulation-usecase/22/fig_3","subset":"bar-chart","split":"validation","classification":[{"panel_id":"a","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The bar chart illustrates the number of ALE reports per year from 1988 to 2014, showing a significant increase during the 'Resurgence' period (between 2006 and 2014).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Year | Number of ALE Reports Per Year |\\n|---|---|\\n| 1988 | 2 |\\n| 1990 | 4 |\\n| 1992 | 15 |\\n| 1994 | 16 |\\n| 1996 | 8 |\\n| 1998 | 7 |\\n| 2000 | 8 |\\n| 2002 | 8 |\\n| 2004 | 5 |\\n| 2006 | 17 |\\n| 2008 | 22 |\\n| 2010 | 5 |\\n| 2012 | 28 |\\n| 2014 | 55 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 1990 and 1998.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of ALE reports have been increased from 1998 to 2014, with a slight decrease in 2001.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2014.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Based on the behavior shown in the figure and the additional publications that have been published over the year, it is reasonable to expect an increase in the number of ALE reports after 2014.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":643,"height":564}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/images/fig_3.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/simulation-usecase/22/Review Paper -- Overview of atomic layer etching in the semiconductor industry.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"bar chart","caption_source":"content.json:img_caption"},"width":648,"height":569,"image_format":"jpeg","image_sha256":"5e2792e29a49b6314f82fdef2bbf0a10850174c5574e35603835696ede8998de","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/box-plot/test/metadata.jsonl

@@ -1,4 +1,4 @@
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG4_a.jpg","caption":"","id":"test/atomic-layer-etching/experimental-usecase/47/FIG4_a","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG4_a","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_a.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_a.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"not_found"},"width":675,"height":575,"image_format":"jpeg","image_sha256":"9b9b64f5f899d85f8a99d620f551e42929dd18301d98c8a3cce103cb104839c9","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":8,"y":5,"width":662,"height":566}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch per cycle EPC (in Å) is plotted against oxygen pulse length (in seconds). As the oxygen pulse length increases, there is a corresponding increase in EPC.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| O₂ pulse length (s) | EPC (Å) |\\n|---|---|\\n| 1 | 0.4 |\\n| 2 | 0.8 |\\n| 3 | 1.2 |\\n| 6 | 1.8 |\\n| 10 | 2.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The pulse is considered to saturate when increasing the pulse length no longer affects the EPC of the process.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The O2 pulse was studied first at 300 °C. The length of the NbCl5 pulse was kept at 4 s and the purge lengths at 1 s. It was found that the O2 pulse does not fully saturate even with 10 s pulse times, but the increase in EPC slows down gradually.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No, not O2 pulse.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the interest of optimizing the etch rate per time unit, 3 s was chosen as the O2 pulse length for further experiments.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG4_b.jpg","caption":"FIG. 4. EPC of molybdenum at  $300^{\\circ}\\mathrm{C}$  (a) as a function of the  $\\mathrm{O}_2$  pulse length when the  $\\mathrm{NbCl}_5$  pulse is kept at 4 s and (b) as a function of the  $\\mathrm{NbCl}_5$  pulse length when the  $\\mathrm{O}_3$  pulse length is kept at 3 s.","id":"test/atomic-layer-etching/experimental-usecase/47/FIG4_b","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG4_b","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_b.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_b.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:image_caption"},"width":711,"height":581,"image_format":"jpeg","image_sha256":"6416c51e2720bca621e13fcd2813342d5a3e4450a3135c24a897942de0644f70","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":6,"width":687,"height":572}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch per cycle EPC  is plotted against the pulse length of NbCl₅. The EPC increases slightly with increasing pulse length.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| NbCl₅ Pulse Length (s) | EPC (Å) | Error Range (Å) |\\n|-------------------------|---------|-----------------|\\n| 0.5                     | 0.95    | ±0.15          |\\n| 1.0                     | 1.05    | ±0.15          |\\n| 1.5                     | 1.10    | ±0.15          |\\n| 2.0                     | 1.20    | ±0.15          |\\n| 4.0                     | 1.20    | ±0.15          |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"NbCl5 pulse length was then varied while keeping the O2 pulse constant.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"the NbCl5 pulse is saturated already at 2 s, and 80% of the saturated EPC is reached within the first 0.2 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The fast initial etching followed by gradual saturation is thought to be caused by etching of MoO3 first, followed by slower etching of the lower oxides and generation of the NbCl3/MoCl3 layer.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"4 s NbCl5 pulses were used in the following experiments to ensure complete reaction\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG5_a.jpg","caption":"FIG. 6. Thickness of film etched as a function of the number of ALE cycles at  $300^{\\circ}\\mathrm{C}$  using cycles of  $4\\mathrm{sNbCl}_5$  and  $3\\mathrm{sO}_2$  pulses with 1 s purges.","id":"test/atomic-layer-etching/experimental-usecase/47/FIG5_a","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG5_a","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG5_a.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG5_a.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:image_caption"},"width":669,"height":578,"image_format":"jpeg","image_sha256":"8af229986b3cba06c62af472a3b30854ba19c8f1c633494062f1336c4098d26b","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":3,"width":662,"height":573}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot relates the temperature with EPC (Å) with error bars indicating variability.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | EPC (Å) |\\n|---|---|\\n| 200 | 0 |\\n| 250 | 0.2 |\\n| 300 | 1.2 |\\n| 350 | 2.6 |\\n| 400 | 3.8 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The EPC was studied as a function of the etching temperature at 200–400 °C using 4 s NbCl5 pulses and 3 s O2 pulses with 1 s purges.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 200 °C, no etching is observed, but at 250– 400 °C, the EPC is seen to depend almost linearly on the temperature with these pulsing parameters\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The lack of etching at 200 °C may arise from the fact that at this temperature Mo is oxidized to only MoO2 which NbCl5 is not able to etch at this low temperature . While the oxidation of both Mo and MoO2 to MoO3 are thermodynamically favorable at this temperature, there could be a kinetic barrier to the reaction beyond MoO2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG6.jpg","caption":"","id":"test/atomic-layer-etching/experimental-usecase/47/FIG6","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG6","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG6.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"not_found"},"width":678,"height":578,"image_format":"jpeg","image_sha256":"90d4552731535a8648595ed14e44132e844b258a65a647cda9cd34df6fd1265f","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":3,"width":665,"height":572}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The images plots the number of ALE cycles and the thicknesses indicating a decreasing trend and etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| no. cycles | Thickness left (nm) |\\n|---|---|\\n| 0 | 52 |\\n| 200 | 41 |\\n| 400 | 25 |\\n| 600 | 12 |\\n| 800 | 2 |\\n| 1000 | 1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~55nm\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch depth as a function of the number of ALE cycles shows the etching increases in a linear fashion as the number of cycles and after 800 cycles the film has been completely etched.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Initial thickness = 55nm,\\nThickness after 200 ALE cycles = 41 nm \\nThickness etched away = 55-41 =14 nm\\nEtch rate = 14/200 = 0.07 nm/cycle\"}]}]"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG4_a.jpg","caption":"","id":"test/atomic-layer-etching/experimental-usecase/47/FIG4_a","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG4_a","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch per cycle EPC (in Å) is plotted against oxygen pulse length (in seconds). As the oxygen pulse length increases, there is a corresponding increase in EPC.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| O₂ pulse length (s) | EPC (Å) |\\n|---|---|\\n| 1 | 0.4 |\\n| 2 | 0.8 |\\n| 3 | 1.2 |\\n| 6 | 1.8 |\\n| 10 | 2.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The pulse is considered to saturate when increasing the pulse length no longer affects the EPC of the process.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The O2 pulse was studied first at 300 °C. The length of the NbCl5 pulse was kept at 4 s and the purge lengths at 1 s. It was found that the O2 pulse does not fully saturate even with 10 s pulse times, but the increase in EPC slows down gradually.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No, not O2 pulse.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the interest of optimizing the etch rate per time unit, 3 s was chosen as the O2 pulse length for further experiments.\"}]}]","bbox":[{"panel_id":"a","x":8,"y":5,"width":662,"height":566}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_a.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_a.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"not_found"},"width":675,"height":575,"image_format":"jpeg","image_sha256":"9b9b64f5f899d85f8a99d620f551e42929dd18301d98c8a3cce103cb104839c9","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG4_b.jpg","caption":"FIG. 4. EPC of molybdenum at  $300^{\\circ}\\mathrm{C}$  (a) as a function of the  $\\mathrm{O}_2$  pulse length when the  $\\mathrm{NbCl}_5$  pulse is kept at 4 s and (b) as a function of the  $\\mathrm{NbCl}_5$  pulse length when the  $\\mathrm{O}_3$  pulse length is kept at 3 s.","id":"test/atomic-layer-etching/experimental-usecase/47/FIG4_b","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG4_b","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch per cycle EPC  is plotted against the pulse length of NbCl₅. The EPC increases slightly with increasing pulse length.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| NbCl₅ Pulse Length (s) | EPC (Å) | Error Range (Å) |\\n|-------------------------|---------|-----------------|\\n| 0.5                     | 0.95    | ±0.15          |\\n| 1.0                     | 1.05    | ±0.15          |\\n| 1.5                     | 1.10    | ±0.15          |\\n| 2.0                     | 1.20    | ±0.15          |\\n| 4.0                     | 1.20    | ±0.15          |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"NbCl5 pulse length was then varied while keeping the O2 pulse constant.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"the NbCl5 pulse is saturated already at 2 s, and 80% of the saturated EPC is reached within the first 0.2 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The fast initial etching followed by gradual saturation is thought to be caused by etching of MoO3 first, followed by slower etching of the lower oxides and generation of the NbCl3/MoCl3 layer.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"4 s NbCl5 pulses were used in the following experiments to ensure complete reaction\"}]}]","bbox":[{"panel_id":"a","x":5,"y":6,"width":687,"height":572}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_b.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG4_b.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:image_caption"},"width":711,"height":581,"image_format":"jpeg","image_sha256":"6416c51e2720bca621e13fcd2813342d5a3e4450a3135c24a897942de0644f70","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG5_a.jpg","caption":"FIG. 6. Thickness of film etched as a function of the number of ALE cycles at  $300^{\\circ}\\mathrm{C}$  using cycles of  $4\\mathrm{sNbCl}_5$  and  $3\\mathrm{sO}_2$  pulses with 1 s purges.","id":"test/atomic-layer-etching/experimental-usecase/47/FIG5_a","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG5_a","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot relates the temperature with EPC (Å) with error bars indicating variability.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | EPC (Å) |\\n|---|---|\\n| 200 | 0 |\\n| 250 | 0.2 |\\n| 300 | 1.2 |\\n| 350 | 2.6 |\\n| 400 | 3.8 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The EPC was studied as a function of the etching temperature at 200–400 °C using 4 s NbCl5 pulses and 3 s O2 pulses with 1 s purges.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 200 °C, no etching is observed, but at 250– 400 °C, the EPC is seen to depend almost linearly on the temperature with these pulsing parameters\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The lack of etching at 200 °C may arise from the fact that at this temperature Mo is oxidized to only MoO2 which NbCl5 is not able to etch at this low temperature . While the oxidation of both Mo and MoO2 to MoO3 are thermodynamically favorable at this temperature, there could be a kinetic barrier to the reaction beyond MoO2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":662,"height":573}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG5_a.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG5_a.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:image_caption"},"width":669,"height":578,"image_format":"jpeg","image_sha256":"8af229986b3cba06c62af472a3b30854ba19c8f1c633494062f1336c4098d26b","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_47_FIG6.jpg","caption":"","id":"test/atomic-layer-etching/experimental-usecase/47/FIG6","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG6","subset":"box-plot","split":"test","classification":[{"panel_id":"a","label":"box plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The images plots the number of ALE cycles and the thicknesses indicating a decreasing trend and etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| no. cycles | Thickness left (nm) |\\n|---|---|\\n| 0 | 52 |\\n| 200 | 41 |\\n| 400 | 25 |\\n| 600 | 12 |\\n| 800 | 2 |\\n| 1000 | 1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~55nm\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch depth as a function of the number of ALE cycles shows the etching increases in a linear fashion as the number of cycles and after 800 cycles the film has been completely etched.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Initial thickness = 55nm,\\nThickness after 200 ALE cycles = 41 nm \\nThickness etched away = 55-41 =14 nm\\nEtch rate = 14/200 = 0.07 nm/cycle\"}]}]","bbox":[{"panel_id":"a","x":5,"y":3,"width":665,"height":572}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG6.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"not_found"},"width":678,"height":578,"image_format":"jpeg","image_sha256":"90d4552731535a8648595ed14e44132e844b258a65a647cda9cd34df6fd1265f","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/box-plot/train/metadata.jsonl

@@ -1,3 +1,3 @@
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_8_figure_5.jpg","caption":"Figure 5. Box plots for the precursor vessel temperatures for the prior introduced precursor classes. Boxes contain the average  $50\\%$  of the respective dataset while white dots represent the arithmetic mean. The first quartile is the median of the lower half of the dataset. The second quartile represents the median of the respective dataset, and the third quartile is the median of the upper half of the dataset. The whiskers are restricted to a  $1.5x$  inter quartile range representing entries outside the box. Outliners are depicted as black dots. If lower or upper whiskers are not displayed an overlap with the respective quartile occurred.","id":"train/atomic-layer-deposition/experimental-usecase/8/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/8/figure_5","subset":"box-plot","split":"train","classification":[{"panel_id":"a","label":"box plot"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/Adv Materials Inter - 2024 - Ghazy - Atomic and Molecular Layer Deposition of Functional Thin Films Based on Rare Earth.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:image_caption"},"width":667,"height":480,"image_format":"jpeg","image_sha256":"15b14ac81a9e579e000abbac7161f32af3dee582c1edf6775571688382f12387","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":4,"y":6,"width":659,"height":471}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The box plot compares precursor vessel temperatures for eight rare earth precursor classes used in ALD: R(amd)3, R(amd)(Cp)2, R(or')x, R(Cp)3, R(thd)x, R(famd)3, R(guan)3, and R(tmsa)3. Temperatures range from approximately 80 to 200 °C. White squares indicate arithmetic means, boxes span the interquartile range, and black diamonds mark outliers. R(or')x shows the highest mean temperature near 165 °C, while R(amd)(Cp)2 displays the lowest mean around 131 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| R(amd)\\\\_3 | R(amd)\\\\(Cp\\\\)_2 | R(or')\\\\_x | R(Cp)_3 | R(thd)_x | R(famd)_3 | R(guan)_3 | R(tmsa)_3 |\\n|---|---|---|---|---|---|---|---|\\n| 120 | 130 | 190 | 170 | 150 | 140 | 150 | 140 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The source temperature controls precursor vapor pressure and delivery rate to the reactor. If it is too low, delivery can be unstable and growth becomes mass transport limited. If it is too high, the precursor can decompose or produce parasitic CVD like contributions. Source temperature is therefore a practical indicator of volatility and thermal robustness for a given precursor family.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The alkoxide family, labeled R(or′)x, shows the highest overall range and high typical values compared with the other families.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Higher volatility at a given operating condition, meaning sufficient vapor pressure can be achieved with less heating.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lower thermal stress on the precursor and delivery hardware\\n\\nReduced risk of precursor decomposition in the source lines\\n\\nImproved process stability and repeatability through easier vapor delivery\"}]}]"}
-{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_29_fig_1.jpg","caption":"Fig. 1. Effect of precursor pulse time on  $\\mathrm{TiO_2}$  ALD growth rate. The vertical error bars indicate film uniformity across the sample (Selvaraj et al., 2013).","id":"train/atomic-layer-deposition/simulation-usecase/29/fig_1","sample_id":"atomic-layer-deposition/simulation-usecase/29/fig_1","subset":"box-plot","split":"train","classification":[{"panel_id":"a","label":"box plot"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:img_caption"},"width":658,"height":378,"image_format":"jpeg","image_sha256":"b563a6fa08a0dfb9e80a8ea043b58d2a87ba0b0e5f530e2fe08d2aeac096bb4f","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":5,"width":657,"height":372}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart plots the experimental growth rate in nanometers per cycle against the precursor pulsing time, illustrating the saturation of the ALD process.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulsing Time (s) | Growth Rate (nm/cycle) |\\n|---|---|\\n| 1 | 0.023 |\\n| 2 | 0.040 |\\n| 4 | 0.040 |\\n| 6 | 0.042 |\\n| 8 |  0.040 |\\n| 14 | 0.042 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After ~ 2s.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The optimal pulse time for the fastest manufacturing would be 2 seconds.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Based on the error bars, higher uniformity is achieved at higher pulsing times.\"}]}]"}
-{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_30_figure_3.jpg","caption":"Figure 3. (a) Adsorption and reaction energies of the bonding steps of TMPS as illustrated in panel c are given as box plots for both  $\\mathrm{a - SiO_2}$  surface slabs. The red line gives the median reaction energy  $(\\Delta E)$  together with its mean electronic (  $\\Delta E_{\\mathrm{elec}}$  green) and dispersion (  $\\Delta E_{\\mathrm{disp}}$  blue) contributions. The dashed red line marks the total reaction energy on the  $\\alpha$  -quartz model.14 (b) Plot showing the inter-silanol distances of the different surface models (I, vertical lines mark the average  $\\mathrm{Si - OH}$  distance of the respective slabs. (c) Reaction schemes illustrating the chemisorption process (I, first row), the first condensation reaction to the SB (II, second row), and the second condensation reaction to the DB bonding configuration (III, third row).","id":"train/atomic-layer-deposition/simulation-usecase/30/figure_3","sample_id":"atomic-layer-deposition/simulation-usecase/30/figure_3","subset":"box-plot","split":"train","classification":[{"panel_id":"a","label":"box plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"reaction scheme"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/Paul Philipp Wellmann et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"30","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:img_caption"},"width":1406,"height":653,"image_format":"jpeg","image_sha256":"e5a538e104869543b3fa0305be99fbe9f80bc6436d5296334907a09311e1b52d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":2,"width":878,"height":652},{"panel_id":"b","x":180,"y":11,"width":453,"height":306},{"panel_id":"c","x":943,"y":2,"width":458,"height":646}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_8_figure_5.jpg","caption":"Figure 5. Box plots for the precursor vessel temperatures for the prior introduced precursor classes. Boxes contain the average  $50\\%$  of the respective dataset while white dots represent the arithmetic mean. The first quartile is the median of the lower half of the dataset. The second quartile represents the median of the respective dataset, and the third quartile is the median of the upper half of the dataset. The whiskers are restricted to a  $1.5x$  inter quartile range representing entries outside the box. Outliners are depicted as black dots. If lower or upper whiskers are not displayed an overlap with the respective quartile occurred.","id":"train/atomic-layer-deposition/experimental-usecase/8/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/8/figure_5","subset":"box-plot","split":"train","classification":[{"panel_id":"a","label":"box plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The box plot compares precursor vessel temperatures for eight rare earth precursor classes used in ALD: R(amd)3, R(amd)(Cp)2, R(or')x, R(Cp)3, R(thd)x, R(famd)3, R(guan)3, and R(tmsa)3. Temperatures range from approximately 80 to 200 °C. White squares indicate arithmetic means, boxes span the interquartile range, and black diamonds mark outliers. R(or')x shows the highest mean temperature near 165 °C, while R(amd)(Cp)2 displays the lowest mean around 131 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| R(amd)\\\\_3 | R(amd)\\\\(Cp\\\\)_2 | R(or')\\\\_x | R(Cp)_3 | R(thd)_x | R(famd)_3 | R(guan)_3 | R(tmsa)_3 |\\n|---|---|---|---|---|---|---|---|\\n| 120 | 130 | 190 | 170 | 150 | 140 | 150 | 140 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The source temperature controls precursor vapor pressure and delivery rate to the reactor. If it is too low, delivery can be unstable and growth becomes mass transport limited. If it is too high, the precursor can decompose or produce parasitic CVD like contributions. Source temperature is therefore a practical indicator of volatility and thermal robustness for a given precursor family.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The alkoxide family, labeled R(or′)x, shows the highest overall range and high typical values compared with the other families.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Higher volatility at a given operating condition, meaning sufficient vapor pressure can be achieved with less heating.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lower thermal stress on the precursor and delivery hardware\\n\\nReduced risk of precursor decomposition in the source lines\\n\\nImproved process stability and repeatability through easier vapor delivery\"}]}]","bbox":[{"panel_id":"a","x":4,"y":6,"width":659,"height":471}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/Adv Materials Inter - 2024 - Ghazy - Atomic and Molecular Layer Deposition of Functional Thin Films Based on Rare Earth.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:image_caption"},"width":667,"height":480,"image_format":"jpeg","image_sha256":"15b14ac81a9e579e000abbac7161f32af3dee582c1edf6775571688382f12387","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_29_fig_1.jpg","caption":"Fig. 1. Effect of precursor pulse time on  $\\mathrm{TiO_2}$  ALD growth rate. The vertical error bars indicate film uniformity across the sample (Selvaraj et al., 2013).","id":"train/atomic-layer-deposition/simulation-usecase/29/fig_1","sample_id":"atomic-layer-deposition/simulation-usecase/29/fig_1","subset":"box-plot","split":"train","classification":[{"panel_id":"a","label":"box plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart plots the experimental growth rate in nanometers per cycle against the precursor pulsing time, illustrating the saturation of the ALD process.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulsing Time (s) | Growth Rate (nm/cycle) |\\n|---|---|\\n| 1 | 0.023 |\\n| 2 | 0.040 |\\n| 4 | 0.040 |\\n| 6 | 0.042 |\\n| 8 |  0.040 |\\n| 14 | 0.042 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After ~ 2s.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The optimal pulse time for the fastest manufacturing would be 2 seconds.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Based on the error bars, higher uniformity is achieved at higher pulsing times.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":657,"height":372}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:img_caption"},"width":658,"height":378,"image_format":"jpeg","image_sha256":"b563a6fa08a0dfb9e80a8ea043b58d2a87ba0b0e5f530e2fe08d2aeac096bb4f","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_30_figure_3.jpg","caption":"Figure 3. (a) Adsorption and reaction energies of the bonding steps of TMPS as illustrated in panel c are given as box plots for both  $\\mathrm{a - SiO_2}$  surface slabs. The red line gives the median reaction energy  $(\\Delta E)$  together with its mean electronic (  $\\Delta E_{\\mathrm{elec}}$  green) and dispersion (  $\\Delta E_{\\mathrm{disp}}$  blue) contributions. The dashed red line marks the total reaction energy on the  $\\alpha$  -quartz model.14 (b) Plot showing the inter-silanol distances of the different surface models (I, vertical lines mark the average  $\\mathrm{Si - OH}$  distance of the respective slabs. (c) Reaction schemes illustrating the chemisorption process (I, first row), the first condensation reaction to the SB (II, second row), and the second condensation reaction to the DB bonding configuration (III, third row).","id":"train/atomic-layer-deposition/simulation-usecase/30/figure_3","sample_id":"atomic-layer-deposition/simulation-usecase/30/figure_3","subset":"box-plot","split":"train","classification":[{"panel_id":"a","label":"box plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"reaction scheme"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":5,"y":2,"width":878,"height":652},{"panel_id":"b","x":180,"y":11,"width":453,"height":306},{"panel_id":"c","x":943,"y":2,"width":458,"height":646}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/30/Paul Philipp Wellmann et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"30","first_classification_panel_id":"a","first_classification_label":"box plot","caption_source":"content.json:img_caption"},"width":1406,"height":653,"image_format":"jpeg","image_sha256":"e5a538e104869543b3fa0305be99fbe9f80bc6436d5296334907a09311e1b52d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/chromaticity-diagram/train/metadata.jsonl

@@ -1 +1 @@
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_8_figure_13.jpg","caption":"Figure 13. Emission color comparison: ALD/MLD grown (Tb, Eu, Er)-TPA thin film versus three different commercial white light emitting products: LED lamp, fluorescent lamp, and incandescent bulb.[320]","id":"train/atomic-layer-deposition/experimental-usecase/8/figure_13","sample_id":"atomic-layer-deposition/experimental-usecase/8/figure_13","subset":"chromaticity-diagram","split":"train","classification":[{"panel_id":"a","label":"chromaticity diagram"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/Adv Materials Inter - 2024 - Ghazy - Atomic and Molecular Layer Deposition of Functional Thin Films Based on Rare Earth.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"chromaticity diagram","caption_source":"content.json:image_caption"},"width":658,"height":542,"image_format":"jpeg","image_sha256":"2a0afeee989dc01869130d4f06c79a70859fcdee8703f461e7de838dd218f847","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":2,"width":655,"height":539}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"CIE 1931 x y chromaticity diagram with a point for an ALD/MLD thin film emission compared against reference lamp chromaticities (LED, fluorescent, incandescent).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample / Light Source | CIE x | CIE y |\\n|---|---|---|\\n| LED lamp (Blue Dot) | 0.40 | 0.39 |\\n| ALD/MLD thin film (Red Dot) | 0.45 | 0.45 |\\n| Incandescent bulb (Green Dot) | 0.48 | 0.42 |\\n| Fluorescent lamp (Black Dot) | 0.50 | 0.46 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent chromaticity coordinates in the CIE 1931 x y color space, describing perceived color independent of intensity.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is closer to the incandescent bulb point.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The emission color can be shifted by changing the emitting center and its local environment, for example adjusting the rare earth activator identity or concentration, modifying the organic linker or host composition in ALD/MLD, or tuning the film stack that controls excitation and energy transfer. These changes alter the emission spectrum, which moves the x y coordinate.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It benchmarks whether the film emission matches practical lighting color points\\n\\nIt helps judge suitability for display, lighting, or sensing where specific color is required\\n\\nIt provides a compact way to compare color output across materials and processing conditions\"}]}]"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_8_figure_13.jpg","caption":"Figure 13. Emission color comparison: ALD/MLD grown (Tb, Eu, Er)-TPA thin film versus three different commercial white light emitting products: LED lamp, fluorescent lamp, and incandescent bulb.[320]","id":"train/atomic-layer-deposition/experimental-usecase/8/figure_13","sample_id":"atomic-layer-deposition/experimental-usecase/8/figure_13","subset":"chromaticity-diagram","split":"train","classification":[{"panel_id":"a","label":"chromaticity diagram"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"CIE 1931 x y chromaticity diagram with a point for an ALD/MLD thin film emission compared against reference lamp chromaticities (LED, fluorescent, incandescent).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample / Light Source | CIE x | CIE y |\\n|---|---|---|\\n| LED lamp (Blue Dot) | 0.40 | 0.39 |\\n| ALD/MLD thin film (Red Dot) | 0.45 | 0.45 |\\n| Incandescent bulb (Green Dot) | 0.48 | 0.42 |\\n| Fluorescent lamp (Black Dot) | 0.50 | 0.46 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent chromaticity coordinates in the CIE 1931 x y color space, describing perceived color independent of intensity.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is closer to the incandescent bulb point.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The emission color can be shifted by changing the emitting center and its local environment, for example adjusting the rare earth activator identity or concentration, modifying the organic linker or host composition in ALD/MLD, or tuning the film stack that controls excitation and energy transfer. These changes alter the emission spectrum, which moves the x y coordinate.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It benchmarks whether the film emission matches practical lighting color points\\n\\nIt helps judge suitability for display, lighting, or sensing where specific color is required\\n\\nIt provides a compact way to compare color output across materials and processing conditions\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":655,"height":539}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/images/figure_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/8/Adv Materials Inter - 2024 - Ghazy - Atomic and Molecular Layer Deposition of Functional Thin Films Based on Rare Earth.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"chromaticity diagram","caption_source":"content.json:image_caption"},"width":658,"height":542,"image_format":"jpeg","image_sha256":"2a0afeee989dc01869130d4f06c79a70859fcdee8703f461e7de838dd218f847","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/contour-heatmap/test/metadata.jsonl

