subsets/line-chart/train/metadata.jsonl

{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_12_fig_10.jpg","caption":"Fig. 10. Tungsten film thickness deposited at  $425 \\mathrm{K}$  versus number of AB cycles. The  $\\mathrm{WF}_6$  and  $\\mathrm{Si}_2\\mathrm{H}_6$  reactant exposures of nine pulses and 40 pulses, respectively, were sufficient for complete half-reactions. The least squares linear fit to the data yields a tungsten growth rate of  $2.5 \\mathrm{\\AA / AB}$  cycle.","id":"train/atomic-layer-deposition/experimental-usecase/12/fig_10","sample_id":"atomic-layer-deposition/experimental-usecase/12/fig_10","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows a linear relationship between the number of AB cycles and the thickness of a tungsten film deposited at 425 K, confirming self-limiting Atomic Layer Deposition (ALD) behavior. A least squares fit gives a tungsten growth rate of 2.5 Å/cycle using reactant exposures of 9 pulses of WF₆ and 40 pulses of Si₂H₆ per cycle, sufficient for complete surface reactions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| AB Cycles | Tungsten Film Thickness (Å) |\\n|---|---|\\n| 0 | 0 |\\n| 10 | 25 |\\n| 20 | 50 |\\n| 30 | 75 |\\n| 40 | 100 |\\n| 50 | 125 |\\n| 60 | 150 |\\n| 70 | 175 |\\n| 80 | 200 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The linear chart shows a consistent, repeatable increase of 2.5 Å per cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"425 K (152 °C).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Immediate nucleation of tungsten from the first cycle.\\n, No incubation period or delay in film growth.\\n, Ideal surface reactivity from the start.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The perfect linearity guarantees uniform, void-free filling of high-aspect-ratio features. The low temperature (425 K) prevents damage to pre-existing chip layers. This enables reliable fabrication of nanoscale tungsten interconnects.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":1,"width":586,"height":576}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/J.W. Klaus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"12","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":594,"height":581,"image_format":"jpeg","image_sha256":"0f04ce1c4e14012a05992cff50250d19cf171ed2c666e9739fb0ab6dbf10123c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_12_fig_11.jpg","caption":"Fig. 11. Tungsten film thickness deposited after three AB cycles versus substrate temperature. The  $\\mathrm{WF}_6$  and  $\\mathrm{Si}_2\\mathrm{H}_6$  reactant exposures at each temperature were sufficient for complete half-reactions.","id":"train/atomic-layer-deposition/experimental-usecase/12/fig_11","sample_id":"atomic-layer-deposition/experimental-usecase/12/fig_11","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line graph represents the tungsten film thickness deposited after 3 AB cycles against substrate temperature (300–600 K). The curve shows a distinct peak: thickness initially rises with temperature, reaches a maximum near 425 K, and then saturates at higher temperatures. WF₆ and Si₂H₆ reactant exposures at each temperature were sufficient for complete half-reactions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Tungsten Film Thickness (Å) |\\n|---|---|\\n| 300 | 3.4 |\\n| 400 | 6.7 |\\n| 500 | 7.5 |\\n| 600 | 7.8 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No.  FTIR confirms the reactions remain self-limiting, surface-saturated half-reactions; the constant rate results from stable surface intermediate coverage, not a gas-phase decomposition mechanism.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The stability of fluorine-containing surface intermediates (WFₓ, SiHᵧF_z) versus the loss of surface hydroxyl groups (dehydroxylation) on oxides.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Their stability keeps the surface consistently saturated with reactive sites, so each cycle deposits the same amount of tungsten. This uniform site availability produces the flat, temperature-independent growth-rate plateau seen on the graph\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It provides a large process window. This minimizes sensitivity to minor temperature fluctuations across a wafer or between tools. It ensures uniform film thickness and properties, improving production yield and repeatability.