@@ -1,3 +1,3 @@
-{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_20_fig_10.jpg","caption":"Fig. 10. Slice contours of transient flow and deposition process in the in-line spatial ALD reactor at 0.005, 0.015, 0.025, and 0.035 s for  $1\\mathrm{mm}$  gap size, 0.22 torr pumping pressure, and  $250^{\\circ}C$  wafer temperature.","id":"test/atomic-layer-deposition/simulation-usecase/20/fig_10","sample_id":"atomic-layer-deposition/simulation-usecase/20/fig_10","subset":"contour-heatmap","split":"test","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/images/fig_10.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/Dongqing Pan et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1264,"height":1219,"image_format":"jpeg","image_sha256":"18781c891dfd08ed668d0d70be9bff3f992c0f094d2667bef2b8656bf75f1e55","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":10,"y":7,"width":619,"height":608},{"panel_id":"b","x":638,"y":4,"width":619,"height":608},{"panel_id":"c","x":1,"y":589,"width":625,"height":634},{"panel_id":"d","x":635,"y":612,"width":622,"height":602}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Simulation snapshot at 0.005 s showing early stage of precursor dosing, with water and TMA beginning to fill their respective cylinders, minimal methane generation, and initial consumption of surface *OH sites under the water cylinder.\"},{\"panel_id\":\"b\",\"text\":\"Simulation snapshot at 0.015 s (critical dosing time) showing fully developed precursor distributions in the dosing cylinders, significant methane byproduct generation on the wafer surface, and near-complete consumption of reactive surface sites under both precursor zones.\"},{\"panel_id\":\"c\",\"text\":\"Simulation snapshot at 0.025 s showing steady-state precursor concentration profiles, saturated methane distribution, and fully reacted surface sites (*AlMe/*Al(Me)2 under TMA, *OH under water).\"},{\"panel_id\":\"d\",\"text\":\"Simulation snapshot at 0.035 s confirming steady-state conditions with no significant changes from (c), demonstrating the self-limiting nature of the ALD surface reactions beyond the critical dosing time.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"},{\"panel_id\":\"b\",\"text\":\"\"},{\"panel_id\":\"c\",\"text\":\"\"},{\"panel_id\":\"d\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 0.005 s, methane concentration is minimal since surface reactions have just begun. By 0.015 s, significant methane appears on the wafer surface beneath both precursor zones, indicating active half-reactions. From 0.025 s to 0.035 s, methane distribution reaches steady state with high concentrations near the surface. This evolution reflects the progression of ALD surface chemistry, where methane is released as a byproduct when TMA reacts with *OH groups and water reacts with methyl-terminated sites.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The nearly identical concentration profiles and surface coverage distributions between 0.025 s and 0.035 s confirm that saturation is achieved, consistent with the paper's finding that critical dosing time is approximately 0.015 s.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent the fraction of reactive surface sites: *OH indicates hydroxyl-terminated sites available for TMA reaction, while *AlMe/*Al(Me)2 indicates methyl-terminated sites (from TMA chemisorption) available for water reaction.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They are the separating gas (nitrogen) injection cylinder in the center and two exhaust/pumping cylinders that remove reaction byproducts and prevent precursor intermixing.\"}]}]"}
-{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_40_fig4.jpg","caption":"Fig.4. a) DEZ 2D concentration profile at its injection channel for different values of substrate velocity, b) the corresponding DEZ concentration profiles projected at the substrate surface level. In this case, the substrate temperature and the gap between the substrate and the injection head were fixed at  $200^{\\circ}C$  and  $180~\\mu \\mathrm{m}$  respectively, c) schematic illustration of the variation of [DEZ] as a function of position for a substrate velocity of  $8cm / s$  as implemented in the proposed model. The data shown here were simulated with Comsol Multiphysics.","id":"test/atomic-layer-deposition/simulation-usecase/40/fig4","sample_id":"atomic-layer-deposition/simulation-usecase/40/fig4","subset":"contour-heatmap","split":"test","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"multi spectra chart"},{"panel_id":"c","label":"multi spectra chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/images/fig4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/Viet Huong Nguyen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1258,"height":855,"image_format":"jpeg","image_sha256":"2c96174a7149e6ee5b5dd8ba94f7e10b27b94b6512cae65db00d9a9651f7a762","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":1258,"height":269},{"panel_id":"b","x":11,"y":312,"width":572,"height":543},{"panel_id":"c","x":639,"y":312,"width":577,"height":505}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This panel shows simulated spatial distributions of DEZ precursor concentration near the substrate for different substrate velocities, illustrating how increasing velocity reduces precursor residence time and alters the overlap between DEZ injection, exhaust, and N₂ carrier gas regions.\"},{\"panel_id\":\"b\",\"text\":\"This panel presents the DEZ concentration at the substrate surface as a function of lateral position for different substrate velocities, demonstrating that higher velocities broaden and lower the effective precursor exposure window, leading to reduced peak surface concentration and shorter saturation regions.\"},{\"panel_id\":\"c\",\"text\":\"This panel schematically defines discrete exposure zones along the substrate motion direction, showing how the total DEZ exposure can be decomposed into successive incremental contributions as the substrate passes under the injector, providing a basis for modeling growth per cycle in spatial ALD.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gas injector | N₂ | Exhaust | DEZ | Exhaust | N₂ |\\n|---|---|---|---|---|---|\\n| 0 cm/s | | | | | |\\n| 4 cm/s | | | | | |\\n| 12 cm/s | | | | | |\\n| 16 cm/s | | | | | |\"},{\"panel_id\":\"b\",\"text\":\"| Substrate velocity (cm/s) | [DEZ] at substrate surface (mol/m³) |\\n|---|---|\\n| 0 | 1.0 |\\n| 4 | 0.95 |\\n| 8 | 0.9 |\\n| 12 | 0.85 |\\n| 16 | 0.8 |\"},{\"panel_id\":\"c\",\"text\":\"| Substrate velocity (cm/s) | [DEZ] at substrate surface (mol/m³) |\\n|---|---|\\n| 8 | 1.0 |\\n| 8 | 0.9 |\\n| 8 | 0.8 |\\n| 8 | 0.7 |\\n| 8 | 0.6 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Substrate velocity directly determines precursor residence time and surface exposure uniformity in spatial ALD. Excessively high velocities limit precursor adsorption and prevent surface saturation, while low velocities may lead to overdosing and parasitic reactions. Optimizing substrate speed ensures self-limited surface chemistry, uniform film thickness, and stable high-throughput deposition.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Substrate velocity, Injector–substrate gap distance, Exhaust channel spacing, Precursor injection geometry\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Increasing substrate velocity narrows the DEZ exposure window and reduces peak surface concentration.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig_9.jpg","caption":"Fig. 9 - Comparison of the projected half-cycle time zones at differing operating conditions for temperature and pressure of the reagents. For Step A, a time contour plot (a) and an operational feasibility plot (b) with a feasible time range of  $1.5\\mathrm{s}$  are shown. For Step B, a time contour plot (c) and an operational feasibility plot (d) with a feasible time range of  $2.5\\mathrm{s}$  are shown.","id":"test/atomic-layer-etching/simulation-usecase/11/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/11/fig_9","subset":"contour-heatmap","split":"test","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"heatmap"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1014,"height":800,"image_format":"jpeg","image_sha256":"0e563c66765fb1786ea19c7de6960cf85c394b07cde0ba8203aa02bb82450dfc","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":1,"y":4,"width":465,"height":384},{"panel_id":"b","x":538,"y":8,"width":475,"height":392},{"panel_id":"c","x":13,"y":422,"width":450,"height":374},{"panel_id":"d","x":536,"y":422,"width":472,"height":377}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Contour plot of FNN-predicted half-cycle times for Step A across the operating window (533-598 K, 0-100 Pa). The horizontal color bands confirm that Step A is essentially pressure-independent, with times ranging from approximately 21 s at low temperatures to under 3 s at high temperatures.\"},{\"panel_id\":\"b\",\"text\":\"Operational feasibility map for Step A using a 1.5 s threshold. The blue region (operable) requires temperatures above approximately 570 K, while the red region (inoperable) indicates conditions where half-cycle times exceed industrial requirements. The boundary is horizontal, confirming pressure independence.\"},{\"panel_id\":\"c\",\"text\":\"Contour plot for Step B showing strong dependence on both temperature and pressure. Half-cycle times range from over 35 s at low temperature/pressure to under 5 s at high temperature and pressure, with curved contour lines indicating coupled effects.\"},{\"panel_id\":\"d\",\"text\":\"Operational feasibility map for Step B using a 2.5 s threshold. The curved boundary shows that achieving operable conditions requires both high temperature (above ~573 K) and sufficient pressure (above ~40-60 Pa depending on temperature).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Time (s) [Approximate, Independent of Pressure] |\\n|---|---|\\n| 525 | 20 |\\n| 530 | 15 |\\n| 540 | 9 |\\n| 550 | 6 |\\n| 560 | 3 |\\n| 570 | 1.5 |\\n| > 575 | < 1 |\"},{\"panel_id\":\"b\",\"text\":\"| Pressure (Pa) | Boundary Temperature (K) | Status Below Boundary | Status Above Boundary |\\n|---|---|---|---|\\n| 0 - 100 | 570 | Inoperable | Operable |\"},{\"panel_id\":\"c\",\"text\":\"| Temperature (K) | Time at P=10 Pa (s) | Time at P=50 Pa (s) | Time at P=90 Pa (s) |\\n|---|---|---|---|\\n| 525 | 38 | 36 | 35 |\\n| 530 | 32 | 30 | 28 |\\n| 540 | 22 | 18 | 16 |\\n| 550 | 15 | 10 | 8 |\\n| 560 | 10 | 6 | 5 |\\n| 570 | 5 | 3 | 2.5 |\\n| 580 | 3 | 1 | < 1 |\"},{\"panel_id\":\"d\",\"text\":\"| Pressure (Pa) | Transition Temperature (K) [Boundary between Inoperable/Operable] |\\n|---|---|\\n| 20 | ~595 |\\n| 25 | 590 |\\n| 30 | 582 |\\n| 40 | 576 |\\n| 50 | 574 |\\n| 60 | 573 |\\n| 80 | 572 |\\n| 100 | 572 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"These thresholds ensure that the microscopic half-cycle times, when combined with macroscopic gas transport times of 2-3 s, yield total process times under the industrial target of approximately 5 s per half-cycle. The thresholds are based on experimental observations showing self-limiting behavior is achieved within these times at the validation temperature of 573 K.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Step A shows virtually no pressure dependence, as evidenced by the horizontal contour lines in (a) and the horizontal feasibility boundary in (b). Only temperature affects the Step A half-cycle time. In contrast, Step B depends on both temperature and pressure, shown by the curved contours in (c) and the curved feasibility boundary in (d). This means Step B optimization requires careful selection of both parameters, while Step A can be optimized by temperature alone.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step A requires temperatures above approximately 570 K to achieve half-cycle times below the 1.5 s feasibility threshold.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Blue indicates operable conditions where half-cycle times meet industrial requirements, while red indicates inoperable conditions where times are too long for practical use.\"}]}]"}
+{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_20_fig_10.jpg","caption":"Fig. 10. Slice contours of transient flow and deposition process in the in-line spatial ALD reactor at 0.005, 0.015, 0.025, and 0.035 s for  $1\\mathrm{mm}$  gap size, 0.22 torr pumping pressure, and  $250^{\\circ}C$  wafer temperature.","id":"test/atomic-layer-deposition/simulation-usecase/20/fig_10","sample_id":"atomic-layer-deposition/simulation-usecase/20/fig_10","subset":"contour-heatmap","split":"test","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Simulation snapshot at 0.005 s showing early stage of precursor dosing, with water and TMA beginning to fill their respective cylinders, minimal methane generation, and initial consumption of surface *OH sites under the water cylinder.\"},{\"panel_id\":\"b\",\"text\":\"Simulation snapshot at 0.015 s (critical dosing time) showing fully developed precursor distributions in the dosing cylinders, significant methane byproduct generation on the wafer surface, and near-complete consumption of reactive surface sites under both precursor zones.\"},{\"panel_id\":\"c\",\"text\":\"Simulation snapshot at 0.025 s showing steady-state precursor concentration profiles, saturated methane distribution, and fully reacted surface sites (*AlMe/*Al(Me)2 under TMA, *OH under water).\"},{\"panel_id\":\"d\",\"text\":\"Simulation snapshot at 0.035 s confirming steady-state conditions with no significant changes from (c), demonstrating the self-limiting nature of the ALD surface reactions beyond the critical dosing time.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"},{\"panel_id\":\"b\",\"text\":\"\"},{\"panel_id\":\"c\",\"text\":\"\"},{\"panel_id\":\"d\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 0.005 s, methane concentration is minimal since surface reactions have just begun. By 0.015 s, significant methane appears on the wafer surface beneath both precursor zones, indicating active half-reactions. From 0.025 s to 0.035 s, methane distribution reaches steady state with high concentrations near the surface. This evolution reflects the progression of ALD surface chemistry, where methane is released as a byproduct when TMA reacts with *OH groups and water reacts with methyl-terminated sites.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The nearly identical concentration profiles and surface coverage distributions between 0.025 s and 0.035 s confirm that saturation is achieved, consistent with the paper's finding that critical dosing time is approximately 0.015 s.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent the fraction of reactive surface sites: *OH indicates hydroxyl-terminated sites available for TMA reaction, while *AlMe/*Al(Me)2 indicates methyl-terminated sites (from TMA chemisorption) available for water reaction.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They are the separating gas (nitrogen) injection cylinder in the center and two exhaust/pumping cylinders that remove reaction byproducts and prevent precursor intermixing.\"}]}]","bbox":[{"panel_id":"a","x":10,"y":7,"width":619,"height":608},{"panel_id":"b","x":638,"y":4,"width":619,"height":608},{"panel_id":"c","x":1,"y":589,"width":625,"height":634},{"panel_id":"d","x":635,"y":612,"width":622,"height":602}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/images/fig_10.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/20/Dongqing Pan et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1264,"height":1219,"image_format":"jpeg","image_sha256":"18781c891dfd08ed668d0d70be9bff3f992c0f094d2667bef2b8656bf75f1e55","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_40_fig4.jpg","caption":"Fig.4. a) DEZ 2D concentration profile at its injection channel for different values of substrate velocity, b) the corresponding DEZ concentration profiles projected at the substrate surface level. In this case, the substrate temperature and the gap between the substrate and the injection head were fixed at  $200^{\\circ}C$  and  $180~\\mu \\mathrm{m}$  respectively, c) schematic illustration of the variation of [DEZ] as a function of position for a substrate velocity of  $8cm / s$  as implemented in the proposed model. The data shown here were simulated with Comsol Multiphysics.","id":"test/atomic-layer-deposition/simulation-usecase/40/fig4","sample_id":"atomic-layer-deposition/simulation-usecase/40/fig4","subset":"contour-heatmap","split":"test","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"multi spectra chart"},{"panel_id":"c","label":"multi spectra chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This panel shows simulated spatial distributions of DEZ precursor concentration near the substrate for different substrate velocities, illustrating how increasing velocity reduces precursor residence time and alters the overlap between DEZ injection, exhaust, and N₂ carrier gas regions.\"},{\"panel_id\":\"b\",\"text\":\"This panel presents the DEZ concentration at the substrate surface as a function of lateral position for different substrate velocities, demonstrating that higher velocities broaden and lower the effective precursor exposure window, leading to reduced peak surface concentration and shorter saturation regions.\"},{\"panel_id\":\"c\",\"text\":\"This panel schematically defines discrete exposure zones along the substrate motion direction, showing how the total DEZ exposure can be decomposed into successive incremental contributions as the substrate passes under the injector, providing a basis for modeling growth per cycle in spatial ALD.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gas injector | N₂ | Exhaust | DEZ | Exhaust | N₂ |\\n|---|---|---|---|---|---|\\n| 0 cm/s | | | | | |\\n| 4 cm/s | | | | | |\\n| 12 cm/s | | | | | |\\n| 16 cm/s | | | | | |\"},{\"panel_id\":\"b\",\"text\":\"| Substrate velocity (cm/s) | [DEZ] at substrate surface (mol/m³) |\\n|---|---|\\n| 0 | 1.0 |\\n| 4 | 0.95 |\\n| 8 | 0.9 |\\n| 12 | 0.85 |\\n| 16 | 0.8 |\"},{\"panel_id\":\"c\",\"text\":\"| Substrate velocity (cm/s) | [DEZ] at substrate surface (mol/m³) |\\n|---|---|\\n| 8 | 1.0 |\\n| 8 | 0.9 |\\n| 8 | 0.8 |\\n| 8 | 0.7 |\\n| 8 | 0.6 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Substrate velocity directly determines precursor residence time and surface exposure uniformity in spatial ALD. Excessively high velocities limit precursor adsorption and prevent surface saturation, while low velocities may lead to overdosing and parasitic reactions. Optimizing substrate speed ensures self-limited surface chemistry, uniform film thickness, and stable high-throughput deposition.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Substrate velocity, Injector–substrate gap distance, Exhaust channel spacing, Precursor injection geometry\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Increasing substrate velocity narrows the DEZ exposure window and reduces peak surface concentration.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":1258,"height":269},{"panel_id":"b","x":11,"y":312,"width":572,"height":543},{"panel_id":"c","x":639,"y":312,"width":577,"height":505}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/images/fig4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/Viet Huong Nguyen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1258,"height":855,"image_format":"jpeg","image_sha256":"2c96174a7149e6ee5b5dd8ba94f7e10b27b94b6512cae65db00d9a9651f7a762","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig_9.jpg","caption":"Fig. 9 - Comparison of the projected half-cycle time zones at differing operating conditions for temperature and pressure of the reagents. For Step A, a time contour plot (a) and an operational feasibility plot (b) with a feasible time range of  $1.5\\mathrm{s}$  are shown. For Step B, a time contour plot (c) and an operational feasibility plot (d) with a feasible time range of  $2.5\\mathrm{s}$  are shown.","id":"test/atomic-layer-etching/simulation-usecase/11/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/11/fig_9","subset":"contour-heatmap","split":"test","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Contour plot of FNN-predicted half-cycle times for Step A across the operating window (533-598 K, 0-100 Pa). The horizontal color bands confirm that Step A is essentially pressure-independent, with times ranging from approximately 21 s at low temperatures to under 3 s at high temperatures.\"},{\"panel_id\":\"b\",\"text\":\"Operational feasibility map for Step A using a 1.5 s threshold. The blue region (operable) requires temperatures above approximately 570 K, while the red region (inoperable) indicates conditions where half-cycle times exceed industrial requirements. The boundary is horizontal, confirming pressure independence.\"},{\"panel_id\":\"c\",\"text\":\"Contour plot for Step B showing strong dependence on both temperature and pressure. Half-cycle times range from over 35 s at low temperature/pressure to under 5 s at high temperature and pressure, with curved contour lines indicating coupled effects.\"},{\"panel_id\":\"d\",\"text\":\"Operational feasibility map for Step B using a 2.5 s threshold. The curved boundary shows that achieving operable conditions requires both high temperature (above ~573 K) and sufficient pressure (above ~40-60 Pa depending on temperature).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Time (s) [Approximate, Independent of Pressure] |\\n|---|---|\\n| 525 | 20 |\\n| 530 | 15 |\\n| 540 | 9 |\\n| 550 | 6 |\\n| 560 | 3 |\\n| 570 | 1.5 |\\n| > 575 | < 1 |\"},{\"panel_id\":\"b\",\"text\":\"| Pressure (Pa) | Boundary Temperature (K) | Status Below Boundary | Status Above Boundary |\\n|---|---|---|---|\\n| 0 - 100 | 570 | Inoperable | Operable |\"},{\"panel_id\":\"c\",\"text\":\"| Temperature (K) | Time at P=10 Pa (s) | Time at P=50 Pa (s) | Time at P=90 Pa (s) |\\n|---|---|---|---|\\n| 525 | 38 | 36 | 35 |\\n| 530 | 32 | 30 | 28 |\\n| 540 | 22 | 18 | 16 |\\n| 550 | 15 | 10 | 8 |\\n| 560 | 10 | 6 | 5 |\\n| 570 | 5 | 3 | 2.5 |\\n| 580 | 3 | 1 | < 1 |\"},{\"panel_id\":\"d\",\"text\":\"| Pressure (Pa) | Transition Temperature (K) [Boundary between Inoperable/Operable] |\\n|---|---|\\n| 20 | ~595 |\\n| 25 | 590 |\\n| 30 | 582 |\\n| 40 | 576 |\\n| 50 | 574 |\\n| 60 | 573 |\\n| 80 | 572 |\\n| 100 | 572 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"These thresholds ensure that the microscopic half-cycle times, when combined with macroscopic gas transport times of 2-3 s, yield total process times under the industrial target of approximately 5 s per half-cycle. The thresholds are based on experimental observations showing self-limiting behavior is achieved within these times at the validation temperature of 573 K.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Step A shows virtually no pressure dependence, as evidenced by the horizontal contour lines in (a) and the horizontal feasibility boundary in (b). Only temperature affects the Step A half-cycle time. In contrast, Step B depends on both temperature and pressure, shown by the curved contours in (c) and the curved feasibility boundary in (d). This means Step B optimization requires careful selection of both parameters, while Step A can be optimized by temperature alone.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step A requires temperatures above approximately 570 K to achieve half-cycle times below the 1.5 s feasibility threshold.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Blue indicates operable conditions where half-cycle times meet industrial requirements, while red indicates inoperable conditions where times are too long for practical use.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":465,"height":384},{"panel_id":"b","x":538,"y":8,"width":475,"height":392},{"panel_id":"c","x":13,"y":422,"width":450,"height":374},{"panel_id":"d","x":536,"y":422,"width":472,"height":377}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1014,"height":800,"image_format":"jpeg","image_sha256":"0e563c66765fb1786ea19c7de6960cf85c394b07cde0ba8203aa02bb82450dfc","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/contour-heatmap/train/metadata.jsonl