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":5,"width":595,"height":600}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/J.W. Klaus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"12","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":600,"height":609,"image_format":"jpeg","image_sha256":"dab9ccea58557a79f71245107afc6b8b75501b68734c468d3819ed59e543cdec","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_18_fig4.jpg","caption":"FiG.4. rPRA  t  t  t  t cursor Ar purge. The process parameters for these depositions are stated in Table I.","id":"train/atomic-layer-deposition/experimental-usecase/18/fig4","sample_id":"atomic-layer-deposition/experimental-usecase/18/fig4","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"line chart"},{"panel_id":"d","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"ZrN growth per cycle versus substrate temperature shows a low-temperature ALD window with stable GPC and a sharp increase at higher temperatures due to thermally activated, non-self-limiting growth.\"},{\"panel_id\":\"b\",\"text\":\"GPC saturation with increasing TDMAZr pulse time confirms self-limiting precursor adsorption and complete surface coverage beyond a critical pulse duration.\"},{\"panel_id\":\"c\",\"text\":\"GPC rapidly reaches a constant value with increasing forming-gas plasma exposure, indicating efficient ligand removal and nitridation within short plasma times.\"},{\"panel_id\":\"d\",\"text\":\"Post-precursor purge time affects GPC: insufficient purging causes elevated growth from parasitic reactions, while longer purge times restore true ALD behavior.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Substrate Temperature (°C) | GPC (nm/cycle) |\\n|---|---|\\n| 100 | ~0.11 |\\n| 150 | ~0.10 |\\n| 200 | ~0.11 |\\n| 250 | ~0.15 |\\n| 300 | ~0.55 |\"},{\"panel_id\":\"b\",\"text\":\"| TDMAZr pulse (s) | GPC (nm/cycle) |\\n|---|---|\\n| 0.02 | ~0.07 |\\n| 0.04 | ~0.09 |\\n| 0.06 | ~0.10 |\\n| 0.08 | ~0.10 |\\n| 0.10 | ~0.10 |\\n| 0.12 | ~0.10 |\\n| 0.16 | ~0.10 |\\n| 0.20 | ~0.10 |\"},{\"panel_id\":\"c\",\"text\":\"| Plasma exposure (s) | GPC (nm/cycle) |\\n|---|---|\\n| 1 | ~0.12 |\\n| 3 | ~0.12 |\\n| 6 | ~0.10 |\\n| 9 | ~0.10 |\\n| 12 | ~0.10 |\\n| 15 | ~0.10 |\\n| 18 | ~0.10 |\"},{\"panel_id\":\"d\",\"text\":\"| Post-TDMAZr purge (s) | GPC (nm/cycle) |\\n|---|---|\\n| 1 | ~0.26 |\\n| 3 | ~0.25 |\\n| 6 | ~0.10 |\\n| 9 | ~0.10 |\\n| 12 | ~0.10 |\\n| 15 | ~0.10 |\\n| 18 | ~0.10 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure identifies an ALD process window that ensures stable, self-limiting ZrN growth at low temperatures by balancing precursor exposure, plasma reactivity, and purge efficiency. These optimized conditions yield consistent growth per cycle while suppressing parasitic CVD reactions, enabling uniform, conductive ZrN films suitable for temperature-sensitive electronic and interconnect applications\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"GPC increases and saturates with increasing TDMAZr pulse time, while it decreases and then stabilizes with increasing post-precursor purge time.\"}]},{\"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\":\"Surface saturation by TDMAZr precursor molecules, Efficient removal of surface ligands by forming-gas plasma, Prevention of gas-phase reactions through sufficient purge time, Thermal stability of surface-bound species within the ALD window\"}]}]","bbox":[{"panel_id":"a","x":10,"y":0,"width":490,"height":330},{"panel_id":"b","x":520,"y":0,"width":477,"height":363},{"panel_id":"c","x":10,"y":389,"width":476,"height":352},{"panel_id":"d","x":520,"y":389,"width":477,"height":352}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/images/fig4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/Triratna Muneshwar et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":997,"height":741,"image_format":"jpeg","image_sha256":"845e3a13ad4767a45632c671a54f084ac8248c6d8b390930dd902684da8710fe","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_18_fig_3.jpg","caption":"FIG. 3. (a) Film thickness vs number of PEALD cycles as determined from d-iSE data analysis for  $\\mathrm{ZrN}$  deposition on Si substrate at  $\\mathrm{T_{sub} = 150^{\\circ}C}$ $\\mathrm{t_1 = 0.12s}$ $\\mathrm{t_2 = 15s}$ $\\mathrm{t_3 = 15s}$  and  $\\mathrm{t_4 = 15s}$  In a steady growth regime, GPC of  $0.10\\mathrm{nm}$  cycle was determined from the linear fit (dashed lines). (b) Comparison between the film thicknesses determined from iSE at  $150^{\\circ}\\mathrm{C}$  and ex-situ XRR at room temperature on identical samples.","id":"train/atomic-layer-deposition/experimental-usecase/18/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/18/fig_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"ZrN film thickness increases linearly with PEALD cycle number, measured by dynamic in-situ spectroscopic ellipsometry. Linear fit gives a growth per cycle of ~0.103 nm, confirming immediate nucleation and self-limiting ALD behavior.