@@ -1,11 +1,11 @@
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_66_figure_2.jpg","caption":"Figure 2. Maps of the coercivity variations for (a)  $\\mathrm{ZnO}$  (oxygen deficient) and (b)  $\\mathrm{ZnO / Fe}$  ALD layers.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_2","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/Paskaleva et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"66","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:image_caption"},"width":1289,"height":555,"image_format":"jpeg","image_sha256":"5559422d2bf1c02ffe00f342c3a80f2f2ab6f615dfe743bc4d1cdb09810de632","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":7,"width":641,"height":543},{"panel_id":"b","x":633,"y":2,"width":650,"height":548}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A contour heatmap showing the magnetic coercivity (Hc) of ZnO (oxygen deficient) particles\"},{\"panel_id\":\"b\",\"text\":\"A contour heatmap showing the magnetic coercivity (Hc) of ZnO:Fe ALD layers\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| X (µm) Range | Y (µm) Range | H_c (Oe) |\\n|---|---|---|\\n| 0–100   | 0–100   | 0–500 |\\n| 100–200 | 0–100   | 500–1500 |\\n| 200–300 | 0–100   | 1500–2500 |\\n| 300–400 | 0–100   | 2500–3500 |\\n| 0–100   | 100–200 | 0–500 |\\n| 100–200 | 100–200 | 500–1500 |\\n| 200–300 | 100–200 | 1500–2500 |\\n| 300–400 | 100–200 | 2500–3500 |\\n| 0–100   | 200–300 | 0–500 |\\n| 100–200 | 200–300 | 500–1500 |\\n| 200–300 | 200–300 | 1500–2500 |\\n| 300–400 | 200–300 | 2500–3500 |\\n| 0–100   | 300–350 | 0–500 |\\n| 100–200 | 300–350 | 500–1500 |\\n| 200–300 | 300–350 | 1500–2500 |\\n| 300–400 | 300–350 | 1500–2500 |\"},{\"panel_id\":\"b\",\"text\":\"| X (µm) Range | Y (µm) Range | H_c (Oe) |\\n|---|---|---|\\n| 0–60   | 0–60   | 0–4000 |\\n| 60–120 | 0–60   | 0–4000 |\\n| 120–180| 0–60   | 0–4000 |\\n| 0–60   | 60–120 | 0–4000 |\\n| 60–120 | 60–120 | 0–4000 |\\n| 120–180| 60–120 | 0–4000 |\\n| 0–60   | 120–180| 0–4000 |\\n| 60–120 | 120–180| 0–4000 |\\n| 120–180| 60–120 | 0–4000 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Coercivity map of the undoped ZnO (oxygen deficient) is uniform\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The observed locations with very high Hc are stochastically distributed and can be assigned to the measurement artefacts.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Hc magnitude, the level of inhomogeneity and the overall magnetization is significantly stronger for the Fe-doped sample\"}]}]"}
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_66_figure_3.jpg","caption":"Figure 3. Maps of the Kerr angle variations for (a)  $\\mathrm{ZnO}$  (oxygen deficient) and (b)  $\\mathrm{ZnO / Fe}$  ALD layers.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_3","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/Paskaleva et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"66","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:image_caption"},"width":1292,"height":556,"image_format":"jpeg","image_sha256":"ed0ddf14ba0921433574f86bea53b6a1a3faaeab914c8878f4e151162b58ed93","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":4,"y":3,"width":652,"height":550},{"panel_id":"b","x":659,"y":1,"width":626,"height":553}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows a contour heatmap of Kerr rotation angle (θKerr) in millidegrees for oxygen deficient ZnO.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows a contour heatmap of Kerr rotation angle (θKerr) in millidegrees for ZnO:Fe samples.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| X (µm) | Y (µm) | θ_Kerr (mdeg) |\\n|---|---|---|\\n| 0–50   | 0–50   | 0–40 |\\n| 50–100 | 0–50   | 40–80 |\\n| 100–150| 0–50   | 80–120 |\\n| 150–200| 0–50   | 120–160 |\\n| 200–250| 0–50   | 160–200 |\\n| 250–300| 0–50   | 200–220 |\\n| 300–350| 0–50   | 160–200 |\\n| 350–400| 0–50   | 160–200 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Undoped ZnO layers are homogeneous, and the values of θKerr are very small, most of the data are in the range [0−20]\\nmdeg.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ZnO/Fe layer demonstrates inhomogeneous distribution of θKerr whose values are significantly (2 orders of magnitude) larger than those of the undoped layer.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"An interesting feature is that both clockwise (positive θKerr) and counterclockwise (negative θKerr) Kerr polarization rotations are observed in the Kerr angle map of Fe-doped ZnO, indicative of complex MO process activity\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_5.jpg","caption":"Fig. 5. Contours of Reynolds Number for Step A of various reactor configurations at 0.025 s in the standard condition in Table 3.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_5","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1011,"height":278,"image_format":"jpeg","image_sha256":"6a9d8377d5db93a5ec1f41102fe5ac8f55326e730473c845e131b3f26182b9cf","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":14,"y":41,"width":360,"height":236},{"panel_id":"b","x":394,"y":41,"width":192,"height":237},{"panel_id":"c","x":605,"y":41,"width":191,"height":234},{"panel_id":"d","x":815,"y":41,"width":196,"height":233}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G0. The color gradient indicates varying values of Re, with darker shades representing higher values.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G1.\"},{\"panel_id\":\"c\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G2.\"},{\"panel_id\":\"d\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G3.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Re Range | Peak Re Region |\\n|------|-------------------|----------------|\\n| G0   | 0.00 – 0.52      | 1.04 – 1.30    |\\n| G1   | 0.00 – 0.52      | 1.04 – 1.30    |\\n| G2   | 0.00 – 0.52      | 1.04 – 1.30    |\\n| G3   | 0.00 – 0.52      | ~1.04          |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The assumption is validated by the contours of the Reynolds number at 0.025 s of half-cycle time elapsed at standard reactor operating conditons, which are illustrated in the images . It is observed that the highest Reynolds number is localized at the precursor injection region for Steps A  with the largest Reynolds number among all reactor configurations being 1.30 for Step A  obtained from the showerhead reactor.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step A configuration is  at 0.025 s\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No turbulent regime is observed throughout all of the reactors due to the atmospheric operating condition which provides more control over the flow pattern. As a result, the flow through the reactor is characterized by laminar behavior\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_6.jpg","caption":"Fig. 6. Contours of Reynolds Number for Step B of various reactor configurations at 0.025 s in the standard condition in Table 3.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_6","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1011,"height":278,"image_format":"jpeg","image_sha256":"903869a0b7189aa53dfbad88e553793f3e5094d9f07cddae5cc9a281efa28158","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":11,"y":44,"width":364,"height":234},{"panel_id":"b","x":395,"y":44,"width":190,"height":232},{"panel_id":"c","x":604,"y":44,"width":193,"height":231},{"panel_id":"d","x":815,"y":44,"width":196,"height":232}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G0. The color gradient indicates varying values of Re, with warmer colors representing higher values.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G1\"},{\"panel_id\":\"c\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G2\"},{\"panel_id\":\"d\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G3\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Re Range | Peak Re Region |\\n|------|-------------------|----------------|\\n| G0   | 0.00 – 0.52      | 2.34 – 2.60    |\\n| G1   | 0.00 – 0.52      | 1.56 – 2.08    |\\n| G2   | 0.00 – 0.52      | 2.34 – 2.60    |\\n| G3   | 0.00 – 0.52      | 1.04 – 1.30    |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The highest Reynolds number is localized at the precursor injection region for Step B, with the largest Reynolds number among all reactor configurations being  2.59 for Step B, the value is  obtained from the showerhead reactor\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step B reactor configurations is at 0.025 s\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No  turbulent regime is observed throughout all of the reactors due to the atmospheric operating condition which provides more control over the flow pattern. As a result, the flow through the reactor is characterized by laminar behavior\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_8.jpg","caption":"Fig. 8. Contours of pressure of HF on the surface of the wafer for a Step A process time of 0.1 s and for a HF feed flow rate of 150 sccm.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_8","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1211,"height":361,"image_format":"jpeg","image_sha256":"083acc9420201fe30d093423d33ebdfb9f66beb055165b017b8688df1be2e1b5","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":7,"y":58,"width":442,"height":299},{"panel_id":"b","x":474,"y":79,"width":220,"height":274},{"panel_id":"c","x":718,"y":79,"width":232,"height":277},{"panel_id":"d","x":974,"y":77,"width":236,"height":279}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G0. The color gradient represents the pressure values, with red indicating higher pressures and blue indicating lower pressures.\"},{\"panel_id\":\"b\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G1.\"},{\"panel_id\":\"c\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G2.\"},{\"panel_id\":\"d\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G3.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Pressure Range (Pa) | High-Pressure Region (Pa) |\\n|------|------------------------------|----------------------------|\\n| G0   | 0 – 28                      | 35 – 49                   |\\n| G1   | 0 – 28                      | 28 – 42                   |\\n| G2   | 0 – 28                      | 49 – 56                   |\\n| G3   | 0 – 28                      | 28 – 42                   |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The flow of the reactor with three inlets (G1) is more evenly distributed than that of G0, and consequently, G1 has less pressure deviation than G0.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A process time of 0.1 s and a HF feed flow rate of 150 sccm.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. G0\\n, 2. G2\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_9.jpg","caption":"Fig. 9. Contours of pressure of TMA on the surface of the wafer for a Step B process time of 0.2 s and for a TMA feed flow rate of 70 sccm.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_9","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1203,"height":364,"image_format":"jpeg","image_sha256":"5f1620ebe3e6e69ff5c53c2cbb3d6bb7763e5769be04777137281150907b32e1","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"c","x":713,"y":60,"width":235,"height":299},{"panel_id":"b","x":458,"y":11,"width":243,"height":350},{"panel_id":"a","x":4,"y":8,"width":443,"height":352},{"panel_id":"d","x":967,"y":62,"width":235,"height":299}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across different reactor models, with color intensity indicating varying pressures for reactor G0\"},{\"panel_id\":\"b\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across reactor G1.\"},{\"panel_id\":\"c\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across reactor G1.\"},{\"panel_id\":\"d\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across reactor G1.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Pressure Range (Pa) | High-Pressure Region (Pa) |\\n|------|------------------------------|----------------------------|\\n| G0   | 0 – 42                      | 56 – 70                   |\\n| G1   | 0 – 42                      | 49 – 56                   |\\n| G2   | 0 – 42                      | 63 – 70                   |\\n| G3   | 0 – 49                      | 42 – 56                   |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. G0 = 2.514 s\\n, 2. G1 = 2.542 s\\n, 3. G2 = 2.528 s\\n, 4. G3 = 2.498 s\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The pressure contours of TMA and the type of species in the thermal ALE cycle of Al2O3 plays a limited role in affecting the flow profile\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step B process time is 0.2 s and TMA feed flow rate is 70 sccm.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It is observed that the shower- head serves to decrease the amount of precursor in central regions of the wafer, and therefore G2 has better uniformity compared to G0 and G1. The half-cycle times  are similar to the half-cycle times of G0 due to the flow resistance of the showerhead divider. Thus, the addition of the showerhead divider improves the uniformity and slightly increases the half-cycle time for Step B.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_11.jpg","caption":"FIG. 11. (Color online) Total Ieads averaged over the wafer during  $\\mathrm{SiO}_2$  PALE in the MERIE reactor. The high magnetic field during step 1 enables low ion energies which promote polymer formation. The high ion energies during step 2 enable sputtering of the passivated layer. Controlling the ion energies during step 2 avoids physical etching and enables self-limiting etching.","id":"train/atomic-layer-etching/simulation-usecase/19/fig_11","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_11","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":364,"height":561,"image_format":"jpeg","image_sha256":"20f82fd4caf1815f2ba04ee744c8d91c5322329218b7618f48cf744898ff900d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":1,"width":218,"height":557},{"panel_id":"b","x":216,"y":1,"width":145,"height":557}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays the ion energy and angular distribution (IEAD) during the first step (passivation). The color gradient represents relative intensity of ions arriving at that specific energy and angle.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays the ion energy and angular distribution (IEAD) during the second step (etching). The color gradient represents relative intensity of ions arriving at that specific energy and angle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Angle (deg) |  Energy (eV) | Intensity |\\n|---|---|---|\\n| -15  | 3 | 0.01 |\\n| 0  | 3 | 0.01 |\\n| 15  | 3 | 0.01 |\\n| -15  | 15 | 0.01 |\\n| -7  | 15 | 0.53 |\\n| 0  | 15 | 1 |\\n| 7  | 15 | 0.53 |\\n| 15  | 15 | 0.01 |\\n| -3  | 58 | 0.01 |\\n| 0  | 58 | 0.53 |\\n| 3  | 58 | 0.01 |\\n| -3  | 83 | 0.01 |\\n| 0  | 83 | 0.53 |\\n| 3  | 83 | 0.01 |\"},{\"panel_id\":\"b\",\"text\":\"| Angle (deg) |  Energy (eV) | Intensity |\\n|---|---|---|\\n| 0  | 42 | 0.01 |\\n| -6 | 63 | 0.01 |\\n| 0  | 63 | 0.85 |\\n| 6  | 63 | 0.01 |\\n| -6 | 68 | 0.01 |\\n| 0  | 68 | 1 |\\n| 6  | 68 | 0.01 |\\n| 0  | 75 | 0.01 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 7 degrees.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 65 - 70 Ev.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 15 degrees.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The most ions arriving at the wafer in are low-energy (with energies below 30 eV). This energy range was chosen to promote passivation.Higher energies would cause sputtering of the surface.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_17.jpg","caption":"FIG. 17. (Color online) Total IEADs averaged over the wafer for PALE with nonsinusoidal biases. First, a main etch is performed with a  $200\\mathrm{V}$  (peak-to-peak) sinusoidal bias followed by PALE utilizing a nonsinusoidal bias waveform with biases of  $50\\mathrm{V}$  (peak-to-peak) for step 1 (passivation) and  $100\\mathrm{V}$  (peak-to-peak) for step 2 (etching). The narrow IEADs with nonsinusoidal biases enable using a single gas mixture for the entire cycle of PALE.","id":"train/atomic-layer-etching/simulation-usecase/19/fig_17","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_17","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_17.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_17.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":378,"height":464,"image_format":"jpeg","image_sha256":"b33f1d053475354cbac7bc0d8312b3cc08c02ebddcbc9dfcd7523bcecd9fdd91","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":3,"width":177,"height":464},{"panel_id":"b","x":176,"y":5,"width":103,"height":460},{"panel_id":"c","x":275,"y":6,"width":102,"height":459}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays  the ion energy and angular distribution (IEAD) for the main etch using a 200 V bias.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays  the ion energy and angular distribution (IEAD) for the passivation step using a 50 V bias.\"},{\"panel_id\":\"c\",\"text\":\"The figure displays  the ion energy and angular distribution (IEAD) for the etching step using a 100 V bias.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Angle (deg) |  Relative Energy (Ei/Vp-p) | Intensity |\\n|---|---|---|\\n| -10  |0.3 | 0.01 |\\n| 0  | 0.3 | 0.01 |\\n| 10  | 0.3 | 0.01 |\\n| -11  |0.4 | 0.01 |\\n| 0  | 0.4 | 0.53 |\\n| 11  | 0.4 | 0.01 |\\n| -8  |0.54 | 0.01 |\\n| 0  | 0.54 | 0.69 |\\n| 8  | 0.54 | 0.01 |\\n| -9  |0.74 | 0.01 |\\n| 0  | 0.74 | 1 |\\n| 9 | 0.74 | 0.01 |\\n| 0  | 0.86 | 0.01 |\"},{\"panel_id\":\"b\",\"text\":\"| Angle (deg) |  Relative Energy (Ei/Vp-p) | Intensity |\\n|---|---|---|\\n| 0  | 0.48 | 0.01 |\\n| -6  |0.58 | 0.01 |\\n| 0  | 0.58 | 0.53 |\\n| 6  | 0.58 | 0.01 |\\n| -9  |0.68 | 0.01 |\\n| 0  | 0.68 | 1 |\\n| 9  | 0.68 | 0.01 |\\n| 0  | 0.85 | 0.01 |\"},{\"panel_id\":\"c\",\"text\":\"| Angle (deg) |  Relative Energy (Ei/Vp-p) | Intensity |\\n|---|---|---|\\n| 0  | 0.54 | 0.01 |\\n| -6  | 0.59 | 0.01 |\\n| 0  |0.59 | 0.53 |\\n| 6  | 0.59 | 0.53 |\\n| -10  | 0.7 | 0.01 |\\n| 0  | 0.7 | 1 |\\n| 10  | 0.7 | 0.53 |\\n| 0  | 0.77 | 0.01 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It decrease, when the angle of arrival deviates from the perpendicular direction (0 degrees).\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 0.35 - 0.85.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It is a bimodal distribution. The heatmap displays two separated regions of maximum intensity (red/orange/yellow areas). They are located at relative energies of approximately 0.35 and 0.75, separated by a lower-intensity green region.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 10 degrees.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_2.jpg","caption":"FIG. 2. (Color online) Total positive ion densities during the PALE of Si using an ICP reactor. (a) Step 1 (passivation) is performed in  $\\mathrm{Ar / Cl}_2 = 80 / 20$  (500 W ICP, 20 mTorr, 100 SCCM) without an applied bias. (b) Step 2 (etching) is performed in pure Ar (500 W ICP, 16 mTorr, 100 SCCM) with a  $100\\mathrm{V}$  sinusoidal bias at  $2\\mathrm{MHz}$ . Contour levels shown are on a scale of  $10^{11}$ . (Densities are plotted using a log scale over 2 decades. The color bar shows the range of the contours relative to the maximum value.)","id":"train/atomic-layer-etching/simulation-usecase/19/fig_2","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_2","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":375,"height":1081,"image_format":"jpeg","image_sha256":"8cc42c079edfb9afacaa844cdb9944b12e84a05a0ef0d77af0c9e23ea5515830","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":373,"height":518},{"panel_id":"b","x":0,"y":524,"width":373,"height":559}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_20_figure_6.jpg","caption":"Figure 6. \"Heat\" maps showing the change in free energy of (a) CVE1 and (b) SL1 reactions for HF exposure on  $\\mathrm{TiO_2}$ . (c) Change in corresponding minimum thermodynamic barrier with respect to reactant pressure and temperature.","id":"train/atomic-layer-etching/simulation-usecase/20/figure_6","sample_id":"atomic-layer-etching/simulation-usecase/20/figure_6","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/Prediction and Validation of the Process Window for Atomic Layer Etching HF Exposure on TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1283,"height":455,"image_format":"jpeg","image_sha256":"e5ce3c82e14b6b63bba61588579fa4035b61b8390b3f4906639cd4b3e28a58ec","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"c","x":844,"y":3,"width":436,"height":447},{"panel_id":"b","x":411,"y":6,"width":428,"height":447},{"panel_id":"a","x":4,"y":3,"width":401,"height":450}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure uses heat maps to visualize the Gibbs free energy (ΔG) for continuous etching (CVE1) and self-limiting modification (SL1) as a function of temperature (X-axis) and reactant pressure (Y-axis). A contour plot showing the reaction conditions favorable or unfavorable for continuous etching (CVE1) across varying temperatures and reactant pressures. Panel (a) shows that CVE1 is “favorable” (green, ΔG < 0) across nearly the entire parameter space, indicating TiO₂ is naturally unstable in HF.\"},{\"panel_id\":\"b\",\"text\":\"A contour plot showing the reaction conditions favorable or unfavorable for self-limiting modification (SL1) across varying temperatures and reactant pressures.\"},{\"panel_id\":\"c\",\"text\":\"A contour plot showing the difference between the two reactions (Y1 − X1) across temperature and pressure. Red regions (Y1 > X1) indicate the self-limiting reaction is energetically preferred (lower ΔG), while green regions (Y1 < X1) indicate continuous etching is preferred. The vertical boundary at 360 K confirms that temperature, not pressure, is the “control knob” dictating the transition between ALE and uncontrolled etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Reactant Pressure [Torr] | 0.0 | 0.5 | 1.0 | 1.5 | 2.0 | 2.5 | 3.0 | 3.5 | 4.0 |\\n|---|---|---|---|---|---|---|---|---|---|\\n| 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 500 K | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 |\\n| 1000 K | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 |\\n| 1500 K | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 |\\n| 2000 K | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature [K] | 0.0 | 500 K | 1000 K | 1500 K | 2000 K |\\n|---|---|---|---|---|---|\\n| 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 500 K | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 |\\n| 1000 K | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 |\\n| 1500 K | -2.4 | -2.4 | -2.4 | -2.4 | -2.4 |\\n| 2000 K | -2.8 | -2.8 | -2.8 | -2.8 | -2.8 |\"},{\"panel_id\":\"c\",\"text\":\"| Temperature [K] | 0.0 | 500 K | 1000 K | 1500 K | 2000 K |\\n|---|---|---|---|---|---|\\n| 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 500 K | -0.4 | -0.4 | -0.4 | -0.4 | -0.4 |\\n| 1000 K | -0.8 | -0.8 | -0.8 | -0.8 | -0.8 |\\n| 1500 K | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 |\\n| 2000 K | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lack of Fluorination / Pristine Surface. At 1500 K, the $\\\\Delta G$ for SL1 is positive (Unfavorable). This means the surface fluorination reaction is non-spontaneous. XPS would show no Fluorine peak (or very low coverage), as the HF gas would simply desorb rather than reacting to form the necessary TiF_x surface layer.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Max Temperature: 360 K. Pressure Effect: No. The boundary line at 360 K is vertical, indicating that the transition is independent of pressure; increasing pressure does not expand the self-limiting window.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 1000 K, the value of Y1 - X1 is negative (Green), meaning the continuous etching reaction (Y1) has a lower energy than the surface modification (X1). The system will bypass the self-limiting step entirely and continuously consume the device material. This destroys the \\\"atomic layer\\\" control, turning the precision process into a rapid, uncontrolled bulk etch.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It indicates that the TiO₂ lattice is thermodynamically unstable relative to the gaseous products (TiF₄ + H₂O). There is no kinetic or thermodynamic barrier preventing the bulk crystal from disintegrating into gas; the reaction is spontaneous everywhere. This necessitates the use of the self-limiting mechanism (Panels b/c) to artificially stop the reaction at the surface.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_10.jpg","caption":"FIG. 10. (Color online) Reactor geometry and total ion densities in the (left) Cl passivation phase and (right)  $\\mathrm{Ar}^+$  ion bombardment phase. A log scale showing two decades is used.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_10","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_10","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":581,"height":505,"image_format":"jpeg","image_sha256":"a15735328218c79069492198bbfccaf1707a0f4a5e0cb25d40d29fb11bff6b02","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":4,"width":577,"height":500}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The visualization represents the total ion density in a cylindrical region, with color gradients indicating varying densities.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Phase        | Location (qualitative) | Total Ion Density (10^11 cm^-3) |\\n|--------------|-------------------------|----------------------------------|\\n| Cl phase     | Inner upper lobe        | 4.71                             |\\n| Cl phase     | Lower-left region       | 2.15                             |\\n| Center line  | Interface (mid-height)  | 3.86                             |\\n| Ar+ phase    | Right mid-height        | 1.47                             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The nonideal fluxes were obtained from reactor scale modeling of an ICP.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. The reactorhas a three-turn flat antenna delivering 300W at 10MHz to the plasma.\\n2. The coil is located above a quartz window 10 cm from the wafer. \\n3. The reactor is 22.5 cm in diameter.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the passivation phase a gas mixture of Ar/Cl2=70/30 at 200 sccm was used with a regulation system maintaining the chamber pressure at 20 mTorr. No bias was applied to the wafer during the passivation phase, however a plasma potential of 25V provided acceleration to ions reaching the wafer surface.\\nThe resulting fluxes to the wafer are 7.0*10^17, 1.1*10^16, 1.6*10^15, and 4.9*10^14 /cm2 s, for Cl, Cl2+, Cl+, and Ar+, with ion energies peaked around 24 V.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_66_figure_2.jpg","caption":"Figure 2. Maps of the coercivity variations for (a)  $\\mathrm{ZnO}$  (oxygen deficient) and (b)  $\\mathrm{ZnO / Fe}$  ALD layers.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_2","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A contour heatmap showing the magnetic coercivity (Hc) of ZnO (oxygen deficient) particles\"},{\"panel_id\":\"b\",\"text\":\"A contour heatmap showing the magnetic coercivity (Hc) of ZnO:Fe ALD layers\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| X (µm) Range | Y (µm) Range | H_c (Oe) |\\n|---|---|---|\\n| 0–100   | 0–100   | 0–500 |\\n| 100–200 | 0–100   | 500–1500 |\\n| 200–300 | 0–100   | 1500–2500 |\\n| 300–400 | 0–100   | 2500–3500 |\\n| 0–100   | 100–200 | 0–500 |\\n| 100–200 | 100–200 | 500–1500 |\\n| 200–300 | 100–200 | 1500–2500 |\\n| 300–400 | 100–200 | 2500–3500 |\\n| 0–100   | 200–300 | 0–500 |\\n| 100–200 | 200–300 | 500–1500 |\\n| 200–300 | 200–300 | 1500–2500 |\\n| 300–400 | 200–300 | 2500–3500 |\\n| 0–100   | 300–350 | 0–500 |\\n| 100–200 | 300–350 | 500–1500 |\\n| 200–300 | 300–350 | 1500–2500 |\\n| 300–400 | 300–350 | 1500–2500 |\"},{\"panel_id\":\"b\",\"text\":\"| X (µm) Range | Y (µm) Range | H_c (Oe) |\\n|---|---|---|\\n| 0–60   | 0–60   | 0–4000 |\\n| 60–120 | 0–60   | 0–4000 |\\n| 120–180| 0–60   | 0–4000 |\\n| 0–60   | 60–120 | 0–4000 |\\n| 60–120 | 60–120 | 0–4000 |\\n| 120–180| 60–120 | 0–4000 |\\n| 0–60   | 120–180| 0–4000 |\\n| 60–120 | 120–180| 0–4000 |\\n| 120–180| 60–120 | 0–4000 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Coercivity map of the undoped ZnO (oxygen deficient) is uniform\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The observed locations with very high Hc are stochastically distributed and can be assigned to the measurement artefacts.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Hc magnitude, the level of inhomogeneity and the overall magnetization is significantly stronger for the Fe-doped sample\"}]}]","bbox":[{"panel_id":"a","x":3,"y":7,"width":641,"height":543},{"panel_id":"b","x":633,"y":2,"width":650,"height":548}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/Paskaleva et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"66","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:image_caption"},"width":1289,"height":555,"image_format":"jpeg","image_sha256":"5559422d2bf1c02ffe00f342c3a80f2f2ab6f615dfe743bc4d1cdb09810de632","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_66_figure_3.jpg","caption":"Figure 3. Maps of the Kerr angle variations for (a)  $\\mathrm{ZnO}$  (oxygen deficient) and (b)  $\\mathrm{ZnO / Fe}$  ALD layers.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_3","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows a contour heatmap of Kerr rotation angle (θKerr) in millidegrees for oxygen deficient ZnO.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows a contour heatmap of Kerr rotation angle (θKerr) in millidegrees for ZnO:Fe samples.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| X (µm) | Y (µm) | θ_Kerr (mdeg) |\\n|---|---|---|\\n| 0–50   | 0–50   | 0–40 |\\n| 50–100 | 0–50   | 40–80 |\\n| 100–150| 0–50   | 80–120 |\\n| 150–200| 0–50   | 120–160 |\\n| 200–250| 0–50   | 160–200 |\\n| 250–300| 0–50   | 200–220 |\\n| 300–350| 0–50   | 160–200 |\\n| 350–400| 0–50   | 160–200 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Undoped ZnO layers are homogeneous, and the values of θKerr are very small, most of the data are in the range [0−20]\\nmdeg.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ZnO/Fe layer demonstrates inhomogeneous distribution of θKerr whose values are significantly (2 orders of magnitude) larger than those of the undoped layer.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"An interesting feature is that both clockwise (positive θKerr) and counterclockwise (negative θKerr) Kerr polarization rotations are observed in the Kerr angle map of Fe-doped ZnO, indicative of complex MO process activity\"}]}]","bbox":[{"panel_id":"a","x":4,"y":3,"width":652,"height":550},{"panel_id":"b","x":659,"y":1,"width":626,"height":553}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/Paskaleva et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"66","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:image_caption"},"width":1292,"height":556,"image_format":"jpeg","image_sha256":"ed0ddf14ba0921433574f86bea53b6a1a3faaeab914c8878f4e151162b58ed93","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_5.jpg","caption":"Fig. 5. Contours of Reynolds Number for Step A of various reactor configurations at 0.025 s in the standard condition in Table 3.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_5","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G0. The color gradient indicates varying values of Re, with darker shades representing higher values.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G1.\"},{\"panel_id\":\"c\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G2.\"},{\"panel_id\":\"d\",\"text\":\"The figure displays a contour of Reynolds Number (Re) for step A configuration across reactor model G3.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Re Range | Peak Re Region |\\n|------|-------------------|----------------|\\n| G0   | 0.00 – 0.52      | 1.04 – 1.30    |\\n| G1   | 0.00 – 0.52      | 1.04 – 1.30    |\\n| G2   | 0.00 – 0.52      | 1.04 – 1.30    |\\n| G3   | 0.00 – 0.52      | ~1.04          |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The assumption is validated by the contours of the Reynolds number at 0.025 s of half-cycle time elapsed at standard reactor operating conditons, which are illustrated in the images . It is observed that the highest Reynolds number is localized at the precursor injection region for Steps A  with the largest Reynolds number among all reactor configurations being 1.30 for Step A  obtained from the showerhead reactor.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step A configuration is  at 0.025 s\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No turbulent regime is observed throughout all of the reactors due to the atmospheric operating condition which provides more control over the flow pattern. As a result, the flow through the reactor is characterized by laminar behavior\"}]}]","bbox":[{"panel_id":"a","x":14,"y":41,"width":360,"height":236},{"panel_id":"b","x":394,"y":41,"width":192,"height":237},{"panel_id":"c","x":605,"y":41,"width":191,"height":234},{"panel_id":"d","x":815,"y":41,"width":196,"height":233}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1011,"height":278,"image_format":"jpeg","image_sha256":"6a9d8377d5db93a5ec1f41102fe5ac8f55326e730473c845e131b3f26182b9cf","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_6.jpg","caption":"Fig. 6. Contours of Reynolds Number for Step B of various reactor configurations at 0.025 s in the standard condition in Table 3.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_6","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G0. The color gradient indicates varying values of Re, with warmer colors representing higher values.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G1\"},{\"panel_id\":\"c\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G2\"},{\"panel_id\":\"d\",\"text\":\"The figure displays a contour of Reynolds number (Re) across four different reactor models labeled G3\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Re Range | Peak Re Region |\\n|------|-------------------|----------------|\\n| G0   | 0.00 – 0.52      | 2.34 – 2.60    |\\n| G1   | 0.00 – 0.52      | 1.56 – 2.08    |\\n| G2   | 0.00 – 0.52      | 2.34 – 2.60    |\\n| G3   | 0.00 – 0.52      | 1.04 – 1.30    |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The highest Reynolds number is localized at the precursor injection region for Step B, with the largest Reynolds number among all reactor configurations being  2.59 for Step B, the value is  obtained from the showerhead reactor\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step B reactor configurations is at 0.025 s\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No  turbulent regime is observed throughout all of the reactors due to the atmospheric operating condition which provides more control over the flow pattern. As a result, the flow through the reactor is characterized by laminar behavior\"}]}]","bbox":[{"panel_id":"a","x":11,"y":44,"width":364,"height":234},{"panel_id":"b","x":395,"y":44,"width":190,"height":232},{"panel_id":"c","x":604,"y":44,"width":193,"height":231},{"panel_id":"d","x":815,"y":44,"width":196,"height":232}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1011,"height":278,"image_format":"jpeg","image_sha256":"903869a0b7189aa53dfbad88e553793f3e5094d9f07cddae5cc9a281efa28158","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_8.jpg","caption":"Fig. 8. Contours of pressure of HF on the surface of the wafer for a Step A process time of 0.1 s and for a HF feed flow rate of 150 sccm.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_8","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G0. The color gradient represents the pressure values, with red indicating higher pressures and blue indicating lower pressures.\"},{\"panel_id\":\"b\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G1.\"},{\"panel_id\":\"c\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G2.\"},{\"panel_id\":\"d\",\"text\":\"The image displays a contours of the pressure of HF (Hydrofluoric acid) over a wafer across four different conditions  for reactor G3.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Pressure Range (Pa) | High-Pressure Region (Pa) |\\n|------|------------------------------|----------------------------|\\n| G0   | 0 – 28                      | 35 – 49                   |\\n| G1   | 0 – 28                      | 28 – 42                   |\\n| G2   | 0 – 28                      | 49 – 56                   |\\n| G3   | 0 – 28                      | 28 – 42                   |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The flow of the reactor with three inlets (G1) is more evenly distributed than that of G0, and consequently, G1 has less pressure deviation than G0.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A process time of 0.1 s and a HF feed flow rate of 150 sccm.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. G0\\n, 2. G2\"}]}]","bbox":[{"panel_id":"a","x":7,"y":58,"width":442,"height":299},{"panel_id":"b","x":474,"y":79,"width":220,"height":274},{"panel_id":"c","x":718,"y":79,"width":232,"height":277},{"panel_id":"d","x":974,"y":77,"width":236,"height":279}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1211,"height":361,"image_format":"jpeg","image_sha256":"083acc9420201fe30d093423d33ebdfb9f66beb055165b017b8688df1be2e1b5","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_17_fig_9.jpg","caption":"Fig. 9. Contours of pressure of TMA on the surface of the wafer for a Step B process time of 0.2 s and for a TMA feed flow rate of 70 sccm.","id":"train/atomic-layer-etching/simulation-usecase/17/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/17/fig_9","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"},{"panel_id":"d","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across different reactor models, with color intensity indicating varying pressures for reactor G0\"},{\"panel_id\":\"b\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across reactor G1.\"},{\"panel_id\":\"c\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across reactor G1.\"},{\"panel_id\":\"d\",\"text\":\"The image displays a contour representing the pressure of TMA (Tetrakis(methoxymethyl)amine) on surface across reactor G1.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Case | Dominant Pressure Range (Pa) | High-Pressure Region (Pa) |\\n|------|------------------------------|----------------------------|\\n| G0   | 0 – 42                      | 56 – 70                   |\\n| G1   | 0 – 42                      | 49 – 56                   |\\n| G2   | 0 – 42                      | 63 – 70                   |\\n| G3   | 0 – 49                      | 42 – 56                   |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. G0 = 2.514 s\\n, 2. G1 = 2.542 s\\n, 3. G2 = 2.528 s\\n, 4. G3 = 2.498 s\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The pressure contours of TMA and the type of species in the thermal ALE cycle of Al2O3 plays a limited role in affecting the flow profile\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Step B process time is 0.2 s and TMA feed flow rate is 70 sccm.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It is observed that the shower- head serves to decrease the amount of precursor in central regions of the wafer, and therefore G2 has better uniformity compared to G0 and G1. The half-cycle times  are similar to the half-cycle times of G0 due to the flow resistance of the showerhead divider. Thus, the addition of the showerhead divider improves the uniformity and slightly increases the half-cycle time for Step B.\"}]}]","bbox":[{"panel_id":"c","x":713,"y":60,"width":235,"height":299},{"panel_id":"b","x":458,"y":11,"width":243,"height":350},{"panel_id":"a","x":4,"y":8,"width":443,"height":352},{"panel_id":"d","x":967,"y":62,"width":235,"height":299}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1203,"height":364,"image_format":"jpeg","image_sha256":"5f1620ebe3e6e69ff5c53c2cbb3d6bb7763e5769be04777137281150907b32e1","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_11.jpg","caption":"FIG. 11. (Color online) Total Ieads averaged over the wafer during  $\\mathrm{SiO}_2$  PALE in the MERIE reactor. The high magnetic field during step 1 enables low ion energies which promote polymer formation. The high ion energies during step 2 enable sputtering of the passivated layer. Controlling the ion energies during step 2 avoids physical etching and enables self-limiting etching.","id":"train/atomic-layer-etching/simulation-usecase/19/fig_11","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_11","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays the ion energy and angular distribution (IEAD) during the first step (passivation). The color gradient represents relative intensity of ions arriving at that specific energy and angle.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays the ion energy and angular distribution (IEAD) during the second step (etching). The color gradient represents relative intensity of ions arriving at that specific energy and angle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Angle (deg) |  Energy (eV) | Intensity |\\n|---|---|---|\\n| -15  | 3 | 0.01 |\\n| 0  | 3 | 0.01 |\\n| 15  | 3 | 0.01 |\\n| -15  | 15 | 0.01 |\\n| -7  | 15 | 0.53 |\\n| 0  | 15 | 1 |\\n| 7  | 15 | 0.53 |\\n| 15  | 15 | 0.01 |\\n| -3  | 58 | 0.01 |\\n| 0  | 58 | 0.53 |\\n| 3  | 58 | 0.01 |\\n| -3  | 83 | 0.01 |\\n| 0  | 83 | 0.53 |\\n| 3  | 83 | 0.01 |\"},{\"panel_id\":\"b\",\"text\":\"| Angle (deg) |  Energy (eV) | Intensity |\\n|---|---|---|\\n| 0  | 42 | 0.01 |\\n| -6 | 63 | 0.01 |\\n| 0  | 63 | 0.85 |\\n| 6  | 63 | 0.01 |\\n| -6 | 68 | 0.01 |\\n| 0  | 68 | 1 |\\n| 6  | 68 | 0.01 |\\n| 0  | 75 | 0.01 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 7 degrees.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 65 - 70 Ev.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 15 degrees.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The most ions arriving at the wafer in are low-energy (with energies below 30 eV). This energy range was chosen to promote passivation.Higher energies would cause sputtering of the surface.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":218,"height":557},{"panel_id":"b","x":216,"y":1,"width":145,"height":557}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":364,"height":561,"image_format":"jpeg","image_sha256":"20f82fd4caf1815f2ba04ee744c8d91c5322329218b7618f48cf744898ff900d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_17.jpg","caption":"FIG. 17. (Color online) Total IEADs averaged over the wafer for PALE with nonsinusoidal biases. First, a main etch is performed with a  $200\\mathrm{V}$  (peak-to-peak) sinusoidal bias followed by PALE utilizing a nonsinusoidal bias waveform with biases of  $50\\mathrm{V}$  (peak-to-peak) for step 1 (passivation) and  $100\\mathrm{V}$  (peak-to-peak) for step 2 (etching). The narrow IEADs with nonsinusoidal biases enable using a single gas mixture for the entire cycle of PALE.","id":"train/atomic-layer-etching/simulation-usecase/19/fig_17","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_17","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays  the ion energy and angular distribution (IEAD) for the main etch using a 200 V bias.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays  the ion energy and angular distribution (IEAD) for the passivation step using a 50 V bias.\"},{\"panel_id\":\"c\",\"text\":\"The figure displays  the ion energy and angular distribution (IEAD) for the etching step using a 100 V bias.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Angle (deg) |  Relative Energy (Ei/Vp-p) | Intensity |\\n|---|---|---|\\n| -10  |0.3 | 0.01 |\\n| 0  | 0.3 | 0.01 |\\n| 10  | 0.3 | 0.01 |\\n| -11  |0.4 | 0.01 |\\n| 0  | 0.4 | 0.53 |\\n| 11  | 0.4 | 0.01 |\\n| -8  |0.54 | 0.01 |\\n| 0  | 0.54 | 0.69 |\\n| 8  | 0.54 | 0.01 |\\n| -9  |0.74 | 0.01 |\\n| 0  | 0.74 | 1 |\\n| 9 | 0.74 | 0.01 |\\n| 0  | 0.86 | 0.01 |\"},{\"panel_id\":\"b\",\"text\":\"| Angle (deg) |  Relative Energy (Ei/Vp-p) | Intensity |\\n|---|---|---|\\n| 0  | 0.48 | 0.01 |\\n| -6  |0.58 | 0.01 |\\n| 0  | 0.58 | 0.53 |\\n| 6  | 0.58 | 0.01 |\\n| -9  |0.68 | 0.01 |\\n| 0  | 0.68 | 1 |\\n| 9  | 0.68 | 0.01 |\\n| 0  | 0.85 | 0.01 |\"},{\"panel_id\":\"c\",\"text\":\"| Angle (deg) |  Relative Energy (Ei/Vp-p) | Intensity |\\n|---|---|---|\\n| 0  | 0.54 | 0.01 |\\n| -6  | 0.59 | 0.01 |\\n| 0  |0.59 | 0.53 |\\n| 6  | 0.59 | 0.53 |\\n| -10  | 0.7 | 0.01 |\\n| 0  | 0.7 | 1 |\\n| 10  | 0.7 | 0.53 |\\n| 0  | 0.77 | 0.01 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It decrease, when the angle of arrival deviates from the perpendicular direction (0 degrees).\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 0.35 - 0.85.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It is a bimodal distribution. The heatmap displays two separated regions of maximum intensity (red/orange/yellow areas). They are located at relative energies of approximately 0.35 and 0.75, separated by a lower-intensity green region.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 10 degrees.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":177,"height":464},{"panel_id":"b","x":176,"y":5,"width":103,"height":460},{"panel_id":"c","x":275,"y":6,"width":102,"height":459}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_17.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_17.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":378,"height":464,"image_format":"jpeg","image_sha256":"b33f1d053475354cbac7bc0d8312b3cc08c02ebddcbc9dfcd7523bcecd9fdd91","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_2.jpg","caption":"FIG. 2. (Color online) Total positive ion densities during the PALE of Si using an ICP reactor. (a) Step 1 (passivation) is performed in  $\\mathrm{Ar / Cl}_2 = 80 / 20$  (500 W ICP, 20 mTorr, 100 SCCM) without an applied bias. (b) Step 2 (etching) is performed in pure Ar (500 W ICP, 16 mTorr, 100 SCCM) with a  $100\\mathrm{V}$  sinusoidal bias at  $2\\mathrm{MHz}$ . Contour levels shown are on a scale of  $10^{11}$ . (Densities are plotted using a log scale over 2 decades. The color bar shows the range of the contours relative to the maximum value.)","id":"train/atomic-layer-etching/simulation-usecase/19/fig_2","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_2","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":373,"height":518},{"panel_id":"b","x":0,"y":524,"width":373,"height":559}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":375,"height":1081,"image_format":"jpeg","image_sha256":"8cc42c079edfb9afacaa844cdb9944b12e84a05a0ef0d77af0c9e23ea5515830","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_20_figure_6.jpg","caption":"Figure 6. \"Heat\" maps showing the change in free energy of (a) CVE1 and (b) SL1 reactions for HF exposure on  $\\mathrm{TiO_2}$ . (c) Change in corresponding minimum thermodynamic barrier with respect to reactant pressure and temperature.","id":"train/atomic-layer-etching/simulation-usecase/20/figure_6","sample_id":"atomic-layer-etching/simulation-usecase/20/figure_6","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure uses heat maps to visualize the Gibbs free energy (ΔG) for continuous etching (CVE1) and self-limiting modification (SL1) as a function of temperature (X-axis) and reactant pressure (Y-axis). A contour plot showing the reaction conditions favorable or unfavorable for continuous etching (CVE1) across varying temperatures and reactant pressures. Panel (a) shows that CVE1 is “favorable” (green, ΔG < 0) across nearly the entire parameter space, indicating TiO₂ is naturally unstable in HF.\"},{\"panel_id\":\"b\",\"text\":\"A contour plot showing the reaction conditions favorable or unfavorable for self-limiting modification (SL1) across varying temperatures and reactant pressures.\"},{\"panel_id\":\"c\",\"text\":\"A contour plot showing the difference between the two reactions (Y1 − X1) across temperature and pressure. Red regions (Y1 > X1) indicate the self-limiting reaction is energetically preferred (lower ΔG), while green regions (Y1 < X1) indicate continuous etching is preferred. The vertical boundary at 360 K confirms that temperature, not pressure, is the “control knob” dictating the transition between ALE and uncontrolled etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Reactant Pressure [Torr] | 0.0 | 0.5 | 1.0 | 1.5 | 2.0 | 2.5 | 3.0 | 3.5 | 4.0 |\\n|---|---|---|---|---|---|---|---|---|---|\\n| 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 500 K | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 |\\n| 1000 K | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 |\\n| 1500 K | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 | -1.8 |\\n| 2000 K | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature [K] | 0.0 | 500 K | 1000 K | 1500 K | 2000 K |\\n|---|---|---|---|---|---|\\n| 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 500 K | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 |\\n| 1000 K | -2.0 | -2.0 | -2.0 | -2.0 | -2.0 |\\n| 1500 K | -2.4 | -2.4 | -2.4 | -2.4 | -2.4 |\\n| 2000 K | -2.8 | -2.8 | -2.8 | -2.8 | -2.8 |\"},{\"panel_id\":\"c\",\"text\":\"| Temperature [K] | 0.0 | 500 K | 1000 K | 1500 K | 2000 K |\\n|---|---|---|---|---|---|\\n| 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 500 K | -0.4 | -0.4 | -0.4 | -0.4 | -0.4 |\\n| 1000 K | -0.8 | -0.8 | -0.8 | -0.8 | -0.8 |\\n| 1500 K | -1.2 | -1.2 | -1.2 | -1.2 | -1.2 |\\n| 2000 K | -1.6 | -1.6 | -1.6 | -1.6 | -1.6 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lack of Fluorination / Pristine Surface. At 1500 K, the $\\\\Delta G$ for SL1 is positive (Unfavorable). This means the surface fluorination reaction is non-spontaneous. XPS would show no Fluorine peak (or very low coverage), as the HF gas would simply desorb rather than reacting to form the necessary TiF_x surface layer.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Max Temperature: 360 K. Pressure Effect: No. The boundary line at 360 K is vertical, indicating that the transition is independent of pressure; increasing pressure does not expand the self-limiting window.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 1000 K, the value of Y1 - X1 is negative (Green), meaning the continuous etching reaction (Y1) has a lower energy than the surface modification (X1). The system will bypass the self-limiting step entirely and continuously consume the device material. This destroys the \\\"atomic layer\\\" control, turning the precision process into a rapid, uncontrolled bulk etch.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It indicates that the TiO₂ lattice is thermodynamically unstable relative to the gaseous products (TiF₄ + H₂O). There is no kinetic or thermodynamic barrier preventing the bulk crystal from disintegrating into gas; the reaction is spontaneous everywhere. This necessitates the use of the self-limiting mechanism (Panels b/c) to artificially stop the reaction at the surface.\"}]}]","bbox":[{"panel_id":"c","x":844,"y":3,"width":436,"height":447},{"panel_id":"b","x":411,"y":6,"width":428,"height":447},{"panel_id":"a","x":4,"y":3,"width":401,"height":450}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/Prediction and Validation of the Process Window for Atomic Layer Etching HF Exposure on TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":1283,"height":455,"image_format":"jpeg","image_sha256":"e5ce3c82e14b6b63bba61588579fa4035b61b8390b3f4906639cd4b3e28a58ec","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_10.jpg","caption":"FIG. 10. (Color online) Reactor geometry and total ion densities in the (left) Cl passivation phase and (right)  $\\mathrm{Ar}^+$  ion bombardment phase. A log scale showing two decades is used.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_10","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_10","subset":"contour-heatmap","split":"train","classification":[{"panel_id":"a","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The visualization represents the total ion density in a cylindrical region, with color gradients indicating varying densities.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Phase        | Location (qualitative) | Total Ion Density (10^11 cm^-3) |\\n|--------------|-------------------------|----------------------------------|\\n| Cl phase     | Inner upper lobe        | 4.71                             |\\n| Cl phase     | Lower-left region       | 2.15                             |\\n| Center line  | Interface (mid-height)  | 3.86                             |\\n| Ar+ phase    | Right mid-height        | 1.47                             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The nonideal fluxes were obtained from reactor scale modeling of an ICP.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. The reactorhas a three-turn flat antenna delivering 300W at 10MHz to the plasma.\\n2. The coil is located above a quartz window 10 cm from the wafer. \\n3. The reactor is 22.5 cm in diameter.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the passivation phase a gas mixture of Ar/Cl2=70/30 at 200 sccm was used with a regulation system maintaining the chamber pressure at 20 mTorr. No bias was applied to the wafer during the passivation phase, however a plasma potential of 25V provided acceleration to ions reaching the wafer surface.\\nThe resulting fluxes to the wafer are 7.0*10^17, 1.1*10^16, 1.6*10^15, and 4.9*10^14 /cm2 s, for Cl, Cl2+, Cl+, and Ar+, with ion energies peaked around 24 V.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":3,"y":4,"width":577,"height":500}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"contour heatmap","caption_source":"content.json:img_caption"},"width":581,"height":505,"image_format":"jpeg","image_sha256":"a15735328218c79069492198bbfccaf1707a0f4a5e0cb25d40d29fb11bff6b02","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/device-structure-diagram/test/metadata.jsonl