\"},{\"panel_id\":\"b\",\"text\":\"Comparison of ZrN thickness measured in-situ by ellipsometry and ex-situ by X-ray reflectivity on identical samples. Close agreement validates the accuracy of the ellipsometric thickness model.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycle # | ZrN thickness (nm) |\\n|---|---|\\n| 5  | ~0.5 |\\n| 10 | ~1.0 |\\n| 20 | ~2.1 |\\n| 30 | ~3.1 |\\n| 40 | ~4.1 |\\n| 50 | ~5.2 |\\n| 60 | ~6.2 |\"},{\"panel_id\":\"b\",\"text\":\"| ALD cycle # | Thickness (nm) |\\n|---|---|\\n| 25 | ~2.6 |\\n| 50 | ~5.2 |\\n| 75 | ~7.7 |\\n| 100 | ~10.3 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear thickness increase with cycle number demonstrates excellent process stability and reproducibility, allowing precise thickness control through cycle counting. The validated growth per cycle and agreement between in-situ and ex-situ measurements ensure reliable scaling of ZrN film thickness, which is essential for integrating conductive nitride layers into advanced electronic and interconnect applications.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ZrN film thickness increases linearly with ALD cycle number with a constant growth rate of approximately 0.10 nm per cycle.\"}]},{\"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\":\"Linear thickness increase without incubation delay, Consistent slope across all measured cycles, Agreement between in-situ SE and ex-situ XRR thickness values, Minimal scatter around the linear fit\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":603,"height":449},{"panel_id":"b","x":351,"y":202,"width":237,"height":172}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/Triratna Muneshwar et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":608,"height":453,"image_format":"jpeg","image_sha256":"449c81a6b1abc01002f00a7b20dc5332c9f680bd05fbde0eedbfe9856490c32d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_18_fig_7.jpg","caption":"FiG. 7. (a) Electrical resistivity of  $\\mathrm{ZrN}$  PEALD films deposited on thermal  $\\mathrm{SiO_2}$  substrate at  $150^{\\circ}\\mathrm{C}$ , as a function of measurement temperature. (b) Plot of  $\\Delta \\rho /\\rho_{\\mathrm{o}}$  vs measurement temperature showing TCR of  $0.0088 / ^{\\circ}\\mathrm{C}$  for  $\\mathrm{ZrN}$  films.","id":"train/atomic-layer-deposition/experimental-usecase/18/fig_7","sample_id":"atomic-layer-deposition/experimental-usecase/18/fig_7","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Electrical resistivity of PEALD-grown ZrN films versus measurement temperature shows a linear increase, characteristic of metallic conduction dominated by electron–phonon scattering.\"},{\"panel_id\":\"b\",\"text\":\"Normalized resistivity change (Δρ/ρ₀) versus temperature. Positive temperature coefficient of resistivity (TCR ≈ 0.0088 %/°C) confirms stable metallic transport over the measured range.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Resistivity (µΩ·cm) |\\n|---|---|\\n| 25  | ~560 |\\n| 50  | ~650 |\\n| 75  | ~750 |\\n| 100 | ~880 |\\n| 125 | ~1020 |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (°C) | Δρ/ρ₀ |\\n|---|---|\\n| 25  | 0.00 |\\n| 50  | ~0.002 |\\n| 75  | ~0.004 |\\n| 100 | ~0.006 |\\n| 125 | ~0.009 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear increase of resistivity with temperature and the positive, stable TCR confirm that the PEALD-grown ZrN films exhibit metallic conduction with predictable thermal behavior. This stability in electrical performance, combined with low-temperature deposition capability, supports the suitability of ZrN films for applications such as conductive diffusion barriers, interconnect layers, and electrodes in microelectronic devices where thermal reliability is essential.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The resistivity increases approximately linearly with temperature, yielding a positive TCR of about 0.0088 %/°C.