@@ -1,4 +1,4 @@
-{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_6_fig_1.jpg","caption":"FIG. 1. (Color online) Schematic of the custom-built, hot-wall ALD reactor. The main part is a cylindrical tube, heated externally. The tube is fitted with inlets for  $\\mathrm{N}_2$  carrier gas and precursors close to its upstream end. A sample stage with in-situ temperature monitoring is positioned near the center of the tube. A separately pumped mass spectrometer is used to analyze gas sampled downstream of the stage by means of a needle valve. The calculated carrier gas flow and diffusion through the tube indicate even distribution of the gas. The flow profile in the mid section of the tube is shown.","id":"test/atomic-layer-deposition/experimental-usecase/6/fig_1","sample_id":"atomic-layer-deposition/experimental-usecase/6/fig_1","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_1.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/Mullings et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":1411,"height":677,"image_format":"jpeg","image_sha256":"940c977d2468c14976b944a6a6693f7f7450f9281cb0f2beda22a1f83fcef395","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":1411,"height":677}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_1_figure_3.jpg","caption":"Figure 3. FinFET CMOS transistor architecture","id":"test/atomic-layer-etching/experimental-usecase/1/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/1/figure_3","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/Atomic Layer Etching An Industry Perspective.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"1","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":472,"height":620,"image_format":"jpeg","image_sha256":"577d2de2336bf974813253183b95900eb142a9e9e180b30aea19f9c2877a6dfa","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":572,"height":620}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_1_figure_4.jpg","caption":"Figure 4. Gate-All-Around CMOS transistor architecture.","id":"test/atomic-layer-etching/experimental-usecase/1/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/1/figure_4","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/Atomic Layer Etching An Industry Perspective.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"1","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":472,"height":630,"image_format":"jpeg","image_sha256":"fa380a83cdc155b17c123c3d81d29f147e95b07e70df641427d99fe8af440d10","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":472,"height":630}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_fig_14.jpg","caption":"Fig. 14. Multiscale CFD pressure data of the generation of DMAF in the TMA injection region (a) with and (b) without an EWMA-based R2R control system for  $\\lambda = 0.3$  when subjected to a shift disturbance of 0.6 at the 10th batch run.","id":"test/atomic-layer-etching/simulation-usecase/34/fig_14","sample_id":"atomic-layer-etching/simulation-usecase/34/fig_14","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_14.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:image_caption"},"width":872,"height":363,"image_format":"jpeg","image_sha256":"957660513003d43362bc0b2124e8f05594874601a6d4d59439e0f9c6ce14d993","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":872,"height":363}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
+{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_6_fig_1.jpg","caption":"FIG. 1. (Color online) Schematic of the custom-built, hot-wall ALD reactor. The main part is a cylindrical tube, heated externally. The tube is fitted with inlets for  $\\mathrm{N}_2$  carrier gas and precursors close to its upstream end. A sample stage with in-situ temperature monitoring is positioned near the center of the tube. A separately pumped mass spectrometer is used to analyze gas sampled downstream of the stage by means of a needle valve. The calculated carrier gas flow and diffusion through the tube indicate even distribution of the gas. The flow profile in the mid section of the tube is shown.","id":"test/atomic-layer-deposition/experimental-usecase/6/fig_1","sample_id":"atomic-layer-deposition/experimental-usecase/6/fig_1","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":1411,"height":677}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_1.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/Mullings et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":1411,"height":677,"image_format":"jpeg","image_sha256":"940c977d2468c14976b944a6a6693f7f7450f9281cb0f2beda22a1f83fcef395","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_1_figure_3.jpg","caption":"Figure 3. FinFET CMOS transistor architecture","id":"test/atomic-layer-etching/experimental-usecase/1/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/1/figure_3","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":572,"height":620}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/Atomic Layer Etching An Industry Perspective.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"1","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":472,"height":620,"image_format":"jpeg","image_sha256":"577d2de2336bf974813253183b95900eb142a9e9e180b30aea19f9c2877a6dfa","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_1_figure_4.jpg","caption":"Figure 4. Gate-All-Around CMOS transistor architecture.","id":"test/atomic-layer-etching/experimental-usecase/1/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/1/figure_4","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":472,"height":630}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/images/figure_4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/1/Atomic Layer Etching An Industry Perspective.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"1","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":472,"height":630,"image_format":"jpeg","image_sha256":"fa380a83cdc155b17c123c3d81d29f147e95b07e70df641427d99fe8af440d10","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_fig_14.jpg","caption":"Fig. 14. Multiscale CFD pressure data of the generation of DMAF in the TMA injection region (a) with and (b) without an EWMA-based R2R control system for  $\\lambda = 0.3$  when subjected to a shift disturbance of 0.6 at the 10th batch run.","id":"test/atomic-layer-etching/simulation-usecase/34/fig_14","sample_id":"atomic-layer-etching/simulation-usecase/34/fig_14","subset":"device-structure-diagram","split":"test","classification":[{"panel_id":"a","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":872,"height":363}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_14.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:image_caption"},"width":872,"height":363,"image_format":"jpeg","image_sha256":"957660513003d43362bc0b2124e8f05594874601a6d4d59439e0f9c6ce14d993","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/device-structure-diagram/train/metadata.jsonl

@@ -1,6 +1,6 @@
-{"file_name":"images/train_atomic-layer-etching_experimental-usecase_9_figure_5.jpg","caption":"Figure 5. (a) Starting heterostructure for InGaAs n-channel FinFETs grown by MBE. (b) Cross-sectional schematics of the FinFET along the source-drain direction and across the fin. Because of the partial ALE selectivity between InGaAs and InAlAs, the InGaAs channel is released. (c) Cross-section TEM image of a finished FinFET with minimum  $W_{\\mathrm{f}} = 2.5 \\mathrm{nm}$  and a highest fin aspect ratio  $\\mathrm{AR} = 20$ . The inset is a close-up image of the InGaAs channel. (d,e) Output and subthreshold characteristics of the most scaled InGaAs FinFET with  $W_{\\mathrm{f}} = 2.5 \\mathrm{nm}$  and  $L_{\\mathrm{g}} = 60 \\mathrm{nm}$ . (f) Benchmark of maximum  $g_{\\mathrm{m}}$  as a function of  $W_{\\mathrm{f}}$  for InGaAs FinFETs and state-of-the-art Si FinFETs.  $V_{\\mathrm{DD}} = 0.5 \\mathrm{V}$  for InGaAs FinFETs and  $V_{\\mathrm{DD}} = 0.7 \\mathrm{V}$  for the latest two generations of Si FinFETs. The blue data represents results from this work, whereas the red data are obtained from the same heterostructure following an identical process to the present one with the exception of in situ thermal ALE and ALD.7","id":"train/atomic-layer-etching/experimental-usecase/9/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/9/figure_5","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"},{"panel_id":"c","label":"image panel"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multi-axis chart"},{"panel_id":"f","label":"multiple scatter plot"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/In Situ Thermal Atomic Layer Etching for Sub5 nm InGaAs.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"9","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":1297,"height":797,"image_format":"jpeg","image_sha256":"bef18d61c989af344c6c2607e6667289e479acb7049e2270219a4ae0385f8e44","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"d","x":17,"y":391,"width":400,"height":363},{"panel_id":"b","x":310,"y":5,"width":539,"height":353},{"panel_id":"c","x":861,"y":3,"width":420,"height":357},{"panel_id":"e","x":433,"y":393,"width":413,"height":355},{"panel_id":"f","x":899,"y":386,"width":402,"height":350},{"panel_id":"a","x":80,"y":21,"width":205,"height":322}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Layer stack of FinFET device\"},{\"panel_id\":\"b\",\"text\":\"A schematic of a FinFET device structure and how it is released\"},{\"panel_id\":\"c\",\"text\":\"Microscopic images of a FinFET device highlighting specific layers and features.\"},{\"panel_id\":\"d\",\"text\":\"Output current (I<sub>D</sub>) versus gate-source voltage (V<sub>GS</sub>) for a FinFET device.\"},{\"panel_id\":\"e\",\"text\":\"Transfer characteristics of a FinFET device showing drain current (I<sub>D</sub>) and gate current (I<sub>G</sub>) versus gate-source voltage (V<sub>GS</sub>).\"},{\"panel_id\":\"f\",\"text\":\"Gate transconductance (g<sub>m</sub>) versus fin width (W<sub>f</sub>) for Si FinFETs and InGaAs FinFETs compared to this work.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Layer Description                | Thickness |\\n|---------------------------------|-----------|\\n| n⁺-In₀.₅₃Ga₀.₄₇As             | 30 nm     |\\n| InP                                        | 4 nm      |\\n| n-In₀.₅₃Ga₀.₄₇As              | 50 nm     |\\n| In₀.₅₂Al₀.₄₈As                    | 5 nm      |\\n| Si δ-doping                       | —         |\\n| In₀.₅₂Al₀.₄₈As buffer       | —         |\\n| S.I. InP substrate              | —         |\"},{\"panel_id\":\"b\",\"text\":\"\"},{\"panel_id\":\"c\",\"text\":\"| Parameter     | Value    |\\n|-------------------|----------|\\n| Fin Width (W)| 50 nm    |\\n| InGaAs             | 2.5 nm   |\\n| HfO₂                 | Present  |\"},{\"panel_id\":\"d\",\"text\":\"| V_DS (V) | I_D @ V_GS=-0.2V (µA/µm) | I_D @ V_GS=0.0V (µA/µm) | I_D @ V_GS=0.2V (µA/µm) | I_D @ V_GS=0.4V (µA/µm) | I_D @ V_GS=0.6V (µA/µm) |\\n|----------|---------------------------|--------------------------|--------------------------|--------------------------|--------------------------|\\n| 0.00     | 0                         | 0                        | 0                        | 0                        | 0                        |\\n| 0.05     | ~2                        | ~12                      | 60                       | 120                      | 180                      |\\n| 0.10     | 5                         | 25                       | 100                      | 180                      | 260                      |\\n| 0.15     | ~6                        | ~32                      | 118                      | 200                      | 285                      |\\n| 0.20     | 8                         | 38                       | 125                      | 210                      | 295                      |\\n| 0.30     | ~9                        | 43                       | 130                      | 215                      | 297                      |\\n| 0.50     | 10                        | 46                       | 132                      | 218                      | 298                      |\"},{\"panel_id\":\"e\",\"text\":\"| V<sub>GS</sub> (V) | I<sub>D</sub> (A/µm) | I<sub>G</sub> (A/µm) |\\n|---|---|---|\\n| -0.2 | 10<sup>-9</sup> | 10<sup>-9</sup> |\\n| -0.1 | 10<sup>-8</sup> | 10<sup>-8</sup> |\\n| 0 | 10<sup>-7</sup> | 10<sup>-7</sup> |\\n| 0.1 | 10<sup>-6</sup> | 10<sup>-6</sup> |\\n| 0.2 | 10<sup>-5</sup> | 10<sup>-5</sup> |\\n| 0.3 | 10<sup>-4</sup> | 10<sup>-4</sup> |\\n| 0.4 | 10<sup>-3</sup> | 10<sup>-3</sup> |\\n| 0.5 | 10<sup>-2</sup> | 10<sup>-2</sup> |\"},{\"panel_id\":\"f\",\"text\":\"| W_f (nm) | g_m This Work (mS/µm) | g_m MIT IEDM '17 (mS/µm) | g_m Si FinFETs (mS/µm) | g_m InGaAs FinFETs (mS/µm) |\\n|----------|------------------------|---------------------------|-------------------------|----------------------------|\\n| 20       | 0.8                   | —                         | —                       | 5                          |\\n| 50       | 0.57                  | 1.3                       | —                       | 8                          |\\n| 10       | 1.45                  | 0.75                      | 1.55                    | 1.11                       |\\n| 12       | 1.5                   | 0.95                      | 1.213                   | 1.0                        |\\n| 14       | 1.7                   | 1.05                      | —                       | 1.517                      |\\n| 17       | 1.8                   | 1.2                       | —                       | 1.25                       |\\n| 20       | 1.9                   | 1.35                      | —                       | 1.222                      |\\n| 22       | —                     | 1.45                      | —                       | 1.123                      |\\n| 25       | —                     | 1.5                       | —                       | 1.252                      |\\n| 27       | —                     | —                         | —                       | 1.352                      |\\n| 28       | —                     | —                         | —                       | 0.65                       |\\n| 30       | —                     | —                         | —                       | 1.153                      |\\n| 33       | —                     | —                         | —                       | 1.537                      |\\n| 37       | —                     | —                         | —                       | 1.4                        |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After 162 cycles of thermal ALE with HF/DMAC 300 °C with PDMAC = 40 mTorr, \\n1. Fin width, Wf ranged from 2.5 to 18 nm\\n2. Gate length, Lg was 60 nm.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The image f benchmarks the maximum gm as a function of Wf for InGaAs FinFETs as well as state-of-the-art Si FinFETs. Among the InGaAs FinFETs, the devices in this paper are the most aggressively scaled in the width dimension. In addition to the results listed above for Wf = 2.5 nm, the InGaAs FinFET in\\nthis paper demonstrated a record gm at Wf below 20 nm. The gm of the InGaAs FinFETs around Wf = 7 nm also approaches that of the second generation Si FinFETs despite a lower VDD = 0.5 V and longer gate length. These results highlight the great potential of this novel technique for future generations of CMOS technology and other device applications.\"}]},{\"panel_id\":\"e\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The device exhibited well-behaved MOSFET characteristics with nearly ideal linear (VDS = 50 mV) and saturation (VDS = 0.5 V) subthreshold swing, Slin and\\nSsat, of 62 and 68 mV/dec, respectively.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_9.jpg","caption":"FIG. 9. (Color online) Plasma properties during  $\\mathrm{SiO}_2$  PALE using a MERIE CCP reactor. (a) Schematic of the reactor. The magnetic field, when applied, is parallel to the electrode. (b)  $\\mathrm{CF}_2$  density. Step 1 is performed in a  $\\mathrm{Ar / C_4F_8 = 75 / 25}$  gas mixture (500 W, 50 mTorr, 100 SCCM, 250 G). The high magnetic field shifts the ions to lower energies. (c)  $\\mathrm{Ar^{+}}$  density during step 2 performed in pure Ar (100 W, 40 mTorr, 300 SCCM). A magnetic field is not used. (Plots are on a log scale over 2 decades.)","id":"train/atomic-layer-etching/simulation-usecase/19/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_9","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":672,"height":591,"image_format":"jpeg","image_sha256":"ee7bdb384431d425d5d0ecd5792630ba756d25d349227620b668510c8786b247","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":1,"y":0,"width":668,"height":160},{"panel_id":"b","x":1,"y":177,"width":669,"height":161},{"panel_id":"c","x":1,"y":354,"width":670,"height":233}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_27_fig_1.jpg","caption":"FIG. 1. (a) Simplified MicroLED structure where surface damage increasingly limits the external quantum efficiency as devices scale down. (b) The general structure of a GaN MISHEMT device showing the recess into the GaN layer. Etching of the recess using reactive ion etching can lead to a rough interface, degrading the device performance.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_1","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/Isotropic atomic layer etching of GaN using SF6 plasma and Al(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"27","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:image_caption"},"width":672,"height":805,"image_format":"jpeg","image_sha256":"0de7b5a0a5337ab1019a03f04fa5d236e45b37d839cf3d1760c012d02a765dba","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":10,"width":672,"height":389},{"panel_id":"b","x":17,"y":452,"width":655,"height":353}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Cross-sectional schematic of a GaN-based MicroLED with a multi-quantum well (MQW) region of alternating InxGa1-xN/GaN layers. The inset highlights etch damage at sidewalls, which limits external quantum efficiency as devices scale down.\"},{\"panel_id\":\"b\",\"text\":\"Cross-sectional schematic of a recessed-gate GaN MISHEMT showing the SiN/AlN/AlGaN/GaN heterostructure on Si substrate. The 2DEG channel at the AlGaN/GaN interface is indicated, and the inset shows etch damage from reactive ion etching of the gate recess.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"},{\"panel_id\":\"b\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Etch damage at the sidewalls increasingly limits the external quantum efficiency.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Reactive ion etching (RIE).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Alternating layers of InxGa1-xN and GaN.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The 2DEG (two-dimensional electron gas) is a highly conductive electron channel that forms at the AlGaN/GaN heterointerface due to spontaneous and piezoelectric polarization effects. This confined electron layer enables extremely high electron mobility because carriers are spatially separated from ionized donors, reducing scattering. The 2DEG is fundamental to HEMT operation as it allows for superior switching frequencies compared to conventional transistors.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_14.jpg","caption":"FIG. 14. (Color online) Initial geometry used for the gate etch case study. The crystalline silicon (c-Si) fin travels back into the page, as indicated by the dotted lines. The resist masks the poly-silicon gate structure, which is pre-etched to a depth just above the top of the fins.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_14","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_14","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":578,"height":452,"image_format":"jpeg","image_sha256":"d254a4e6677541b7935e1888a29c61e6ea96f3287c16063aa2b97ab1f0fda4fa","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":1,"width":574,"height":448}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_6_fig_1.jpg","caption":"FIG. 1. Schematic of the experimental apparatus. The plasma is powered by an inductively coupled coil at the top and by a RF biased substrate stage. The major measurements are of the etch rate (via laser interferometry) and optical emission spectroscopy of species near the silicon substrate being etched during the  $\\mathsf{Ar}^+$  bombardment stage of ALE.","id":"train/atomic-layer-etching/simulation-usecase/6/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/6/fig_1","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/Dynamics of plasma atomic layer etching Molecular.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":672,"height":500,"image_format":"jpeg","image_sha256":"23d424d5ea8e667ce1c7e1e199603122594711e4cef6b3a84d8ff51c90356628","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":672,"height":500}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_8_fig_1.jpg","caption":"FIG. 1. (a) Simplified MicroLED structure where surface damage increasingly limits the external quantum efficiency as devices scale down. (b) The general structure of a GaN MISHEMT device showing the recess into the GaN layer. Etching of the recess using reactive ion etching can lead to a rough interface, degrading the device performance.","id":"train/atomic-layer-etching/simulation-usecase/8/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_1","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/Isotropic atomic layer etching of GaN using SF6 plasma and AI(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":672,"height":805,"image_format":"jpeg","image_sha256":"0de7b5a0a5337ab1019a03f04fa5d236e45b37d839cf3d1760c012d02a765dba","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":2,"width":669,"height":396},{"panel_id":"b","x":0,"y":436,"width":671,"height":368}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A cross-sectional view of a multi-quantum well semiconductor structure, including etch damage due to ion etching.\"},{\"panel_id\":\"b\",\"text\":\"A cross-sectional view of a GaN MISHEMT structure, including etch damage due to ion etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"},{\"panel_id\":\"b\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Silicon nitride is used as an isolating layer between the source, drain, and gate regions to overcome parasitic capacitance.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Typical dielectrics include aluminum oxide and hafnium oxide. In the past, silicon oxide was widely used.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The well is doped with indium.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Reactive ion and radical species in the plasma can damage the interface between the different layers in the device, resulting in less sharp edges leading to lower device efficiencies.\"}]}]"}
+{"file_name":"images/train_atomic-layer-etching_experimental-usecase_9_figure_5.jpg","caption":"Figure 5. (a) Starting heterostructure for InGaAs n-channel FinFETs grown by MBE. (b) Cross-sectional schematics of the FinFET along the source-drain direction and across the fin. Because of the partial ALE selectivity between InGaAs and InAlAs, the InGaAs channel is released. (c) Cross-section TEM image of a finished FinFET with minimum  $W_{\\mathrm{f}} = 2.5 \\mathrm{nm}$  and a highest fin aspect ratio  $\\mathrm{AR} = 20$ . The inset is a close-up image of the InGaAs channel. (d,e) Output and subthreshold characteristics of the most scaled InGaAs FinFET with  $W_{\\mathrm{f}} = 2.5 \\mathrm{nm}$  and  $L_{\\mathrm{g}} = 60 \\mathrm{nm}$ . (f) Benchmark of maximum  $g_{\\mathrm{m}}$  as a function of  $W_{\\mathrm{f}}$  for InGaAs FinFETs and state-of-the-art Si FinFETs.  $V_{\\mathrm{DD}} = 0.5 \\mathrm{V}$  for InGaAs FinFETs and  $V_{\\mathrm{DD}} = 0.7 \\mathrm{V}$  for the latest two generations of Si FinFETs. The blue data represents results from this work, whereas the red data are obtained from the same heterostructure following an identical process to the present one with the exception of in situ thermal ALE and ALD.7","id":"train/atomic-layer-etching/experimental-usecase/9/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/9/figure_5","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"},{"panel_id":"c","label":"image panel"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multi-axis chart"},{"panel_id":"f","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Layer stack of FinFET device\"},{\"panel_id\":\"b\",\"text\":\"A schematic of a FinFET device structure and how it is released\"},{\"panel_id\":\"c\",\"text\":\"Microscopic images of a FinFET device highlighting specific layers and features.\"},{\"panel_id\":\"d\",\"text\":\"Output current (I<sub>D</sub>) versus gate-source voltage (V<sub>GS</sub>) for a FinFET device.\"},{\"panel_id\":\"e\",\"text\":\"Transfer characteristics of a FinFET device showing drain current (I<sub>D</sub>) and gate current (I<sub>G</sub>) versus gate-source voltage (V<sub>GS</sub>).\"},{\"panel_id\":\"f\",\"text\":\"Gate transconductance (g<sub>m</sub>) versus fin width (W<sub>f</sub>) for Si FinFETs and InGaAs FinFETs compared to this work.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Layer Description                | Thickness |\\n|---------------------------------|-----------|\\n| n⁺-In₀.₅₃Ga₀.₄₇As             | 30 nm     |\\n| InP                                        | 4 nm      |\\n| n-In₀.₅₃Ga₀.₄₇As              | 50 nm     |\\n| In₀.₅₂Al₀.₄₈As                    | 5 nm      |\\n| Si δ-doping                       | —         |\\n| In₀.₅₂Al₀.₄₈As buffer       | —         |\\n| S.I. InP substrate              | —         |\"},{\"panel_id\":\"b\",\"text\":\"\"},{\"panel_id\":\"c\",\"text\":\"| Parameter     | Value    |\\n|-------------------|----------|\\n| Fin Width (W)| 50 nm    |\\n| InGaAs             | 2.5 nm   |\\n| HfO₂                 | Present  |\"},{\"panel_id\":\"d\",\"text\":\"| V_DS (V) | I_D @ V_GS=-0.2V (µA/µm) | I_D @ V_GS=0.0V (µA/µm) | I_D @ V_GS=0.2V (µA/µm) | I_D @ V_GS=0.4V (µA/µm) | I_D @ V_GS=0.6V (µA/µm) |\\n|----------|---------------------------|--------------------------|--------------------------|--------------------------|--------------------------|\\n| 0.00     | 0                         | 0                        | 0                        | 0                        | 0                        |\\n| 0.05     | ~2                        | ~12                      | 60                       | 120                      | 180                      |\\n| 0.10     | 5                         | 25                       | 100                      | 180                      | 260                      |\\n| 0.15     | ~6                        | ~32                      | 118                      | 200                      | 285                      |\\n| 0.20     | 8                         | 38                       | 125                      | 210                      | 295                      |\\n| 0.30     | ~9                        | 43                       | 130                      | 215                      | 297                      |\\n| 0.50     | 10                        | 46                       | 132                      | 218                      | 298                      |\"},{\"panel_id\":\"e\",\"text\":\"| V<sub>GS</sub> (V) | I<sub>D</sub> (A/µm) | I<sub>G</sub> (A/µm) |\\n|---|---|---|\\n| -0.2 | 10<sup>-9</sup> | 10<sup>-9</sup> |\\n| -0.1 | 10<sup>-8</sup> | 10<sup>-8</sup> |\\n| 0 | 10<sup>-7</sup> | 10<sup>-7</sup> |\\n| 0.1 | 10<sup>-6</sup> | 10<sup>-6</sup> |\\n| 0.2 | 10<sup>-5</sup> | 10<sup>-5</sup> |\\n| 0.3 | 10<sup>-4</sup> | 10<sup>-4</sup> |\\n| 0.4 | 10<sup>-3</sup> | 10<sup>-3</sup> |\\n| 0.5 | 10<sup>-2</sup> | 10<sup>-2</sup> |\"},{\"panel_id\":\"f\",\"text\":\"| W_f (nm) | g_m This Work (mS/µm) | g_m MIT IEDM '17 (mS/µm) | g_m Si FinFETs (mS/µm) | g_m InGaAs FinFETs (mS/µm) |\\n|----------|------------------------|---------------------------|-------------------------|----------------------------|\\n| 20       | 0.8                   | —                         | —                       | 5                          |\\n| 50       | 0.57                  | 1.3                       | —                       | 8                          |\\n| 10       | 1.45                  | 0.75                      | 1.55                    | 1.11                       |\\n| 12       | 1.5                   | 0.95                      | 1.213                   | 1.0                        |\\n| 14       | 1.7                   | 1.05                      | —                       | 1.517                      |\\n| 17       | 1.8                   | 1.2                       | —                       | 1.25                       |\\n| 20       | 1.9                   | 1.35                      | —                       | 1.222                      |\\n| 22       | —                     | 1.45                      | —                       | 1.123                      |\\n| 25       | —                     | 1.5                       | —                       | 1.252                      |\\n| 27       | —                     | —                         | —                       | 1.352                      |\\n| 28       | —                     | —                         | —                       | 0.65                       |\\n| 30       | —                     | —                         | —                       | 1.153                      |\\n| 33       | —                     | —                         | —                       | 1.537                      |\\n| 37       | —                     | —                         | —                       | 1.4                        |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After 162 cycles of thermal ALE with HF/DMAC 300 °C with PDMAC = 40 mTorr, \\n1. Fin width, Wf ranged from 2.5 to 18 nm\\n2. Gate length, Lg was 60 nm.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The image f benchmarks the maximum gm as a function of Wf for InGaAs FinFETs as well as state-of-the-art Si FinFETs. Among the InGaAs FinFETs, the devices in this paper are the most aggressively scaled in the width dimension. In addition to the results listed above for Wf = 2.5 nm, the InGaAs FinFET in\\nthis paper demonstrated a record gm at Wf below 20 nm. The gm of the InGaAs FinFETs around Wf = 7 nm also approaches that of the second generation Si FinFETs despite a lower VDD = 0.5 V and longer gate length. These results highlight the great potential of this novel technique for future generations of CMOS technology and other device applications.\"}]},{\"panel_id\":\"e\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The device exhibited well-behaved MOSFET characteristics with nearly ideal linear (VDS = 50 mV) and saturation (VDS = 0.5 V) subthreshold swing, Slin and\\nSsat, of 62 and 68 mV/dec, respectively.\"}]}]","bbox":[{"panel_id":"d","x":17,"y":391,"width":400,"height":363},{"panel_id":"b","x":310,"y":5,"width":539,"height":353},{"panel_id":"c","x":861,"y":3,"width":420,"height":357},{"panel_id":"e","x":433,"y":393,"width":413,"height":355},{"panel_id":"f","x":899,"y":386,"width":402,"height":350},{"panel_id":"a","x":80,"y":21,"width":205,"height":322}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/9/In Situ Thermal Atomic Layer Etching for Sub5 nm InGaAs.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"9","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":1297,"height":797,"image_format":"jpeg","image_sha256":"bef18d61c989af344c6c2607e6667289e479acb7049e2270219a4ae0385f8e44","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_9.jpg","caption":"FIG. 9. (Color online) Plasma properties during  $\\mathrm{SiO}_2$  PALE using a MERIE CCP reactor. (a) Schematic of the reactor. The magnetic field, when applied, is parallel to the electrode. (b)  $\\mathrm{CF}_2$  density. Step 1 is performed in a  $\\mathrm{Ar / C_4F_8 = 75 / 25}$  gas mixture (500 W, 50 mTorr, 100 SCCM, 250 G). The high magnetic field shifts the ions to lower energies. (c)  $\\mathrm{Ar^{+}}$  density during step 2 performed in pure Ar (100 W, 40 mTorr, 300 SCCM). A magnetic field is not used. (Plots are on a log scale over 2 decades.)","id":"train/atomic-layer-etching/simulation-usecase/19/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_9","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":1,"y":0,"width":668,"height":160},{"panel_id":"b","x":1,"y":177,"width":669,"height":161},{"panel_id":"c","x":1,"y":354,"width":670,"height":233}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":672,"height":591,"image_format":"jpeg","image_sha256":"ee7bdb384431d425d5d0ecd5792630ba756d25d349227620b668510c8786b247","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_27_fig_1.jpg","caption":"FIG. 1. (a) Simplified MicroLED structure where surface damage increasingly limits the external quantum efficiency as devices scale down. (b) The general structure of a GaN MISHEMT device showing the recess into the GaN layer. Etching of the recess using reactive ion etching can lead to a rough interface, degrading the device performance.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_1","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Cross-sectional schematic of a GaN-based MicroLED with a multi-quantum well (MQW) region of alternating InxGa1-xN/GaN layers. The inset highlights etch damage at sidewalls, which limits external quantum efficiency as devices scale down.\"},{\"panel_id\":\"b\",\"text\":\"Cross-sectional schematic of a recessed-gate GaN MISHEMT showing the SiN/AlN/AlGaN/GaN heterostructure on Si substrate. The 2DEG channel at the AlGaN/GaN interface is indicated, and the inset shows etch damage from reactive ion etching of the gate recess.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"},{\"panel_id\":\"b\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Etch damage at the sidewalls increasingly limits the external quantum efficiency.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Reactive ion etching (RIE).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Alternating layers of InxGa1-xN and GaN.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The 2DEG (two-dimensional electron gas) is a highly conductive electron channel that forms at the AlGaN/GaN heterointerface due to spontaneous and piezoelectric polarization effects. This confined electron layer enables extremely high electron mobility because carriers are spatially separated from ionized donors, reducing scattering. The 2DEG is fundamental to HEMT operation as it allows for superior switching frequencies compared to conventional transistors.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":10,"width":672,"height":389},{"panel_id":"b","x":17,"y":452,"width":655,"height":353}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/Isotropic atomic layer etching of GaN using SF6 plasma and Al(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"27","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:image_caption"},"width":672,"height":805,"image_format":"jpeg","image_sha256":"0de7b5a0a5337ab1019a03f04fa5d236e45b37d839cf3d1760c012d02a765dba","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_14.jpg","caption":"FIG. 14. (Color online) Initial geometry used for the gate etch case study. The crystalline silicon (c-Si) fin travels back into the page, as indicated by the dotted lines. The resist masks the poly-silicon gate structure, which is pre-etched to a depth just above the top of the fins.","id":"train/atomic-layer-etching/simulation-usecase/3/fig_14","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_14","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":5,"y":1,"width":574,"height":448}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":578,"height":452,"image_format":"jpeg","image_sha256":"d254a4e6677541b7935e1888a29c61e6ea96f3287c16063aa2b97ab1f0fda4fa","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_6_fig_1.jpg","caption":"FIG. 1. Schematic of the experimental apparatus. The plasma is powered by an inductively coupled coil at the top and by a RF biased substrate stage. The major measurements are of the etch rate (via laser interferometry) and optical emission spectroscopy of species near the silicon substrate being etched during the  $\\mathsf{Ar}^+$  bombardment stage of ALE.","id":"train/atomic-layer-etching/simulation-usecase/6/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/6/fig_1","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":672,"height":500}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/Dynamics of plasma atomic layer etching Molecular.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":672,"height":500,"image_format":"jpeg","image_sha256":"23d424d5ea8e667ce1c7e1e199603122594711e4cef6b3a84d8ff51c90356628","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_8_fig_1.jpg","caption":"FIG. 1. (a) Simplified MicroLED structure where surface damage increasingly limits the external quantum efficiency as devices scale down. (b) The general structure of a GaN MISHEMT device showing the recess into the GaN layer. Etching of the recess using reactive ion etching can lead to a rough interface, degrading the device performance.","id":"train/atomic-layer-etching/simulation-usecase/8/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_1","subset":"device-structure-diagram","split":"train","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A cross-sectional view of a multi-quantum well semiconductor structure, including etch damage due to ion etching.\"},{\"panel_id\":\"b\",\"text\":\"A cross-sectional view of a GaN MISHEMT structure, including etch damage due to ion etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"},{\"panel_id\":\"b\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Silicon nitride is used as an isolating layer between the source, drain, and gate regions to overcome parasitic capacitance.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Typical dielectrics include aluminum oxide and hafnium oxide. In the past, silicon oxide was widely used.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The well is doped with indium.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Reactive ion and radical species in the plasma can damage the interface between the different layers in the device, resulting in less sharp edges leading to lower device efficiencies.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":669,"height":396},{"panel_id":"b","x":0,"y":436,"width":671,"height":368}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/Isotropic atomic layer etching of GaN using SF6 plasma and AI(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":672,"height":805,"image_format":"jpeg","image_sha256":"0de7b5a0a5337ab1019a03f04fa5d236e45b37d839cf3d1760c012d02a765dba","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/device-structure-diagram/validation/metadata.jsonl

@@ -1 +1 @@
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_10_fig_3.jpg","caption":"Fig. 3 Schematic reresentations of the ractors used in this work: a) an Oxford Instruments FlexAL reactor and (b) a home-built ALD reactor equipped with an in situ FTIR setup, shown here in transmission mode for powder samples. The angle of the IR mirrors on the detector and source side can be adjusted such that planar samples can be studied in reflection mode. Both systems are pumped with turbo pumps and use an inductively coupled plasma source supplied with 13.56 MHz RF power.","id":"validation/atomic-layer-etching/experimental-usecase/10/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/10/fig_3","subset":"device-structure-diagram","split":"validation","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/images/fig_3.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/Investigation of the atomic layer etching mechanism for Al2O3.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":1122,"height":497,"image_format":"jpeg","image_sha256":"0696a2f4d9b52ae8006667213f7520d7359e7395b7fa42d2cc4a28c02f727283","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":8,"y":7,"width":502,"height":490},{"panel_id":"b","x":524,"y":6,"width":591,"height":489}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_10_fig_3.jpg","caption":"Fig. 3 Schematic reresentations of the ractors used in this work: a) an Oxford Instruments FlexAL reactor and (b) a home-built ALD reactor equipped with an in situ FTIR setup, shown here in transmission mode for powder samples. The angle of the IR mirrors on the detector and source side can be adjusted such that planar samples can be studied in reflection mode. Both systems are pumped with turbo pumps and use an inductively coupled plasma source supplied with 13.56 MHz RF power.","id":"validation/atomic-layer-etching/experimental-usecase/10/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/10/fig_3","subset":"device-structure-diagram","split":"validation","classification":[{"panel_id":"a","label":"device structure diagram"},{"panel_id":"b","label":"device structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":8,"y":7,"width":502,"height":490},{"panel_id":"b","x":524,"y":6,"width":591,"height":489}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/images/fig_3.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/10/Investigation of the atomic layer etching mechanism for Al2O3.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"device structure diagram","caption_source":"content.json:img_caption"},"width":1122,"height":497,"image_format":"jpeg","image_sha256":"0696a2f4d9b52ae8006667213f7520d7359e7395b7fa42d2cc4a28c02f727283","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/formula/train/metadata.jsonl

@@ -1 +1 @@
-{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_31_6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052.jpg","caption":"","id":"train/atomic-layer-deposition/simulation-usecase/31/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052","sample_id":"atomic-layer-deposition/simulation-usecase/31/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052","subset":"formula","split":"train","classification":[{"panel_id":"a","label":"formula"},{"panel_id":"b","label":"formula"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/images/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/images/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/Petteri Peltonen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"31","first_classification_panel_id":"a","first_classification_label":"formula","caption_source":"not_found"},"width":653,"height":245,"image_format":"jpeg","image_sha256":"ff07c6fdb5274201c02fb9eb63d40c828fffa7e93587ec713d0c8c37abf6114d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":6,"y":6,"width":647,"height":109},{"panel_id":"b","x":6,"y":113,"width":646,"height":131}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
+{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_31_6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052.jpg","caption":"","id":"train/atomic-layer-deposition/simulation-usecase/31/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052","sample_id":"atomic-layer-deposition/simulation-usecase/31/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052","subset":"formula","split":"train","classification":[{"panel_id":"a","label":"formula"},{"panel_id":"b","label":"formula"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":6,"y":6,"width":647,"height":109},{"panel_id":"b","x":6,"y":113,"width":646,"height":131}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/images/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/images/6d9d789b47e735586a02443aad8b1a7ba33fbab48fbe9bad4ef64f0aa85a2052.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/31/Petteri Peltonen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"31","first_classification_panel_id":"a","first_classification_label":"formula","caption_source":"not_found"},"width":653,"height":245,"image_format":"jpeg","image_sha256":"ff07c6fdb5274201c02fb9eb63d40c828fffa7e93587ec713d0c8c37abf6114d","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/grouped-bar-chart/test/metadata.jsonl