\"}]},{\"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\":\"Presence of metallic Zr 4d conduction states, Electron–phonon scattering dominating charge transport, Continuous, dense ZrN film microstructure, Low defect-induced carrier localization\"}]}]","bbox":[{"panel_id":"a","x":3,"y":6,"width":606,"height":434},{"panel_id":"b","x":344,"y":202,"width":242,"height":161}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/18/Triratna Muneshwar et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":608,"height":441,"image_format":"jpeg","image_sha256":"3b9ccd4183aad7ff41a96c2f349a61475a33c8211b4b8d3c2fb6dd6d72085479","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_2_figure_2.jpg","caption":"Figure 2. Film thickness vs the number of reaction cycles completed.","id":"train/atomic-layer-deposition/experimental-usecase/2/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/2/figure_2","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the evolution of film thickness as a function of the total number of ALD reaction cycles. The relationship is strongly linear, with thickness increasing from ~35 nm at 1000 cycles to ~105 nm at 3000 cycles. The linear fit and narrow error bars suggest a stable, fully saturated ALD process in which the growth per cycle (GPC) remains constant throughout the experiment.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of reaction cycles | Film thickness (nm) |\\n|---------------------------|----------------------|\\n| 1000                      | ~35                  |\\n| 1500                      | ~50                  |\\n| 2000                      | ~70                  |\\n| 3000                      | ~105                 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"- Constant surface saturation during each precursor exposure step\\n- Efficient purging that prevents gas-phase or parasitic reactions\\n- Stable substrate temperature across all cycles\\n- Reproducible adsorption and ligand-removal reactions in every half-cycle\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A constant and uniform growth per cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"CVD-like growth would introduce continuous deposition during reactant exposure, causing the growth rate to depend on time rather than discrete cycles. This typically results in super-linear thickness increases with cycle count. The strictly linear trend observed confirms that growth occurs only during surface-limited reactions, consistent with true ALD rather than CVD behavior.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A linear relationship allows thickness to be defined solely by the number of ALD cycles, enabling deterministic control without reliance on in-situ end-point detection. This is critical for nanoscale devices where deviations of even a few nanometers can significantly alter electrical, optical, or mechanical performance. Linear scaling simplifies process integration and improves reproducibility across wafers and batches.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":529,"height":378}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/2/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/2/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/2/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/2/Aaltonen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"2","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":533,"height":381,"image_format":"jpeg","image_sha256":"38b33f84a9c8cf172d21ae022e8832e387c82e443279f91b84d02ab2781cb8be","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_22_fig_1.jpg","caption":"Fig. 1. Growth rate of LiF thin films as a function of deposition temperature. Lithd and  $\\mathrm{TiF_4}$  pulse lengths were 2 s and purge times between these pulses were 4 s.","id":"train/atomic-layer-deposition/experimental-usecase/22/fig_1","sample_id":"atomic-layer-deposition/experimental-usecase/22/fig_1","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the growth rate (Å/cycle) as a function of the deposition temperature (°C). The growth rate of the material decreases as the deposition temperature increases, reaching at minimum at around 320 °C before slightly increasing again.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition temperature / °C | Growth rate Å / cycle |\\n|---|---|\\n| 250 | 1.5 ± 0.1 |\\n| 275 | 1.3 ± 0.1 |\\n| 300 | 1.2 ± 0.1 |\\n| 310 | 1.0 ± 0.1 |\\n| 350 | 1.2 ± 0.1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The figure shows the growth per cycles (Å/cycle) for different temperatures of deposition (between 250 °C and 350 °C).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the deposition temperature increases, the growth per cycle decreases up to a temperature of 25 °C. After this point, growth per cycle increases again.