@@ -1,3 +1,3 @@
-{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_33_figure_5.jpg","caption":"Figure 5. Mobility of graphene determined by Hall measurements, after transfer, after plasma treatment, after 100 cycles  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD, and after  $400^{\\circ}\\mathrm{C}$  anneal for pristine,  $\\mathrm{O}_2$ , and  $\\mathrm{H}_2$  plasma treated graphene.","id":"test/atomic-layer-deposition/simulation-usecase/33/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/33/figure_5","subset":"grouped-bar-chart","split":"test","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/images/figure_5.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/René H. J. Vervuurt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:img_caption"},"width":642,"height":347,"image_format":"jpeg","image_sha256":"5b6f3bd42626024de78ac94ff57ea874c1b05843a64fe9984efba26577688fe2","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":642,"height":347}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Grouped bar chart of Hall mobility for graphene samples after transfer, after plasma treatment, after 100 cycles Al2O3 ALD, and after a 400 C anneal, comparing pristine graphene to 30 s O2 plasma and 35 s H2 plasma conditions. Plasma treatment strongly reduces mobility (about 195 cm2/(V s) for O2 and 467 cm2/(V s) for H2), while Al2O3 ALD and subsequent anneal recover mobility, with the H2 plasma sample reaching about 2190 cm2/(V s) after the 400 C anneal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample | Pristine (cm²/(V·s)) | Plasma Treatment (cm²/(V·s)) | Al₂O₃ ALD (cm²/(V·s)) | 400°C Anneal (cm²/(V·s)) |\\n|---|---|---|---|---|\\n| Pristine | 1300 | X | 1500 | 1850 |\\n| O₂ Plasma 30s | 1800 | 180 | 1400 | 1620 |\\n| H₂ Plasma 35s | 1450 | 450 | 1480 | 2200 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent the Hall mobility measured after graphene transfer, before any plasma treatment, for each sample set.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both plasmas reduce mobility, but O2 plasma causes a larger drop, to about 195 cm2/(V s), while H2 plasma reduces it to about 467 cm2/(V s). After 100 Al2O3 ALD cycles, the O2 treated sample partially recovers to about 1390 cm2/(V s), while the H2 treated sample recovers to about 1470 cm2/(V s), roughly returning to its initial value. After the 400 C anneal, the O2 treated sample reaches about 1630 cm2/(V s), whereas the H2 treated sample improves further to about 2190 cm2/(V s).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The paper attributes the mobility drop to plasma induced conversion of sp2 carbon toward sp3 bonding and the formation of out of plane bonds. These added bonds and associated functional groups act as scattering centers for charge carriers, which reduces mobility. This interpretation is consistent with the XPS and Raman evidence of increased sp3 character after plasma exposure.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"H2 plasma functionalization plus Al2O3 ALD plus 400 C anneal: mobility recovers to the initial value after ALD and then increases to about 152% of the original value after anneal (about 2190 cm2/(V s)).\\n\\nO2 plasma functionalization plus Al2O3 ALD plus 400 C anneal: mobility improves strongly after ALD and anneal but remains below the initial value (about 91% recovery, around 1630 cm2/(V s)), indicating incomplete reversibility.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_20_figure_14.jpg","caption":"Figure 14. Atomic percentage of various elements measured by XPS after different etching conditions: (orange) unetched AlN sample, (green) after 50 cycles of  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3$ , (blue) after 50 cycles of  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3 / \\mathrm{H}_2\\mathrm{O}$ , and (red) after 50 cycles of  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3$  ending with a  $\\mathrm{H}_2\\mathrm{O}$  exposure for  $5\\mathrm{~s~}$ .","id":"test/atomic-layer-etching/experimental-usecase/20/figure_14","sample_id":"atomic-layer-etching/experimental-usecase/20/figure_14","subset":"grouped-bar-chart","split":"test","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_14.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/Thermal Atomic Layer Etching of Aluminum Nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:img_caption"},"width":658,"height":547,"image_format":"jpeg","image_sha256":"7faed0e08c7ea40f46143502bdfc0fc330f4e3e089903f02c2ca11d904e7349e","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":3,"width":653,"height":541}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart compares the atomic percentage of various elements (Al, N, O, Cl, F) across four different conditions: AlN Survey, 50 cycles XeF₂/BCl₃, 50 cycles XeF₂/BCl₃/H₂O, and 50 cycles XeF₂/BCl₃ End w/ H₂O. The atomic % decreases from Al to F for each condition.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Element | AlN Survey | 50 cycles XeF₂/BCl₃ | 50 cycles XeF₂/BCl₃/H₂O | 50 cycles XeF₂/BCl₃ End w/ H₂O |\\n| --- | --- | --- | --- | --- |\\n| Al | 49 | 50 | 38 | 47 |\\n| N | 24 | 26 | 22 | 36 |\\n| O | 26 | 6 | 25 | 13 |\\n| Cl | - | 10 | 11 | 1 |\\n| F | - | 5 | 4 | 3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The unreacted AlN sample, after 50 cycles of XeF2/BCl3, after 50 cycles of XeF2/BCl3/H2O and after 50 cycles of XeF2/BCl3 ending with a H2O exposure.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After 50 cycles of XeF2/BCl3, the atomic percentage increases slightly, but then decreases with the addition of water. It increases again after 50 cycles of XeF2/BCl2 end with H2O.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The lowest oxygen atomic percentage can be observed after 50 cycles of XeF2/BCl3.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The water step at the end of the cycle reduces the atomic percentage of F, but not that of Cl.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_experimental-usecase_41_fig_7.jpg","caption":"FIG. 7. Atomic fraction of a TiN surface after the oxidation, fluorination, and removal steps of the TiN ALE process. The oxidation step was performed using  $\\Omega_{2}$  at 5 SCCM, 0.05 Torr,  $100^{\\circ}\\mathrm{C}$ , and  $15\\mathrm{W}$ , and the fluorination step was performed using  $\\mathrm{CF_4}$  at 2 SCCM, 0.03 Torr,  $100^{\\circ}\\mathrm{C}$ , and  $15\\mathrm{W}$ . At the removal step, the temperature was increased to  $150^{\\circ}\\mathrm{C}$  with an Ar of 50 SCCM for  $30\\mathrm{s}$  without RF power.","id":"test/atomic-layer-etching/experimental-usecase/41/fig_7","sample_id":"atomic-layer-etching/experimental-usecase/41/fig_7","subset":"grouped-bar-chart","split":"test","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/images/fig_7.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/Plasma atomic layer etching for titanium nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"41","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":539,"height":411,"image_format":"jpeg","image_sha256":"91c8ec2ab0af965be3af0cae2a163c12b3c2eddc2073da01bd1c1e06abb5e8d3","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":4,"y":2,"width":530,"height":406}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"XPS-derived atomic composition of a TiN surface tracked through a full thermal ALE cycle. Oxidation strongly increases surface oxygen (~40 at.%), indicating formation of an oxide layer. Fluorination introduces fluorine (~13 at.%) while reducing oxygen, consistent with formation of a fluorinated surface layer. The final thermal removal step nearly eliminates fluorine (~1 at.%) and restores a TiN-like surface composition with recovered Ti and N fractions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Process step        | Ti (%) | N (%) | O (%) | F (%) | C (%) |\\n|---------------------|--------|-------|-------|-------|-------|\\n| Reference (TiN)     | ~33    | ~32   | ~19   | ~0    | ~17   |\\n| Oxidation           | ~27    | ~18   | ~40   | ~0    | ~15   |\\n| Fluorination        | ~27    | ~19   | ~20   | ~13   | ~20   |\\n| Removal             | ~34    | ~34   | ~19   | ~1    | ~12   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The auto summary correctly listed the elements and processing stages but did not interpret the chemical evolution of the surface. I replaced it with a concise, mechanism-aware description that links composition changes to oxidation, fluorination, and fluorine removal.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After the removal step, the nitrogen fraction (~34 at.%) slightly exceeds that of the reference surface (~32 at.%). This indicates effective restoration of a TiN-like surface after elimination of the modified fluorinated layer.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The fluorinated surface species formed during CF₄ exposure are thermally less stable than bulk TiN. Mild heating under inert Ar promotes desorption or decomposition of these fluorine-containing species, as evidenced by the sharp reduction in F atomic fraction, while leaving the underlying TiN largely intact.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
+{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_33_figure_5.jpg","caption":"Figure 5. Mobility of graphene determined by Hall measurements, after transfer, after plasma treatment, after 100 cycles  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD, and after  $400^{\\circ}\\mathrm{C}$  anneal for pristine,  $\\mathrm{O}_2$ , and  $\\mathrm{H}_2$  plasma treated graphene.","id":"test/atomic-layer-deposition/simulation-usecase/33/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/33/figure_5","subset":"grouped-bar-chart","split":"test","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Grouped bar chart of Hall mobility for graphene samples after transfer, after plasma treatment, after 100 cycles Al2O3 ALD, and after a 400 C anneal, comparing pristine graphene to 30 s O2 plasma and 35 s H2 plasma conditions. Plasma treatment strongly reduces mobility (about 195 cm2/(V s) for O2 and 467 cm2/(V s) for H2), while Al2O3 ALD and subsequent anneal recover mobility, with the H2 plasma sample reaching about 2190 cm2/(V s) after the 400 C anneal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample | Pristine (cm²/(V·s)) | Plasma Treatment (cm²/(V·s)) | Al₂O₃ ALD (cm²/(V·s)) | 400°C Anneal (cm²/(V·s)) |\\n|---|---|---|---|---|\\n| Pristine | 1300 | X | 1500 | 1850 |\\n| O₂ Plasma 30s | 1800 | 180 | 1400 | 1620 |\\n| H₂ Plasma 35s | 1450 | 450 | 1480 | 2200 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent the Hall mobility measured after graphene transfer, before any plasma treatment, for each sample set.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both plasmas reduce mobility, but O2 plasma causes a larger drop, to about 195 cm2/(V s), while H2 plasma reduces it to about 467 cm2/(V s). After 100 Al2O3 ALD cycles, the O2 treated sample partially recovers to about 1390 cm2/(V s), while the H2 treated sample recovers to about 1470 cm2/(V s), roughly returning to its initial value. After the 400 C anneal, the O2 treated sample reaches about 1630 cm2/(V s), whereas the H2 treated sample improves further to about 2190 cm2/(V s).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The paper attributes the mobility drop to plasma induced conversion of sp2 carbon toward sp3 bonding and the formation of out of plane bonds. These added bonds and associated functional groups act as scattering centers for charge carriers, which reduces mobility. This interpretation is consistent with the XPS and Raman evidence of increased sp3 character after plasma exposure.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"H2 plasma functionalization plus Al2O3 ALD plus 400 C anneal: mobility recovers to the initial value after ALD and then increases to about 152% of the original value after anneal (about 2190 cm2/(V s)).\\n\\nO2 plasma functionalization plus Al2O3 ALD plus 400 C anneal: mobility improves strongly after ALD and anneal but remains below the initial value (about 91% recovery, around 1630 cm2/(V s)), indicating incomplete reversibility.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":642,"height":347}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/images/figure_5.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/33/René H. J. Vervuurt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:img_caption"},"width":642,"height":347,"image_format":"jpeg","image_sha256":"5b6f3bd42626024de78ac94ff57ea874c1b05843a64fe9984efba26577688fe2","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_20_figure_14.jpg","caption":"Figure 14. Atomic percentage of various elements measured by XPS after different etching conditions: (orange) unetched AlN sample, (green) after 50 cycles of  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3$ , (blue) after 50 cycles of  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3 / \\mathrm{H}_2\\mathrm{O}$ , and (red) after 50 cycles of  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3$  ending with a  $\\mathrm{H}_2\\mathrm{O}$  exposure for  $5\\mathrm{~s~}$ .","id":"test/atomic-layer-etching/experimental-usecase/20/figure_14","sample_id":"atomic-layer-etching/experimental-usecase/20/figure_14","subset":"grouped-bar-chart","split":"test","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart compares the atomic percentage of various elements (Al, N, O, Cl, F) across four different conditions: AlN Survey, 50 cycles XeF₂/BCl₃, 50 cycles XeF₂/BCl₃/H₂O, and 50 cycles XeF₂/BCl₃ End w/ H₂O. The atomic % decreases from Al to F for each condition.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Element | AlN Survey | 50 cycles XeF₂/BCl₃ | 50 cycles XeF₂/BCl₃/H₂O | 50 cycles XeF₂/BCl₃ End w/ H₂O |\\n| --- | --- | --- | --- | --- |\\n| Al | 49 | 50 | 38 | 47 |\\n| N | 24 | 26 | 22 | 36 |\\n| O | 26 | 6 | 25 | 13 |\\n| Cl | - | 10 | 11 | 1 |\\n| F | - | 5 | 4 | 3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The unreacted AlN sample, after 50 cycles of XeF2/BCl3, after 50 cycles of XeF2/BCl3/H2O and after 50 cycles of XeF2/BCl3 ending with a H2O exposure.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After 50 cycles of XeF2/BCl3, the atomic percentage increases slightly, but then decreases with the addition of water. It increases again after 50 cycles of XeF2/BCl2 end with H2O.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The lowest oxygen atomic percentage can be observed after 50 cycles of XeF2/BCl3.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The water step at the end of the cycle reduces the atomic percentage of F, but not that of Cl.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":653,"height":541}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_14.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/Thermal Atomic Layer Etching of Aluminum Nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:img_caption"},"width":658,"height":547,"image_format":"jpeg","image_sha256":"7faed0e08c7ea40f46143502bdfc0fc330f4e3e089903f02c2ca11d904e7349e","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_experimental-usecase_41_fig_7.jpg","caption":"FIG. 7. Atomic fraction of a TiN surface after the oxidation, fluorination, and removal steps of the TiN ALE process. The oxidation step was performed using  $\\Omega_{2}$  at 5 SCCM, 0.05 Torr,  $100^{\\circ}\\mathrm{C}$ , and  $15\\mathrm{W}$ , and the fluorination step was performed using  $\\mathrm{CF_4}$  at 2 SCCM, 0.03 Torr,  $100^{\\circ}\\mathrm{C}$ , and  $15\\mathrm{W}$ . At the removal step, the temperature was increased to  $150^{\\circ}\\mathrm{C}$  with an Ar of 50 SCCM for  $30\\mathrm{s}$  without RF power.","id":"test/atomic-layer-etching/experimental-usecase/41/fig_7","sample_id":"atomic-layer-etching/experimental-usecase/41/fig_7","subset":"grouped-bar-chart","split":"test","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"XPS-derived atomic composition of a TiN surface tracked through a full thermal ALE cycle. Oxidation strongly increases surface oxygen (~40 at.%), indicating formation of an oxide layer. Fluorination introduces fluorine (~13 at.%) while reducing oxygen, consistent with formation of a fluorinated surface layer. The final thermal removal step nearly eliminates fluorine (~1 at.%) and restores a TiN-like surface composition with recovered Ti and N fractions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Process step        | Ti (%) | N (%) | O (%) | F (%) | C (%) |\\n|---------------------|--------|-------|-------|-------|-------|\\n| Reference (TiN)     | ~33    | ~32   | ~19   | ~0    | ~17   |\\n| Oxidation           | ~27    | ~18   | ~40   | ~0    | ~15   |\\n| Fluorination        | ~27    | ~19   | ~20   | ~13   | ~20   |\\n| Removal             | ~34    | ~34   | ~19   | ~1    | ~12   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The auto summary correctly listed the elements and processing stages but did not interpret the chemical evolution of the surface. I replaced it with a concise, mechanism-aware description that links composition changes to oxidation, fluorination, and fluorine removal.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After the removal step, the nitrogen fraction (~34 at.%) slightly exceeds that of the reference surface (~32 at.%). This indicates effective restoration of a TiN-like surface after elimination of the modified fluorinated layer.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The fluorinated surface species formed during CF₄ exposure are thermally less stable than bulk TiN. Mild heating under inert Ar promotes desorption or decomposition of these fluorine-containing species, as evidenced by the sharp reduction in F atomic fraction, while leaving the underlying TiN largely intact.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":4,"y":2,"width":530,"height":406}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/images/fig_7.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/41/Plasma atomic layer etching for titanium nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"41","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":539,"height":411,"image_format":"jpeg","image_sha256":"91c8ec2ab0af965be3af0cae2a163c12b3c2eddc2073da01bd1c1e06abb5e8d3","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/grouped-bar-chart/train/metadata.jsonl

@@ -1,6 +1,6 @@
-{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_28_figure_3.jpg","caption":"Figure 3. (a) Growth rates of the GeSe films grown via the ALD-DFM and conventional ALD process. (b) Growth rates according to the pulse pressure adjusted through the metering valve. (c) Model for the film growth during the ALD-DFM process for the efficient removal of the physisorbed molecules.","id":"train/atomic-layer-deposition/experimental-usecase/28/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/28/figure_3","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"reaction scheme"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/Woohyun Kim et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":1005,"height":1811,"image_format":"jpeg","image_sha256":"0143cf2631edf240b330c458d5ae98a3be6a87e6f5ecf1b9789ec9167a788c59","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":5,"y":56,"width":473,"height":401},{"panel_id":"b","x":535,"y":62,"width":450,"height":398},{"panel_id":"c","x":79,"y":506,"width":926,"height":1299}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_57_figure_2.jpg","caption":"Figure 2. Variation of contribution of entropy relative to the empty surface. Values are in  $\\mathrm{kJ}\\mathrm{mol}^{-1}$  over the selected reaction  $\\mathrm{A01}\\rightarrow \\mathrm{A02}$ $\\rightarrow \\mathrm{A03TS}\\rightarrow \\mathrm{A04}$ . See Figures 3 and S2 for details.","id":"train/atomic-layer-deposition/simulation-usecase/57/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/57/figure_2","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/C. Goehry et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"57","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":545,"height":411,"image_format":"jpeg","image_sha256":"8f0632233faedf5c28b12993adc6777cde7aca967a1d2352df7640367722a36c","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":4,"width":540,"height":403}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart displays the entropy contribution in kJ·mol⁻¹ for different categories (A02, A03TS, A04) across various subcategories (N1, N3, N4, N5, Nf2, Nf3, Nf4). Each category has multiple subcategories represented by different patterns. The A04 category shows the lowest energy contribution, against A03TS which has the highest values.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Entropy contribution (kJ·mol⁻¹) | A02 | A03TS | A04 |\\n|---|---|---|---|\\n| N1 | 105 | 155 | 18 |\\n| N3 | 110 | 121 | 30 |\\n| N4 | 110 | 122 | 22 |\\n| N5 | 118 | - | 32 |\\n| Nf2 | 98 | 113 | 37 |\\n| Nf3 | - | 125 | 32 |\\n| Nf4 | 111 | - | 22 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The N1, N3, N4, N5, Nf2, Nf3 and Nf4 clusters were used to determine the entropy contribution.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The entropy contribution increases for all clusters from A01 to A03TS, but decreases for all clusters at A04.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The A04 reaction exhibits the lowest energy contribution of all the clusters.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Overall, the entropy contribution along the reactions is very similar for both cases involving frozen and non-frozen atoms.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_experimental-usecase_30_fig_5.jpg","caption":"Fig. 5. Results of XPS analysis of AlGaN surface. (a) Initial, (b)  $\\mathrm{Cl}_2$  ALE, and (c)  $\\mathrm{BCl}_3 / \\mathrm{Cl}_2$  ALE.","id":"train/atomic-layer-etching/experimental-usecase/30/fig_5","sample_id":"atomic-layer-etching/experimental-usecase/30/fig_5","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/Atomic layer etching of GaN and AlGaN.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"30","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":547,"height":334,"image_format":"jpeg","image_sha256":"f997f8aa84ad3c5556e000daeb99fb609168b3ad1234674c6eb490a7a463c5d7","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":2,"width":543,"height":331}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart shows XPS atom percent for the initial AlGaN surface, after Cl2 ALE, and after BCl3 and Cl2 ALE.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample | Ga (Atom %) | N (Atom %) | Al (Atom %) | O (Atom %) |\\n|---|---|---|---|---|\\n| (a) Initial | 50 | 45 | 5 | 3 |\\n| (b) Cl₂ | 35 | 42 | 5 | 18 |\\n| (c) BCl₃/Cl₂ | 48 | 42 | 5 | 5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ga has the highest atom percent in (a).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"O is very low in (a), increases strongly in (b) after Cl2 ALE, then decreases in (c) after BCl3 and Cl2 ALE to a level closer to the initial condition.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It suggests the surface remains GaN like or Ga rich and N rich at the top surface within the XPS sampling depth, rather than being dominated by an oxide or a thick overlayer of other species.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Cl2 process may leave a surface that is more prone to post process oxidation before XPS measurement.\\n\\nThe BCl3 and Cl2 process may produce a surface condition that results in less oxygen containing residue or less surface oxidation when exposed to air.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_26_fig_8.jpg","caption":"FIG. 8. Dominant desorbed species during the desorption steps for different  $\\mathrm{Ar^{+}}$  ion energies in the thin (a) and thick (b) Cl layer deposition cases. Reflected or desorbed Ar ions or atoms are not counted.","id":"train/atomic-layer-etching/simulation-usecase/26/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/26/fig_8","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"},{"panel_id":"b","label":"grouped bar chart"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/Surface damage formation during atomic layer etching of silicon with chlorine adsorption.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:img_caption"},"width":1339,"height":525,"image_format":"jpeg","image_sha256":"d217d6886735661e16f4ca46320e6b24c35934f9488f79ff32fb959593ca38ea","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":13,"y":15,"width":629,"height":470},{"panel_id":"b","x":699,"y":15,"width":640,"height":461}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the percentage of Si/Cl based desorbed species  for thin Cl layer deposition at different electron energies (20 eV, 30 eV, 40 eV, 50 eV, 60 eV).\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the percentage of Si/Cl based desorbed species  for thick Cl layer deposition at different electron energies (20 eV, 30 eV, 40 eV, 50 eV, 60 eV).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Desorbed Species | Cl | Si | SiCl | SiCl2 |\\n|---|---|---|---|---|\\n| 20 eV | 35% | 38% | 21% | 6% |\\n| 30 eV | 30% | 50% | 17% | 3% |\\n| 40 eV | 18% | 69% | 11% | 2% |\\n| 50 eV | 10% | 78% | 6% | 1% |\\n| 60 eV | 8% | 82% | 5% | 1% |\"},{\"panel_id\":\"b\",\"text\":\"| Desorbed Species | Cl | Si | SiCl | SiCl2 | Cl2 | SiCl3 | SiCl4 |\\n|---|---|---|---|---|---|---|---|\\n| 20 eV | 65% | 30% | 5% | 2% | 1% | 1% | 1% |\\n| 30 eV | 45% | 45% | 10% | 5% | 1% | 1% | 1% |\\n| 40 eV | 35% | 38% | 12% | 6% | 1% | 1% | 1% |\\n| 50 eV | 25% | 30% | 15% | 7% | 1% | 1% | 1% |\\n| 60 eV | 20% | 25% | 18% | 8% | 1% | 1% | 1% |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. Cl\\n, 2. Si\\n, 3. SiCl\\n, 4. SiCl2\\n, 5. Cl2\\n, 6. SiCl3\\n, 7. SiCl4\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No Ar atom or ion emitted or reflected from the surface is counted because most of the incident Arþ ions are reflected once they hit the surface.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the case of thin Cl layer deposition, only a small percentage of Cl bonded Si species such as SiCl and SiCl2 desorbed from the surface. This is most likely due to the limited availability of Cl atoms on the surface under the thin Cl layer deposition conditions. The percentage of monoatomic Si atoms as desorbed species increases with the increasing Ar+ ion energy, which indicates that physical sputtering In case of Si by Ar+ ion impact dominates at higher ion incident energy. For thick Cl deposited films, tt is seen in (b), where more Cl atoms are available for surface reactions, that desorbed species with higher Cl content, including SiCl3 and SiCl4, are more visible.\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_figure_10.jpg","caption":"Figure 10. Number distribution of etching products with regard to  $\\theta_{\\mathrm{in}}$ . The error bars indicate the standard deviations among three independent runs.","id":"train/atomic-layer-etching/simulation-usecase/33/figure_10","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_10","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/Atomistic Simulation of HF Etching Process of Amorphous.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":875,"height":416,"image_format":"jpeg","image_sha256":"c3b87b092ecfc5031d0f66907fcb1dc87fdab5651641564b120c085c5390d8ba","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":875,"height":416}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart displays the number of molecular products generated for a range of gas species—including SiFₓ (x = 1–4), N-based species (N, NH, NH₂, NH₃, N₂), and composite species (SiₓNᵧ)—as a function of incident angle from 0° to 75°. Fluorinated silicon products like SiF₃ and SiF₄ show higher formation at intermediate angles (15°–60°), while nitrogen-containing products like NH₃ and N₂ vary more moderately across the angle range. Notably, the counts for SiₓNᵧ products tend to decrease at higher angles. The variation in product distribution with angle suggests angular dependence in surface reaction dynamics, affecting both fluorination and nitridation pathways.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gas species | Si | SiF | SiF<sub>2</sub> | SiF<sub>3</sub> | SiF<sub>4</sub> | N | NH | NH<sub>2</sub> | NH<sub>3</sub> | N<sub>2</sub> | Si<sub>x</sub>N<sub>y</sub> |\\n|---|---|---|---|---|---|---|---|---|---|---|---|\\n| 0° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 15° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 30° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 45° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 60° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 75° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The number of molecules formed for each gas species varies with the angle of incidence, indicating that reaction dynamics are angle-dependent. For instance, the formation of fluorinated silicon species such as SiF₃ and SiF₄ increases notably at mid-range angles (e.g., 30°–60°), while their production is lower at steeper angles (75°) and normal incidence (0°). Similarly, the generation of nitrogen-containing species (NHₓ, N₂) appears more stable across angles, although NH₃ shows a slight decrease at high angles. Composite products like SiₓNᵧ also decline in number with increasing angle, suggesting reduced recombination at oblique incidences.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiF₄\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
-{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_figure_7.jpg","caption":"Figure 7. Number distribution of etching products with regard to  $E_{\\mathrm{in}}$ . The error bars indicate the standard deviations among three independent runs.","id":"train/atomic-layer-etching/simulation-usecase/33/figure_7","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_7","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/Atomistic Simulation of HF Etching Process of Amorphous.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":875,"height":414,"image_format":"jpeg","image_sha256":"fe5a0323ac9dad15d2da34f5b87a6e3cc4dde693e2f4aab7dd032a9cbc902200","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":875,"height":414}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This grouped bar chart illustrates how the formation of various gas-phase species evolves with increasing incident energy from 10 eV to 50 eV. Fluorinated silicon products (e.g., SiF₂, SiF₃, SiF₄) show a marked increase with energy, with SiF₄ becoming the dominant species at energies ≥30 eV. Nitrogen-containing species like NH₃ and N₂ also increase in abundance, particularly NH₃, which appears early at 10 eV and continues rising through 50 eV. Notably, the composite product SiₓNᵧ only emerges at 50 eV, indicating that higher energy input is necessary for the formation of complex bonded species. This energy-dependent profile provides insight into how plasma etching or energetic surface reactions can be tuned by controlling incident energy.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gas species | Si | SiF | SiF₂ | SiF₃ | SiF₄ | N | NH | NH₂ | NH₃ | N₂ | SiₓNy |\\n|---|---|---|---|---|---|---|---|---|---|---|---|\\n| 10 eV | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 15 | 5 | 0 |\\n| 20 eV | 0 | 0 | 0 | 0 | 5 | 0 | 0 | 0 | 30 | 10 | 0 |\\n| 30 eV | 0 | 0 | 5 | 5 | 35 | 0 | 0 | 0 | 40 | 20 | 0 |\\n| 40 eV | 0 | 0 | 10 | 5 | 40 | 0 | 0 | 0 | 45 | 25 | 0 |\\n| 50 eV | 0 | 2 | 25 | 5 | 40 | 0 | 0 | 0 | 45 | 30 | 15 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the incident energy increases from 10 eV to 50 eV, the number of fluorinated silicon species rises substantially, with SiF₄ becoming the most abundant at higher energies. Low-energy exposures (10–20 eV) yield primarily NH₃ and N₂, while fluorinated species like SiF₂ and SiF₃ only begin to appear around 30 eV. SiₓNᵧ species, indicative of more complex reaction products, are only observed at 50 eV, suggesting that higher energy is required to break and reform bonds involving both Si and N. This trend highlights a shift from simple desorption products to more reactive, fragmented, or recombined species as energy increases.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"NH₃.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
+{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_28_figure_3.jpg","caption":"Figure 3. (a) Growth rates of the GeSe films grown via the ALD-DFM and conventional ALD process. (b) Growth rates according to the pulse pressure adjusted through the metering valve. (c) Model for the film growth during the ALD-DFM process for the efficient removal of the physisorbed molecules.","id":"train/atomic-layer-deposition/experimental-usecase/28/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/28/figure_3","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"reaction scheme"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":5,"y":56,"width":473,"height":401},{"panel_id":"b","x":535,"y":62,"width":450,"height":398},{"panel_id":"c","x":79,"y":506,"width":926,"height":1299}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/28/Woohyun Kim et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":1005,"height":1811,"image_format":"jpeg","image_sha256":"0143cf2631edf240b330c458d5ae98a3be6a87e6f5ecf1b9789ec9167a788c59","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_57_figure_2.jpg","caption":"Figure 2. Variation of contribution of entropy relative to the empty surface. Values are in  $\\mathrm{kJ}\\mathrm{mol}^{-1}$  over the selected reaction  $\\mathrm{A01}\\rightarrow \\mathrm{A02}$ $\\rightarrow \\mathrm{A03TS}\\rightarrow \\mathrm{A04}$ . See Figures 3 and S2 for details.","id":"train/atomic-layer-deposition/simulation-usecase/57/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/57/figure_2","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart displays the entropy contribution in kJ·mol⁻¹ for different categories (A02, A03TS, A04) across various subcategories (N1, N3, N4, N5, Nf2, Nf3, Nf4). Each category has multiple subcategories represented by different patterns. The A04 category shows the lowest energy contribution, against A03TS which has the highest values.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Entropy contribution (kJ·mol⁻¹) | A02 | A03TS | A04 |\\n|---|---|---|---|\\n| N1 | 105 | 155 | 18 |\\n| N3 | 110 | 121 | 30 |\\n| N4 | 110 | 122 | 22 |\\n| N5 | 118 | - | 32 |\\n| Nf2 | 98 | 113 | 37 |\\n| Nf3 | - | 125 | 32 |\\n| Nf4 | 111 | - | 22 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The N1, N3, N4, N5, Nf2, Nf3 and Nf4 clusters were used to determine the entropy contribution.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The entropy contribution increases for all clusters from A01 to A03TS, but decreases for all clusters at A04.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The A04 reaction exhibits the lowest energy contribution of all the clusters.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Overall, the entropy contribution along the reactions is very similar for both cases involving frozen and non-frozen atoms.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":4,"width":540,"height":403}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/57/C. Goehry et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"57","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":545,"height":411,"image_format":"jpeg","image_sha256":"8f0632233faedf5c28b12993adc6777cde7aca967a1d2352df7640367722a36c","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_experimental-usecase_30_fig_5.jpg","caption":"Fig. 5. Results of XPS analysis of AlGaN surface. (a) Initial, (b)  $\\mathrm{Cl}_2$  ALE, and (c)  $\\mathrm{BCl}_3 / \\mathrm{Cl}_2$  ALE.","id":"train/atomic-layer-etching/experimental-usecase/30/fig_5","sample_id":"atomic-layer-etching/experimental-usecase/30/fig_5","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart shows XPS atom percent for the initial AlGaN surface, after Cl2 ALE, and after BCl3 and Cl2 ALE.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample | Ga (Atom %) | N (Atom %) | Al (Atom %) | O (Atom %) |\\n|---|---|---|---|---|\\n| (a) Initial | 50 | 45 | 5 | 3 |\\n| (b) Cl₂ | 35 | 42 | 5 | 18 |\\n| (c) BCl₃/Cl₂ | 48 | 42 | 5 | 5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ga has the highest atom percent in (a).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"O is very low in (a), increases strongly in (b) after Cl2 ALE, then decreases in (c) after BCl3 and Cl2 ALE to a level closer to the initial condition.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It suggests the surface remains GaN like or Ga rich and N rich at the top surface within the XPS sampling depth, rather than being dominated by an oxide or a thick overlayer of other species.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Cl2 process may leave a surface that is more prone to post process oxidation before XPS measurement.\\n\\nThe BCl3 and Cl2 process may produce a surface condition that results in less oxygen containing residue or less surface oxidation when exposed to air.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":543,"height":331}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/30/Atomic layer etching of GaN and AlGaN.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"30","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":547,"height":334,"image_format":"jpeg","image_sha256":"f997f8aa84ad3c5556e000daeb99fb609168b3ad1234674c6eb490a7a463c5d7","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_26_fig_8.jpg","caption":"FIG. 8. Dominant desorbed species during the desorption steps for different  $\\mathrm{Ar^{+}}$  ion energies in the thin (a) and thick (b) Cl layer deposition cases. Reflected or desorbed Ar ions or atoms are not counted.","id":"train/atomic-layer-etching/simulation-usecase/26/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/26/fig_8","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"},{"panel_id":"b","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the percentage of Si/Cl based desorbed species  for thin Cl layer deposition at different electron energies (20 eV, 30 eV, 40 eV, 50 eV, 60 eV).\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the percentage of Si/Cl based desorbed species  for thick Cl layer deposition at different electron energies (20 eV, 30 eV, 40 eV, 50 eV, 60 eV).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Desorbed Species | Cl | Si | SiCl | SiCl2 |\\n|---|---|---|---|---|\\n| 20 eV | 35% | 38% | 21% | 6% |\\n| 30 eV | 30% | 50% | 17% | 3% |\\n| 40 eV | 18% | 69% | 11% | 2% |\\n| 50 eV | 10% | 78% | 6% | 1% |\\n| 60 eV | 8% | 82% | 5% | 1% |\"},{\"panel_id\":\"b\",\"text\":\"| Desorbed Species | Cl | Si | SiCl | SiCl2 | Cl2 | SiCl3 | SiCl4 |\\n|---|---|---|---|---|---|---|---|\\n| 20 eV | 65% | 30% | 5% | 2% | 1% | 1% | 1% |\\n| 30 eV | 45% | 45% | 10% | 5% | 1% | 1% | 1% |\\n| 40 eV | 35% | 38% | 12% | 6% | 1% | 1% | 1% |\\n| 50 eV | 25% | 30% | 15% | 7% | 1% | 1% | 1% |\\n| 60 eV | 20% | 25% | 18% | 8% | 1% | 1% | 1% |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. Cl\\n, 2. Si\\n, 3. SiCl\\n, 4. SiCl2\\n, 5. Cl2\\n, 6. SiCl3\\n, 7. SiCl4\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No Ar atom or ion emitted or reflected from the surface is counted because most of the incident Arþ ions are reflected once they hit the surface.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the case of thin Cl layer deposition, only a small percentage of Cl bonded Si species such as SiCl and SiCl2 desorbed from the surface. This is most likely due to the limited availability of Cl atoms on the surface under the thin Cl layer deposition conditions. The percentage of monoatomic Si atoms as desorbed species increases with the increasing Ar+ ion energy, which indicates that physical sputtering In case of Si by Ar+ ion impact dominates at higher ion incident energy. For thick Cl deposited films, tt is seen in (b), where more Cl atoms are available for surface reactions, that desorbed species with higher Cl content, including SiCl3 and SiCl4, are more visible.\"}]}]","bbox":[{"panel_id":"a","x":13,"y":15,"width":629,"height":470},{"panel_id":"b","x":699,"y":15,"width":640,"height":461}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/Surface damage formation during atomic layer etching of silicon with chlorine adsorption.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:img_caption"},"width":1339,"height":525,"image_format":"jpeg","image_sha256":"d217d6886735661e16f4ca46320e6b24c35934f9488f79ff32fb959593ca38ea","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_figure_10.jpg","caption":"Figure 10. Number distribution of etching products with regard to  $\\theta_{\\mathrm{in}}$ . The error bars indicate the standard deviations among three independent runs.","id":"train/atomic-layer-etching/simulation-usecase/33/figure_10","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_10","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart displays the number of molecular products generated for a range of gas species—including SiFₓ (x = 1–4), N-based species (N, NH, NH₂, NH₃, N₂), and composite species (SiₓNᵧ)—as a function of incident angle from 0° to 75°. Fluorinated silicon products like SiF₃ and SiF₄ show higher formation at intermediate angles (15°–60°), while nitrogen-containing products like NH₃ and N₂ vary more moderately across the angle range. Notably, the counts for SiₓNᵧ products tend to decrease at higher angles. The variation in product distribution with angle suggests angular dependence in surface reaction dynamics, affecting both fluorination and nitridation pathways.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gas species | Si | SiF | SiF<sub>2</sub> | SiF<sub>3</sub> | SiF<sub>4</sub> | N | NH | NH<sub>2</sub> | NH<sub>3</sub> | N<sub>2</sub> | Si<sub>x</sub>N<sub>y</sub> |\\n|---|---|---|---|---|---|---|---|---|---|---|---|\\n| 0° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 15° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 30° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 45° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 60° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\\n| 75° | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The number of molecules formed for each gas species varies with the angle of incidence, indicating that reaction dynamics are angle-dependent. For instance, the formation of fluorinated silicon species such as SiF₃ and SiF₄ increases notably at mid-range angles (e.g., 30°–60°), while their production is lower at steeper angles (75°) and normal incidence (0°). Similarly, the generation of nitrogen-containing species (NHₓ, N₂) appears more stable across angles, although NH₃ shows a slight decrease at high angles. Composite products like SiₓNᵧ also decline in number with increasing angle, suggesting reduced recombination at oblique incidences.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiF₄\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":875,"height":416}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/Atomistic Simulation of HF Etching Process of Amorphous.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":875,"height":416,"image_format":"jpeg","image_sha256":"c3b87b092ecfc5031d0f66907fcb1dc87fdab5651641564b120c085c5390d8ba","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_figure_7.jpg","caption":"Figure 7. Number distribution of etching products with regard to  $E_{\\mathrm{in}}$ . The error bars indicate the standard deviations among three independent runs.","id":"train/atomic-layer-etching/simulation-usecase/33/figure_7","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_7","subset":"grouped-bar-chart","split":"train","classification":[{"panel_id":"a","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This grouped bar chart illustrates how the formation of various gas-phase species evolves with increasing incident energy from 10 eV to 50 eV. Fluorinated silicon products (e.g., SiF₂, SiF₃, SiF₄) show a marked increase with energy, with SiF₄ becoming the dominant species at energies ≥30 eV. Nitrogen-containing species like NH₃ and N₂ also increase in abundance, particularly NH₃, which appears early at 10 eV and continues rising through 50 eV. Notably, the composite product SiₓNᵧ only emerges at 50 eV, indicating that higher energy input is necessary for the formation of complex bonded species. This energy-dependent profile provides insight into how plasma etching or energetic surface reactions can be tuned by controlling incident energy.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gas species | Si | SiF | SiF₂ | SiF₃ | SiF₄ | N | NH | NH₂ | NH₃ | N₂ | SiₓNy |\\n|---|---|---|---|---|---|---|---|---|---|---|---|\\n| 10 eV | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 15 | 5 | 0 |\\n| 20 eV | 0 | 0 | 0 | 0 | 5 | 0 | 0 | 0 | 30 | 10 | 0 |\\n| 30 eV | 0 | 0 | 5 | 5 | 35 | 0 | 0 | 0 | 40 | 20 | 0 |\\n| 40 eV | 0 | 0 | 10 | 5 | 40 | 0 | 0 | 0 | 45 | 25 | 0 |\\n| 50 eV | 0 | 2 | 25 | 5 | 40 | 0 | 0 | 0 | 45 | 30 | 15 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the incident energy increases from 10 eV to 50 eV, the number of fluorinated silicon species rises substantially, with SiF₄ becoming the most abundant at higher energies. Low-energy exposures (10–20 eV) yield primarily NH₃ and N₂, while fluorinated species like SiF₂ and SiF₃ only begin to appear around 30 eV. SiₓNᵧ species, indicative of more complex reaction products, are only observed at 50 eV, suggesting that higher energy is required to break and reform bonds involving both Si and N. This trend highlights a shift from simple desorption products to more reactive, fragmented, or recombined species as energy increases.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"NH₃.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":875,"height":414}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/Atomistic Simulation of HF Etching Process of Amorphous.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"grouped bar chart","caption_source":"content.json:image_caption"},"width":875,"height":414,"image_format":"jpeg","image_sha256":"fe5a0323ac9dad15d2da34f5b87a6e3cc4dde693e2f4aab7dd032a9cbc902200","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/multi-axis-chart/validation/metadata.jsonl