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"325 °C leads to the lowest growth rate.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At this point, growth per cycle reaches its maximum value, enabling deeper etching in a single cycle.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":634,"height":495}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/Miia Mantymaki et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":637,"height":500,"image_format":"jpeg","image_sha256":"7874440260e4e6451b2701b6df09a0dee832af9620fedb15f77ee586a95157c9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_22_fig_2.jpg","caption":"Fig. 2. Growth rate of LiF thin films as a function of  $\\mathrm{TiF_4}$  pulse length at  $325^{\\circ}C$  The Lithd pulse was kept at 2 s and purge times were 4 s.","id":"train/atomic-layer-deposition/experimental-usecase/22/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/22/fig_2","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the growth rate of a material in Å per cycle as a function of TiF pulse length in seconds. The growth rate increases with increasing pulse length but it slightly decreases at 2 s.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TiF, pulse length / s | Growth rate Å/cycle |\\n|---|---|\\n| 0.5 | 0.95 |\\n| 1 | 1.00 |\\n| 2 | 0.95 |\\n| 3 | 1.20 |\\n| 4 | 1.30 |\\n| 6 | 1.40 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The growth rate increases almost linearly with the length of the TiF4 pulse, but decreases slightly at 2s.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.5 s and 2 s.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No for this case, because the growth rate doesn't remain constant.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":636,"height":493}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/Miia Mantymaki et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":641,"height":500,"image_format":"jpeg","image_sha256":"2cdebae3b5a139d9b917822c2d261b0df6245ac167cf3834ab551cded63c518d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_22_fig_3.jpg","caption":"Fig. 3. Growth rate of LiF thin films as a function of Lithd pulse length at  $325^{\\circ}C$  The  $\\mathrm{TiF_4}$  pulse was kept at 1 s and purge times were 4 s.","id":"train/atomic-layer-deposition/experimental-usecase/22/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/22/fig_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the growth rate (Å/cycle)  as a function of the Lithd pulse length (s). The growth rate increases linearly with increasing lithd pulse length.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Lithd pulse length / s | Growth rate / Å / cycle |\\n|---|---|\\n| 1 | 0.75 ± 0.05 |\\n| 2 | 1.00 ± 0.05 |\\n| 3 | 1.05 ± 0.05 |\\n| 4 | 1.10 ± 0.05 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is not possible to observe this because the growth rate does not remain constant throughout the Lithd pulse length.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The growth rate increases almost linearly with an increase in the length of the Lithd pulse.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1 s of Lithd pulse length show the lowest growth rate with a value of approximately of 0.7 Å/cycle.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Growth rate saturation can be observed as a constant values over changes in the  x-axis values. In this case, increasing the Lithd pulse length to over 4 s would reveal whether the growth rate remains constant. However, it is also necessary to consider the limitations of the process itself, such as the maximum amount of Lithd that can be loaded into the reactor.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":644,"height":490}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/22/Miia Mantymaki et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":647,"height":495,"image_format":"jpeg","image_sha256":"dfd9d1dbdc53bade726626cae20ddcc6a7f686f24bf302efc29e6218967d377a","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_26_figure_7.jpg","caption":"Figure 7. (a) Resistivity versus percentage of  $\\mathrm{SnO_2}$  cycles measured using four-point probe for ALD ITO films prepared on glass at  $275^{\\circ}C$  using 300 cycles. (b) Hall probe measurements of the same ITO films showing carrier concentration and mobility.","id":"train/atomic-layer-deposition/experimental-usecase/26/figure_7","sample_id":"atomic-layer-deposition/experimental-usecase/26/figure_7","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows that resistivity initially drops sharply with increasing % SnO₂ cycles and then increases again beyond 20%.