@@ -1,5 +1,5 @@
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_24_figure_5.jpg","caption":"Figure 5. Mass change versus time for  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE using sequential TMA and HF exposures at 250, 275, 300, and  $325^{\\circ}C$","id":"validation/atomic-layer-etching/experimental-usecase/24/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/24/figure_5","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/images/figure_5.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/Trimethylaluminum as the Metal Precursor for the Atomic LayerEtching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"24","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":667,"height":486,"image_format":"jpeg","image_sha256":"9722bbee3fbccb297a7aa63c36e9af6538233ba23be1ad874fca144304ed8921","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":667,"height":486}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows mass loss from Al2O3 during repeated TMA/HF cycles at four temperatures. All traces decrease roughly linearly with time, but the slope becomes steeper at higher temperatures, indicating faster etching. The 250 °C curve shows the smallest mass loss while the 325 °C trace shows the largest, consistent with thermally enhanced reaction rates. The right-hand axis converts the mass loss to an equivalent thickness removed.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Time (s) | Mass change (ng/cm²) |\\n|------------------|----------|-----------------------|\\n| 250              | 0        | ~0                    |\\n| 250              | 4000     | ~-500                 |\\n| 250              | 8000     | ~-900                 |\\n| 275              | 0        | ~0                    |\\n| 275              | 4000     | ~-800                 |\\n| 275              | 8000     | ~-1600                |\\n| 300              | 0        | ~0                    |\\n| 300              | 4000     | ~-1100                |\\n| 300              | 8000     | ~-2200                |\\n| 325              | 0        | ~0                    |\\n| 325              | 4000     | ~-1500                |\\n| 325              | 8000     | ~-3000                |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2.0 s of TMA exposure, 30.0 s of N2 purge, 1.0 s of HF exposure, and 30.0 s of N2 purge.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The mass change per cycle increases with higher temperature values.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"325 °C\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The information shown in the figure makes it possible to determine that the temperature of the ALE cycle affects the rates of the process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process with the highest etch rate is the one at 325 degree Celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"This would take 6000 seconds.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Industrial players usually aim for the highest throughput. This would be achieved with the process at 325 degrees Celcius.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The process at 250 degrees Celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At each temperature, the mass decreases approximately linearly with time. The lines do not show curvature or abrupt changes over the plotted interval. This indicates a steady accumulation of mass loss during the process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The 250 °C and 275 °C curves remain closest together.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Mass change on the left y-axis directly represents cumulative material removal.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The thickness axis provides an alternative way to interpret the same trend in length units. It allows readers to estimate film removal depth directly from the plot. This is useful for process planning without additional conversion.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Etch rate increases approximately 5-fold from 0.14 Å/cycle at 250°C to 0.75 Å/cycle at 325°C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The dual axes allow direct comparison of the raw QCM measurement (mass change in ng/cm²) with the more practically relevant thickness change (in Å). The thickness conversion uses the known Al₂O₃ density of 3.1 g/cm³, enabling readers to immediately understand the etch depth without manual calculation. This is particularly useful for process development where target etch depths are typically specified in thickness units.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Linear behavior confirms that the etch rate remains constant throughout 100 cycles at each temperature, indicating stable, reproducible etching.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 75 Å (based on 0.75 Å/cycle × 100 cycles), as shown on the right y-axis.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"TMA dose (2 s) → 2. Purge (30 s) → 3. HF dose (1 s) → 4. Purge (30 s).\\nThe notation defines the precise timing: reactant dose length and purge length. This controls reactant exposure and byproduct removal, ensuring self-limiting surface reactions and preventing gas-phase mixing.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"325°C is ruled out. Using 300°C (the highest allowed) maximizes the etch rate within the constraint, reducing process time. The trade-off is operating near the budget limit, which may increase thermal stress compared to 250°C, but offers significantly faster throughput.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing temperature increases the slope (etch rate). This follows the Arrhenius law, where higher thermal energy accelerates the surface reaction kinetics of both the fluorination (HF) and ligand-exchange (TMA) steps, leading to more material removed per unit time.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"If CVD-like growth occurred (e.g., TMA and HF mixing in the gas phase), the net mass change would likely turn positive (mass gain) or the etch rate would slow down drastically, causing the curves to flatten or curve upward. The fact that the lines remain strictly negative and linear confirms that the process is operating in a pure etching mode with no competing deposition artifacts.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"These MCPCs correspond to etch rates hence a mass loss is a direct indicator of thickness decreasing and vice versa.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch rates are\\n1. 0.14 Å/cycle at 250 °C \\n2.  0.75 Å/cycle at 325 °C\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The determination of these etch rates is based on the Al2O3 ALD film density of 3.1 g/cm3, the MCPC, the number of ALE cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"100\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The MCPC increases with temperature from -4.2 ng/(cm^2 cycle) at 250 °C to -23.3 ng/(cm2 cycle) at 325 °C. This corresponds to etch rates of 0.14 Å/cycle at 250 °C to 0.75 Å/cycle at 325 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For this study, a quartz crystal microbalance (QCM) was used, which yields mass change in ng/cm^2. Before the experiments, the density of the Al2O3 substrate was determined to be 3.1 g/cm^3. Dividing mass change (per area) by the density in this case results in the material thickness that was etched per cycle, expressed in Å.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2 s TMA exposure, 30 s purge, 1 s HF exposure, 30 s purge.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Temperature\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"325 °C, 300 °C, 275 °C, 250 °C\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Since one cycle is about 63 s, 100 cycles correspond to roughly 6300 s, so you can read the expected thickness removal near that time on the right-hand axis. At 325 °C, the curve reaches roughly 70 Å (≥6 nm). All lower-temperature curves show less than 6 nm after 100 cycles. Therefore, 325 °C is the best choice because it meets the ≥6 nm target within 100 cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300 °C\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Higher temperatures produce steeper mass-loss slopes. The 250 °C curve decreases slowly, while the 300 °C and 325 °C curves show progressively faster loss. This indicates that the etching reaction accelerates significantly with temperature.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TMA / purge / HF / purge (“2–30–1–30”).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"250 °C, 275 °C, 300 °C, and 325 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"MCPC increases from −4.2 ng/(cm² cycle) at 250 °C to −23.3 ng/(cm² cycle) at 325 °C.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Consistent removal of Al₂O₃ per cycle, Predictable thickness reduction, Minimal fluctuations between cycles\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Temperature control is crucial because it directly determines the etch rate. By selecting appropriate temperatures, practitioners can achieve precise and uniform removal of Al₂O₃, tailor etch depths, and ensure consistent results across multiple samples or production batches. Proper temperature management is therefore essential for reproducible, high-quality ALE in both research and industrial applications.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_2.jpg","caption":"Figure 2. Relative atomic percentages among In, Ga, and As in InGaAs, and the atomic percentage of Cl on the InGaAs surface measured by XPS as a function of Cl adsorption time from 2.5 to  $20\\mathrm{s}$ . The power to the ICP ion source was  $200\\mathrm{W}$  at  $\\mathrm{Cl}_2$ $1.0~\\mathrm{mTorr}$  of process chamber pressure. The grid voltages to the ICP ion gun were maintained at  $-10\\mathrm{V}$  for the first grid and  $-20\\mathrm{V}$  for the second grid.","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_2","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_2.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":658,"height":461,"image_format":"jpeg","image_sha256":"469cc7db3f29fb94c823b9b47ee159e5144071ec780e5b3658bd95f74b229200","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":653,"height":458}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the changes in In:Ga:As atomic percentage and Cl atomic percentage on the surface over time during Cl adsorption.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cl Adsorption Time (sec) | In Atomic Percentage (%) | Ga Atomic Percentage (%) |As Atomic Percentage (%) |Cl Atomic Percentage (%) |\\n|---|---|---|---|---|\\n| Pristine | 45 |36|15| 0 |\\n| 2.5 | 48 | 39|14|7 |\\n| 5 | 49 |35|15|12 |\\n| 10 | 49 |35|15|16 |\\n| 20 | 46 | 35|16 |16|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Eventhough there is a slight increase in indium content for short times, the overal ratio remains constant, so it does not affect the ratios.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3.2:2.4:1\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The chlorine is only added to the surface. If the full surface is covered in chlorine, there cannot be any additional chlorine added.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"15 seconds\"}]}]"}
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_4.jpg","caption":"Figure 4. Sputter etch depth of InGaAs itself during the  $\\mathrm{Ar^{+}}$  ion gun operation as a function of (a) different first grid voltages for the  $\\mathrm{Ar^{+}}$  ion exposure time of 50 s/cycle and (b)  $\\mathrm{Ar^{+}}$  ion exposure time for the first grid voltages of  $+10$  and  $+30\\mathrm{V}$ . The sputtering was cyclic processed for 100 cycles without adsorption of chlorine during the chlorine adsorption step. The power of the ICP ion gun for the  $\\mathrm{Ar^{+}}$  ion was maintained at  $200\\mathrm{W}$  at the process chamber pressure of  $3.0\\mathrm{mTorr}$  Ar. The second grid voltage to the ICP ion gun was maintained at  $-100\\mathrm{V}$  for a directional ion beam while the third grid voltage was grounded.","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_4","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_4.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":653,"height":444,"image_format":"jpeg","image_sha256":"965e6fddb0022eb82ebbe0a3b65d0f8de71e8855f0066e243349c4384beacdf8","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":3,"width":649,"height":445}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the etch depth and etch rate of various materials (InGaAs, Si, SiO<sub>2</sub>, HfO<sub>2</sub>, PR, ACL) as a function of Ar<sup>+</sup> ion exposure time. The etch depth increases with exposure time for InGaAs, while the etch rate saturates at around 50 seconds.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ar<sup>+</sup> Ion Exposure Time (sec/cycle) | Etch Depth (Å) InGaAs|Etch Depth (Å) Other| Etch Rate (Å/Cycle) InGaAs|Etch Rate (Å/Cycle) Other|\\n|---|---|---|---|---|\\n| 0 | 0 | 0 |0|0|\\n| 20 | 40 | 0 |0.4|0|\\n| 30 | 60 | 0|0.6 |0|\\n| 50 | 100 | 0|1.0 |0|\\n| 80 | 100 |0| 1.0 |0|\\n| 100 | 100 | 0|1.0 |0|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Only the InGaAs is etched upon the plasma exposure, so the process is selective for InGaAs.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The argon ions only remove the modified layer, which is 1.1 angstrom thick.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Instead of using a modification with chlorine, use something else, such as fluorine.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"100 cycles.\"}]}]"}
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_6.jpg","caption":"Figure 6. Etch depth  $(\\mathring{\\mathrm{A}})$  and etch rate (A/cycle) of InGaAs measured as a function of the number of ALE cycles.","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_6","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_6","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_6.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":669,"height":450,"image_format":"jpeg","image_sha256":"b0d1b89df148cef029667379e1942de01f583ff63652cfb3e76e082cc00f5c6c","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":1,"y":1,"width":666,"height":447}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the etch depth and etch rate as a function of the number of ALE cycles. The etch depth increases with the number of cycles, whereas the etch rate remains constant.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| No. of ALE Cycles | Etch Depth (Å) | Etch Rate (Å/Cycle) |\\n|---|---|---|\\n| 100 | 100 ± 5 | 1.1 ± 0.1 |\\n| 200 | 210 ± 5 | 1.1 ± 0.1 |\\n| 300 | 325 ± 5 | 1.1 ± 0.1 |\\n| 400 | 450 ± 5 | 1.1 ± 0.1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate is normalized per cycle, for ALE, this is not dependent on the number of cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The value is 1.1 angstrom/cycle. This can also be determined by determining the slope of the etch depth.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there is a region. This is most likely to occur in the initial cycles. This also holds for ALD, where it is called nucleation delay.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"44 nm.\"}]}]"}
-{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_7.jpg","caption":"Figure 7. Surface roughness and surface composition after the 100 cycles of ALE, investigated using AFM and XPS, respectively. As references, the surface roughness and surface composition of as-received un-etched InGaAs and the InGaAs etched by conventional RIE were included. For the InGaAs RIE, InGaAs was etched for 1 min using an ICP etcher operated at  $13.56\\mathrm{MHz}200\\mathrm{W}$ $-50\\mathrm{V}$  of bias voltage, and  $\\mathrm{Cl}_2$  gas pressure of  $50\\mathrm{mTorr}$","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_7","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_7","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_7.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":669,"height":425,"image_format":"jpeg","image_sha256":"fd8be89b63a523530024c7306f006c2326d06df9f46dea04d6bd4c136211d814","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":4,"y":3,"width":664,"height":423}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart displays atomic percentages of various elements (In, Ga, As, Cl, C) across three conditions (Ref., ALE, RIE) along with RMS roughness measurements.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Atomic Percentage (%) | Ref. | ALE | RIE |\\n|---|---|---|---|\\n| In | 20.6% | 29.5% | 38.0% |\\n| Ga | 43.4% | 21.8% | 18.4% |\\n| As | 6.6% | 5.3% | 2.4% |\\n| Cl | 0.0% | 0.0% | 7.8% |\\n| RMS Roughness (nm) | 4.5 | 4.8 | 5.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both for the ALE and the Ref. the percentages are roughly equal. This shows that there is no selectivity towards specific species in the film.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there is an influence, where Ga and As seem to be sputtered selectively.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ALE shows a lower RMS roughness than RIE so ALE is more gentle and the ion impact from RIE has bigger impact on the\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The roughness is not extremely low, but there has been some added processing with C incorporation so it is unclear, based on this figure whether this deposition leads to low roughness.\"}]}]"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_24_figure_5.jpg","caption":"Figure 5. Mass change versus time for  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE using sequential TMA and HF exposures at 250, 275, 300, and  $325^{\\circ}C$","id":"validation/atomic-layer-etching/experimental-usecase/24/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/24/figure_5","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows mass loss from Al2O3 during repeated TMA/HF cycles at four temperatures. All traces decrease roughly linearly with time, but the slope becomes steeper at higher temperatures, indicating faster etching. The 250 °C curve shows the smallest mass loss while the 325 °C trace shows the largest, consistent with thermally enhanced reaction rates. The right-hand axis converts the mass loss to an equivalent thickness removed.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Time (s) | Mass change (ng/cm²) |\\n|------------------|----------|-----------------------|\\n| 250              | 0        | ~0                    |\\n| 250              | 4000     | ~-500                 |\\n| 250              | 8000     | ~-900                 |\\n| 275              | 0        | ~0                    |\\n| 275              | 4000     | ~-800                 |\\n| 275              | 8000     | ~-1600                |\\n| 300              | 0        | ~0                    |\\n| 300              | 4000     | ~-1100                |\\n| 300              | 8000     | ~-2200                |\\n| 325              | 0        | ~0                    |\\n| 325              | 4000     | ~-1500                |\\n| 325              | 8000     | ~-3000                |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2.0 s of TMA exposure, 30.0 s of N2 purge, 1.0 s of HF exposure, and 30.0 s of N2 purge.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The mass change per cycle increases with higher temperature values.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"325 °C\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The information shown in the figure makes it possible to determine that the temperature of the ALE cycle affects the rates of the process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process with the highest etch rate is the one at 325 degree Celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"This would take 6000 seconds.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Industrial players usually aim for the highest throughput. This would be achieved with the process at 325 degrees Celcius.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The process at 250 degrees Celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At each temperature, the mass decreases approximately linearly with time. The lines do not show curvature or abrupt changes over the plotted interval. This indicates a steady accumulation of mass loss during the process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The 250 °C and 275 °C curves remain closest together.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Mass change on the left y-axis directly represents cumulative material removal.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The thickness axis provides an alternative way to interpret the same trend in length units. It allows readers to estimate film removal depth directly from the plot. This is useful for process planning without additional conversion.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Etch rate increases approximately 5-fold from 0.14 Å/cycle at 250°C to 0.75 Å/cycle at 325°C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The dual axes allow direct comparison of the raw QCM measurement (mass change in ng/cm²) with the more practically relevant thickness change (in Å). The thickness conversion uses the known Al₂O₃ density of 3.1 g/cm³, enabling readers to immediately understand the etch depth without manual calculation. This is particularly useful for process development where target etch depths are typically specified in thickness units.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Linear behavior confirms that the etch rate remains constant throughout 100 cycles at each temperature, indicating stable, reproducible etching.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 75 Å (based on 0.75 Å/cycle × 100 cycles), as shown on the right y-axis.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"TMA dose (2 s) → 2. Purge (30 s) → 3. HF dose (1 s) → 4. Purge (30 s).\\nThe notation defines the precise timing: reactant dose length and purge length. This controls reactant exposure and byproduct removal, ensuring self-limiting surface reactions and preventing gas-phase mixing.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"325°C is ruled out. Using 300°C (the highest allowed) maximizes the etch rate within the constraint, reducing process time. The trade-off is operating near the budget limit, which may increase thermal stress compared to 250°C, but offers significantly faster throughput.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing temperature increases the slope (etch rate). This follows the Arrhenius law, where higher thermal energy accelerates the surface reaction kinetics of both the fluorination (HF) and ligand-exchange (TMA) steps, leading to more material removed per unit time.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"If CVD-like growth occurred (e.g., TMA and HF mixing in the gas phase), the net mass change would likely turn positive (mass gain) or the etch rate would slow down drastically, causing the curves to flatten or curve upward. The fact that the lines remain strictly negative and linear confirms that the process is operating in a pure etching mode with no competing deposition artifacts.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"These MCPCs correspond to etch rates hence a mass loss is a direct indicator of thickness decreasing and vice versa.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch rates are\\n1. 0.14 Å/cycle at 250 °C \\n2.  0.75 Å/cycle at 325 °C\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The determination of these etch rates is based on the Al2O3 ALD film density of 3.1 g/cm3, the MCPC, the number of ALE cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"100\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The MCPC increases with temperature from -4.2 ng/(cm^2 cycle) at 250 °C to -23.3 ng/(cm2 cycle) at 325 °C. This corresponds to etch rates of 0.14 Å/cycle at 250 °C to 0.75 Å/cycle at 325 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For this study, a quartz crystal microbalance (QCM) was used, which yields mass change in ng/cm^2. Before the experiments, the density of the Al2O3 substrate was determined to be 3.1 g/cm^3. Dividing mass change (per area) by the density in this case results in the material thickness that was etched per cycle, expressed in Å.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2 s TMA exposure, 30 s purge, 1 s HF exposure, 30 s purge.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Temperature\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"325 °C, 300 °C, 275 °C, 250 °C\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Since one cycle is about 63 s, 100 cycles correspond to roughly 6300 s, so you can read the expected thickness removal near that time on the right-hand axis. At 325 °C, the curve reaches roughly 70 Å (≥6 nm). All lower-temperature curves show less than 6 nm after 100 cycles. Therefore, 325 °C is the best choice because it meets the ≥6 nm target within 100 cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300 °C\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Higher temperatures produce steeper mass-loss slopes. The 250 °C curve decreases slowly, while the 300 °C and 325 °C curves show progressively faster loss. This indicates that the etching reaction accelerates significantly with temperature.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TMA / purge / HF / purge (“2–30–1–30”).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"250 °C, 275 °C, 300 °C, and 325 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"MCPC increases from −4.2 ng/(cm² cycle) at 250 °C to −23.3 ng/(cm² cycle) at 325 °C.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Consistent removal of Al₂O₃ per cycle, Predictable thickness reduction, Minimal fluctuations between cycles\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Temperature control is crucial because it directly determines the etch rate. By selecting appropriate temperatures, practitioners can achieve precise and uniform removal of Al₂O₃, tailor etch depths, and ensure consistent results across multiple samples or production batches. Proper temperature management is therefore essential for reproducible, high-quality ALE in both research and industrial applications.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":667,"height":486}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/images/figure_5.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/24/Trimethylaluminum as the Metal Precursor for the Atomic LayerEtching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"24","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":667,"height":486,"image_format":"jpeg","image_sha256":"9722bbee3fbccb297a7aa63c36e9af6538233ba23be1ad874fca144304ed8921","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_2.jpg","caption":"Figure 2. Relative atomic percentages among In, Ga, and As in InGaAs, and the atomic percentage of Cl on the InGaAs surface measured by XPS as a function of Cl adsorption time from 2.5 to  $20\\mathrm{s}$ . The power to the ICP ion source was  $200\\mathrm{W}$  at  $\\mathrm{Cl}_2$ $1.0~\\mathrm{mTorr}$  of process chamber pressure. The grid voltages to the ICP ion gun were maintained at  $-10\\mathrm{V}$  for the first grid and  $-20\\mathrm{V}$  for the second grid.","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_2","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the changes in In:Ga:As atomic percentage and Cl atomic percentage on the surface over time during Cl adsorption.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cl Adsorption Time (sec) | In Atomic Percentage (%) | Ga Atomic Percentage (%) |As Atomic Percentage (%) |Cl Atomic Percentage (%) |\\n|---|---|---|---|---|\\n| Pristine | 45 |36|15| 0 |\\n| 2.5 | 48 | 39|14|7 |\\n| 5 | 49 |35|15|12 |\\n| 10 | 49 |35|15|16 |\\n| 20 | 46 | 35|16 |16|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Eventhough there is a slight increase in indium content for short times, the overal ratio remains constant, so it does not affect the ratios.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3.2:2.4:1\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The chlorine is only added to the surface. If the full surface is covered in chlorine, there cannot be any additional chlorine added.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"15 seconds\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":653,"height":458}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_2.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":658,"height":461,"image_format":"jpeg","image_sha256":"469cc7db3f29fb94c823b9b47ee159e5144071ec780e5b3658bd95f74b229200","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_4.jpg","caption":"Figure 4. Sputter etch depth of InGaAs itself during the  $\\mathrm{Ar^{+}}$  ion gun operation as a function of (a) different first grid voltages for the  $\\mathrm{Ar^{+}}$  ion exposure time of 50 s/cycle and (b)  $\\mathrm{Ar^{+}}$  ion exposure time for the first grid voltages of  $+10$  and  $+30\\mathrm{V}$ . The sputtering was cyclic processed for 100 cycles without adsorption of chlorine during the chlorine adsorption step. The power of the ICP ion gun for the  $\\mathrm{Ar^{+}}$  ion was maintained at  $200\\mathrm{W}$  at the process chamber pressure of  $3.0\\mathrm{mTorr}$  Ar. The second grid voltage to the ICP ion gun was maintained at  $-100\\mathrm{V}$  for a directional ion beam while the third grid voltage was grounded.","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_4","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the etch depth and etch rate of various materials (InGaAs, Si, SiO<sub>2</sub>, HfO<sub>2</sub>, PR, ACL) as a function of Ar<sup>+</sup> ion exposure time. The etch depth increases with exposure time for InGaAs, while the etch rate saturates at around 50 seconds.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ar<sup>+</sup> Ion Exposure Time (sec/cycle) | Etch Depth (Å) InGaAs|Etch Depth (Å) Other| Etch Rate (Å/Cycle) InGaAs|Etch Rate (Å/Cycle) Other|\\n|---|---|---|---|---|\\n| 0 | 0 | 0 |0|0|\\n| 20 | 40 | 0 |0.4|0|\\n| 30 | 60 | 0|0.6 |0|\\n| 50 | 100 | 0|1.0 |0|\\n| 80 | 100 |0| 1.0 |0|\\n| 100 | 100 | 0|1.0 |0|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Only the InGaAs is etched upon the plasma exposure, so the process is selective for InGaAs.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The argon ions only remove the modified layer, which is 1.1 angstrom thick.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Instead of using a modification with chlorine, use something else, such as fluorine.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"100 cycles.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":649,"height":445}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_4.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":653,"height":444,"image_format":"jpeg","image_sha256":"965e6fddb0022eb82ebbe0a3b65d0f8de71e8855f0066e243349c4384beacdf8","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_6.jpg","caption":"Figure 6. Etch depth  $(\\mathring{\\mathrm{A}})$  and etch rate (A/cycle) of InGaAs measured as a function of the number of ALE cycles.","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_6","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_6","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the etch depth and etch rate as a function of the number of ALE cycles. The etch depth increases with the number of cycles, whereas the etch rate remains constant.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| No. of ALE Cycles | Etch Depth (Å) | Etch Rate (Å/Cycle) |\\n|---|---|---|\\n| 100 | 100 ± 5 | 1.1 ± 0.1 |\\n| 200 | 210 ± 5 | 1.1 ± 0.1 |\\n| 300 | 325 ± 5 | 1.1 ± 0.1 |\\n| 400 | 450 ± 5 | 1.1 ± 0.1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate is normalized per cycle, for ALE, this is not dependent on the number of cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The value is 1.1 angstrom/cycle. This can also be determined by determining the slope of the etch depth.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there is a region. This is most likely to occur in the initial cycles. This also holds for ALD, where it is called nucleation delay.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"44 nm.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":666,"height":447}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_6.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":669,"height":450,"image_format":"jpeg","image_sha256":"b0d1b89df148cef029667379e1942de01f583ff63652cfb3e76e082cc00f5c6c","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/validation_atomic-layer-etching_experimental-usecase_5_figure_7.jpg","caption":"Figure 7. Surface roughness and surface composition after the 100 cycles of ALE, investigated using AFM and XPS, respectively. As references, the surface roughness and surface composition of as-received un-etched InGaAs and the InGaAs etched by conventional RIE were included. For the InGaAs RIE, InGaAs was etched for 1 min using an ICP etcher operated at  $13.56\\mathrm{MHz}200\\mathrm{W}$ $-50\\mathrm{V}$  of bias voltage, and  $\\mathrm{Cl}_2$  gas pressure of  $50\\mathrm{mTorr}$","id":"validation/atomic-layer-etching/experimental-usecase/5/figure_7","sample_id":"atomic-layer-etching/experimental-usecase/5/figure_7","subset":"multi-axis-chart","split":"validation","classification":[{"panel_id":"a","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The grouped bar chart displays atomic percentages of various elements (In, Ga, As, Cl, C) across three conditions (Ref., ALE, RIE) along with RMS roughness measurements.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Atomic Percentage (%) | Ref. | ALE | RIE |\\n|---|---|---|---|\\n| In | 20.6% | 29.5% | 38.0% |\\n| Ga | 43.4% | 21.8% | 18.4% |\\n| As | 6.6% | 5.3% | 2.4% |\\n| Cl | 0.0% | 0.0% | 7.8% |\\n| RMS Roughness (nm) | 4.5 | 4.8 | 5.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both for the ALE and the Ref. the percentages are roughly equal. This shows that there is no selectivity towards specific species in the film.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there is an influence, where Ga and As seem to be sputtered selectively.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ALE shows a lower RMS roughness than RIE so ALE is more gentle and the ion impact from RIE has bigger impact on the\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The roughness is not extremely low, but there has been some added processing with C incorporation so it is unclear, based on this figure whether this deposition leads to low roughness.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":3,"width":664,"height":423}],"source":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/images/figure_7.json","source_content":"icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/dev/atomic-layer-etching/experimental-usecase/5/Atomic layer etching of InGaAs by controlled ion beam.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multi-axis chart","caption_source":"content.json:img_caption"},"width":669,"height":425,"image_format":"jpeg","image_sha256":"fd8be89b63a523530024c7306f006c2326d06df9f46dea04d6bd4c136211d814","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/network-diagram/test/metadata.jsonl

@@ -1,2 +1,2 @@
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig2.jpg","caption":"Fig.2-Two-input-one-output FNN model with two hidden layers.","id":"test/atomic-layer-etching/simulation-usecase/11/fig2","sample_id":"atomic-layer-etching/simulation-usecase/11/fig2","subset":"network-diagram","split":"test","classification":[{"panel_id":"a","label":"network diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"network diagram","caption_source":"content.json:img_caption"},"width":614,"height":433,"image_format":"jpeg","image_sha256":"cd960a1af61e0e98fd8e0b064c79009063d12cb77c022a92be502f450b109242","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":3,"y":5,"width":604,"height":427}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A neural network architecture with input nodes, hidden layers, and an output node.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"x₁ represents temperature (K) and x₂ represents pressure (Pa) of the ALE process operating conditions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Two identical FNN architectures are trained separately for Step A (HF fluorination) and Step B (TMA etching). Each model takes temperature and pressure as inputs and predicts the corresponding half-cycle time. The models are trained on 80% of kMC simulation data and validated on the remaining 20%. Both use 50 neurons in each hidden layer and the exponential linear unit (ELU) activation function.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The exponential linear unit (ELU) activation function is used for effective gradient propagation and smoother prediction compared to other activation functions.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The FNN predicts half-cycle times in less than one second, compared to approximately 20 minutes per kMC simulation, enabling real-time process control applications.\"}]}]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_fig_9.jpg","caption":"Fig. 9. An artificial neural network architecture consisting of an input, output, and hidden layer. In this work, the input layer has  $l = 3$  inputs ( $x_{1}$ ,  $x_{2}$ , and  $x_{3}$ ), the hidden layer is defined with  $p = 20$  neurons ( $h_{j}$  for  $j = 1, 2, \\ldots, p$ ), and the output layer consists of one output ( $\\hat{y}$ ).","id":"test/atomic-layer-etching/simulation-usecase/34/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/34/fig_9","subset":"network-diagram","split":"test","classification":[{"panel_id":"a","label":"network diagram"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"network diagram","caption_source":"content.json:image_caption"},"width":983,"height":536,"image_format":"jpeg","image_sha256":"347051b21b7fbff1ec1c993b597d645af43f4bdc75e126fc5a49735660b5c36b","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":983,"height":536}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_11_fig2.jpg","caption":"Fig.2-Two-input-one-output FNN model with two hidden layers.","id":"test/atomic-layer-etching/simulation-usecase/11/fig2","sample_id":"atomic-layer-etching/simulation-usecase/11/fig2","subset":"network-diagram","split":"test","classification":[{"panel_id":"a","label":"network diagram"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A neural network architecture with input nodes, hidden layers, and an output node.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"x₁ represents temperature (K) and x₂ represents pressure (Pa) of the ALE process operating conditions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Two identical FNN architectures are trained separately for Step A (HF fluorination) and Step B (TMA etching). Each model takes temperature and pressure as inputs and predicts the corresponding half-cycle time. The models are trained on 80% of kMC simulation data and validated on the remaining 20%. Both use 50 neurons in each hidden layer and the exponential linear unit (ELU) activation function.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The exponential linear unit (ELU) activation function is used for effective gradient propagation and smoother prediction compared to other activation functions.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The FNN predicts half-cycle times in less than one second, compared to approximately 20 minutes per kMC simulation, enabling real-time process control applications.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":5,"width":604,"height":427}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/images/fig2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/11/Microscopic and data-driven modeling andoperation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"network diagram","caption_source":"content.json:img_caption"},"width":614,"height":433,"image_format":"jpeg","image_sha256":"cd960a1af61e0e98fd8e0b064c79009063d12cb77c022a92be502f450b109242","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_34_fig_9.jpg","caption":"Fig. 9. An artificial neural network architecture consisting of an input, output, and hidden layer. In this work, the input layer has  $l = 3$  inputs ( $x_{1}$ ,  $x_{2}$ , and  $x_{3}$ ), the hidden layer is defined with  $p = 20$  neurons ( $h_{j}$  for  $j = 1, 2, \\ldots, p$ ), and the output layer consists of one output ( $\\hat{y}$ ).","id":"test/atomic-layer-etching/simulation-usecase/34/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/34/fig_9","subset":"network-diagram","split":"test","classification":[{"panel_id":"a","label":"network diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":983,"height":536}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/34/Machine learning-based run-to-run control of a spatial thermal.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"network diagram","caption_source":"content.json:image_caption"},"width":983,"height":536,"image_format":"jpeg","image_sha256":"347051b21b7fbff1ec1c993b597d645af43f4bdc75e126fc5a49735660b5c36b","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}

subsets/periodic-table-map/test/metadata.jsonl

@@ -1,2 +1,2 @@
-{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_47_figure_5.jpg","caption":"Figure 5. Pidc table of metals and semi-metals, highiging those that have ben deosited as elements or metal lloys by thmal ALD (solid black box) or pulsed-CVD (dashed black box). Atomic numbers are shown above the atom symbols and Pauling electronegativities are shown below for values with two decimal places, the oxidation state is: I for groups 1 and 11, II for groups 2, 4-10 and 12, III for groups 3 and 13, and IV for group 14. Non-metals (including semi-metals) and elements without significant natural abundance are shown in white. Transition metals are shaded in light grey, while other metals are dark grey. Reproduced with permission.[72] Copyright 2013, Elsevier.","id":"test/atomic-layer-deposition/simulation-usecase/47/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/47/figure_5","subset":"periodic-table-map","split":"test","classification":[{"panel_id":"a","label":"periodic table map"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/images/figure_5.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/Simon D. Elliott et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"periodic table map","caption_source":"content.json:image_caption"},"width":1011,"height":622,"image_format":"jpeg","image_sha256":"69acfce56d4f7e25be15e1567d2679afe640a016727db9fa0588b915f8556b1e","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":0,"y":0,"width":1011,"height":622}],"summarization":"[]","data_extraction":"[]","vqa":"[]"}
-{"file_name":"images/test_atomic-layer-etching_simulation-usecase_21_figure_3.jpg","caption":"Figure 3. Color map of surface binding energies  $(E_0)$  of elemental materials, from values independently given by Yamamura and Tawara. Materials with high  $E_0$  values, such as carbon and the refractory metals, are most likely to exhibit high synergy in directional ALE.","id":"test/atomic-layer-etching/simulation-usecase/21/figure_3","sample_id":"atomic-layer-etching/simulation-usecase/21/figure_3","subset":"periodic-table-map","split":"test","classification":[{"panel_id":"a","label":"periodic table map"}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/images/figure_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/Review Paper -- Atomic Layer Etching Rethinking the Art of Etch.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"21","first_classification_panel_id":"a","first_classification_label":"periodic table map","caption_source":"content.json:img_caption"},"width":1408,"height":375,"image_format":"jpeg","image_sha256":"5d71082e5cdfe8dbaa25741a238d55079e439f3b749a9b386626a292f159f4b6","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0","bbox":[{"panel_id":"a","x":2,"y":0,"width":1407,"height":374}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The periodic table is color-coded to show surface binding energy E₀ for elemental solids using discrete ranges from <2 eV to >7 eV. Elements with high E₀, including carbon and refractory transition metals, occupy the darkest bins, indicating strong lattice cohesion and resistance to removal. Alkali metals and weakly bound elements fall into the lowest categories, reflecting low surface stability. The map provides a qualitative basis for anticipating where strong chemical–physical synergy is required for effective directional ALE\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Element | Surface binding energy E₀ (categorical) |\\n|---------|------------------------------------------|\\n| H, Li,Na, K, Rb, Cs, Mg, Ca, Sr, Ba, Zn, Cd, Hg, Tl, Po, Br, I, At, He, Ne, Ar, Kr, Xe, Rn| < 2 eV |\\n| Be, Sc, Y, La, Ti, Cr, Mn, Fe, Co, Ni, Pd, Cu, Ag, Au,Al, Si, Ge, Sn, N, P, As,Sb, Cl| 2–5 eV   |\\n| B, O, F, V, Rh, Pt,  | 5-6 eV  |\\n| Zr, Mo, Tc, Ru, Hf, Ir | 6-7 eV  |\\n| C, Nb, Ta, W, Re, Os | 7+ eV  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Materials with low surface binding energies can often be removed using chemically driven or thermally assisted steps, since only modest bond weakening is required for detachment. In contrast, high-E₀ materials such as refractory metals and carbon require a deliberate sequence in which surface chemistry first reduces local binding strength, followed by directional energy input to complete removal. This separation of chemical modification and physical activation enables atomic-scale control even for strongly bound solids.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"-Refractory transition metals and carbon cluster in the highest E₀ ranges.\\n-Alkali metals appear in the lowest E₀ category, indicating weak surface cohesion.\\n-Several p-block elements (e.g., B, O, F) show elevated E₀ relative to neighbors.\\n-Intermediate E₀ values dominate among common semiconductors and metals, suggesting mixed chemical and physical contributions.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Carbon and refractory transition metals (e.g., Ta, W, Re, Os).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]"}
+{"file_name":"images/test_atomic-layer-deposition_simulation-usecase_47_figure_5.jpg","caption":"Figure 5. Pidc table of metals and semi-metals, highiging those that have ben deosited as elements or metal lloys by thmal ALD (solid black box) or pulsed-CVD (dashed black box). Atomic numbers are shown above the atom symbols and Pauling electronegativities are shown below for values with two decimal places, the oxidation state is: I for groups 1 and 11, II for groups 2, 4-10 and 12, III for groups 3 and 13, and IV for group 14. Non-metals (including semi-metals) and elements without significant natural abundance are shown in white. Transition metals are shaded in light grey, while other metals are dark grey. Reproduced with permission.[72] Copyright 2013, Elsevier.","id":"test/atomic-layer-deposition/simulation-usecase/47/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/47/figure_5","subset":"periodic-table-map","split":"test","classification":[{"panel_id":"a","label":"periodic table map"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":1011,"height":622}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/images/figure_5.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/47/Simon D. Elliott et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"periodic table map","caption_source":"content.json:image_caption"},"width":1011,"height":622,"image_format":"jpeg","image_sha256":"69acfce56d4f7e25be15e1567d2679afe640a016727db9fa0588b915f8556b1e","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}
+{"file_name":"images/test_atomic-layer-etching_simulation-usecase_21_figure_3.jpg","caption":"Figure 3. Color map of surface binding energies  $(E_0)$  of elemental materials, from values independently given by Yamamura and Tawara. Materials with high  $E_0$  values, such as carbon and the refractory metals, are most likely to exhibit high synergy in directional ALE.","id":"test/atomic-layer-etching/simulation-usecase/21/figure_3","sample_id":"atomic-layer-etching/simulation-usecase/21/figure_3","subset":"periodic-table-map","split":"test","classification":[{"panel_id":"a","label":"periodic table map"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The periodic table is color-coded to show surface binding energy E₀ for elemental solids using discrete ranges from <2 eV to >7 eV. Elements with high E₀, including carbon and refractory transition metals, occupy the darkest bins, indicating strong lattice cohesion and resistance to removal. Alkali metals and weakly bound elements fall into the lowest categories, reflecting low surface stability. The map provides a qualitative basis for anticipating where strong chemical–physical synergy is required for effective directional ALE\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Element | Surface binding energy E₀ (categorical) |\\n|---------|------------------------------------------|\\n| H, Li,Na, K, Rb, Cs, Mg, Ca, Sr, Ba, Zn, Cd, Hg, Tl, Po, Br, I, At, He, Ne, Ar, Kr, Xe, Rn| < 2 eV |\\n| Be, Sc, Y, La, Ti, Cr, Mn, Fe, Co, Ni, Pd, Cu, Ag, Au,Al, Si, Ge, Sn, N, P, As,Sb, Cl| 2–5 eV   |\\n| B, O, F, V, Rh, Pt,  | 5-6 eV  |\\n| Zr, Mo, Tc, Ru, Hf, Ir | 6-7 eV  |\\n| C, Nb, Ta, W, Re, Os | 7+ eV  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Materials with low surface binding energies can often be removed using chemically driven or thermally assisted steps, since only modest bond weakening is required for detachment. In contrast, high-E₀ materials such as refractory metals and carbon require a deliberate sequence in which surface chemistry first reduces local binding strength, followed by directional energy input to complete removal. This separation of chemical modification and physical activation enables atomic-scale control even for strongly bound solids.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"-Refractory transition metals and carbon cluster in the highest E₀ ranges.\\n-Alkali metals appear in the lowest E₀ category, indicating weak surface cohesion.\\n-Several p-block elements (e.g., B, O, F) show elevated E₀ relative to neighbors.\\n-Intermediate E₀ values dominate among common semiconductors and metals, suggesting mixed chemical and physical contributions.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Carbon and refractory transition metals (e.g., Ta, W, Re, Os).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":1407,"height":374}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/images/figure_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/21/Review Paper -- Atomic Layer Etching Rethinking the Art of Etch.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"21","first_classification_panel_id":"a","first_classification_label":"periodic table map","caption_source":"content.json:img_caption"},"width":1408,"height":375,"image_format":"jpeg","image_sha256":"5d71082e5cdfe8dbaa25741a238d55079e439f3b749a9b386626a292f159f4b6","metadata_license":"MIT for repository metadata where applicable; see LICENSE.","image_license":"mixed-rights; inspect source article and provenance before reuse.","image_reuse_status":"requires rights review before reuse or redistribution.","schema_version":"1.0.0"}