\"},{\"panel_id\":\"b\",\"text\":\"The multi-axis chart indicates that carrier concentration increases significantly with SnO₂ cycles, while mobility remains relatively constant.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| % SnO₂ Cycles | Resistivity (Ohm·cm) | Carrier Concentration (cm⁻³) | Mobility (cm²/V·s) |\\n|---------------|----------------------|------------------------------|---------------------|\\n| 0             | 1e2                  | 1e19                         | 10                  |\\n| 5             | 1e-3                 | 1e20                         | 100                 |\\n| 10            | 1e-3                 | 1e21                         | 100                 |\\n| 15            | 1e-3                 | —                            | —                   |\\n| 20            | 1e-3                 | —                            | —                   |\\n| 25            | 1e3                  | 1e22                         | 100                 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"25%\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Resistivity drops sharply at low SnO₂ percentages, plateaus, and then increases again beyond 20%.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Resisitivity, Carrier concentration, Mobility\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":585,"height":568},{"panel_id":"b","x":650,"y":10,"width":615,"height":559}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/Jeffrey W. Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":1270,"height":570,"image_format":"jpeg","image_sha256":"09502c89c8d6cbea94a8b031ecfff0e3d67f8a1b28fbcfcdb229c08f15cb33a4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_29_figure_5.jpg","caption":"Figure 5. Normalized integrated absorbances for the  $\\mathrm{W - F}_x$  stretching mode  $(\\sim 680~\\mathrm{cm}^{-1})$  and the  $\\mathrm{N - H}_2$  stretching mode  $(\\sim 3400~\\mathrm{cm}^{-1})$  vs. reactant exposure during the (a)  $\\mathrm{NH}_3$  and (b)  $\\mathrm{WF}_6$  half-reactions at  $600~\\mathrm{K}$ .","id":"train/atomic-layer-deposition/experimental-usecase/29/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/29/figure_5","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The normalized integrated absorbance of N-H₂ Stretch and W-Fₓ Stretch against NH3 exposure at T = 600 K.\"},{\"panel_id\":\"b\",\"text\":\"The normalized integrated absorbance of N-H₂ Stretch and W-Fₓ Stretch against WF6 exposure at T = 600 K.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Reactant Exposure (Torr min) | N-H₂ Stretch | W-Fₓ Stretch |\\n|---|---|---|\\n| 0 | 0.0 | 0.0 |\\n| 10⁻³ | 0.1 | 0.9 |\\n| 10⁻² | 0.2 | 0.8 |\\n| 10⁻¹ | 0.3 | 0.7 |\\n| 1 | 0.4 | 0.6 |\\n| 10 | 0.5 | 0.5 |\\n| 10² | 0.6 | 0.4 |\\n| 10³ | 0.7 | 0.3 |\"},{\"panel_id\":\"b\",\"text\":\"| Reactant Exposure (Torr min) | N-H₂ Stretch | W-Fₓ Stretch |\\n|---|---|---|\\n| 0 | 0.0 | 0.0 |\\n| 10⁻³ | 0.1 | 0.9 |\\n| 10⁻² | 0.2 | 0.8 |\\n| 10⁻¹ | 0.3 | 0.7 |\\n| 1 | 0.4 | 0.6 |\\n| 10 | 0.5 | 0.5 |\\n| 10² | 0.6 | 0.4 |\\n| 10³ | 0.7 | 0.3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Roughly 0.1.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 0.9.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"During the NH3 exposure, the loss of WFx surface species occurs more rapidly than the adsorption of NH2 surface species. When compared to the WF6 exposure, the rate of loss and gain of surface species is roughly equal. The authors relate this phenomenon to NH3 molecules reacting with multiple WFx surface species.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 0.7.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":653,"height":399},{"panel_id":"b","x":0,"y":403,"width":658,"height":436}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_5.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":"line chart","caption_source":"content.json:image_caption"},"width":658,"height":875,"image_format":"jpeg","image_sha256":"81b8e397e377e1afaaee38dcd141f8a33bc67c3ad3fb16b1093e960e5defab42","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_29_figure_8.jpg","caption":"Figure 8. Tungsten nitride film thickness measured by ellipsometry vs. number of AB cycles at  $600~\\mathrm{K}$ . The  $\\mathrm{NH_3}$  exposure of  $12,000\\mathrm{L}$  and  $\\mathrm{WF_6}$  exposure of  $4300\\mathrm{L}$  were sufficient for complete half-reactions during each AB cycle.","id":"train/atomic-layer-deposition/experimental-usecase/29/figure_8","sample_id":"atomic-layer-deposition/experimental-usecase/29/figure_8","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows a linear relationship between the number of AB cycles and film thickness, indicating a linear growth rate.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| AB Cycles | Film Thickness (Å) |\\n|---|---|\\n| 3 | 7.5 |\\n| 6 | 15 |\\n| 12 | 30.5 |\\n| 19 | 48.5 |\\n| 25 | 64 |\\n| 32 | 81.5 |\\n| 48 | 122.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"75 Å.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"12,000 L.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"600 K.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2.55 Å/cycle.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":667,"height":641}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_8.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":"line chart","caption_source":"content.json:image_caption"},"width":672,"height":647,"image_format":"jpeg","image_sha256":"5ae8985e885d69d0381b98cd173842f15dbe9cab28e0e5fd67268c22d1865492","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_29_figure_9.jpg","caption":"Figure 9. Ellipsometric measurements of the tungsten nitride film thickness deposited by three AB cycles at various substrate temperatures. Reactant exposures were sufficient for complete half-reactions at each temperature.","id":"train/atomic-layer-deposition/experimental-usecase/29/figure_9","sample_id":"atomic-layer-deposition/experimental-usecase/29/figure_9","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the relationship between film thickness and temperature, showing an increase up to 600 K followed by a plateau.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Film Thickness (Å) |\\n|---|---|\\n| 400 | 0.5 |\\n| 500 | 3.5 |\\n| 550 | 7.0 |\\n| 600 | 7.75 |\\n| 650 | 8.0 |\\n| 700 | 7.75 |\\n| 750 | 8.0 |\\n| 800 | 8.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It increases rapidly with increasing substrate temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 600 and 800 K.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the substrate temperature increases from 400 to 600 K, the tungsten nitride film thickness deposited during three AB cycles rises rapidly, indicating that the surface reactions are becoming more efficient. In this lower temperature regime, FTIR measurements suggest that the half-reactions do not fully complete, and may even result in the formation of a WF₆:NH₃ adduct species. Once the temperature exceeds 600 K, the deposited film thickness stabilizes and increases only slightly up to 800 K. This plateau corresponds to approximately 2.5 Å per AB cycle, consistent with the formation of a W₂N monolayer and confirms that the surface half-reactions proceed to completion in this higher temperature range.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":1,"width":672,"height":675}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/29/images/figure_9.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":"line chart","caption_source":"content.json:image_caption"},"width":680,"height":680,"image_format":"jpeg","image_sha256":"29f460e6f2e3250d768c02000c662bad4a5a73c3956a7e024ca9908c3fcd9e17","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_33_figure_3.jpg","caption":"$0.9\\mu \\mathrm{mol}\\mathrm{cm}^{-2}$  .b, Growth rate of cobalt at  $300^{\\circ}C$  as a function of hydrogen dose; the cobalt dose was fixed at 2 nmol  $\\mathbb{C}\\mathbb{m}^{-2}$ Figure 3 Data showing that the thicknesses of copper films approach a saturated value as the doses of the precursors are increased. a, Growth rate of copper at  $250^{\\circ}C$  as a function of copper precursor dose; the hydrogen dose was fixed at  $3\\mu \\mathrm{mol}\\mathrm{cm}^{-2}$ . b, Growth rate of copper at  $250^{\\circ}C$  as a function of hydrogen dose; the copper dose was fixed at  $4\\mathrm{nmol}\\mathrm{cm}^{-2}$ .","id":"train/atomic-layer-deposition/experimental-usecase/33/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/33/figure_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"line chart"},{"panel_id":"d","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Panel (a) shows the growth rate increases with precursor temperature and plateaus at ~0.045 Å/cycle above 60 °C.\"},{\"panel_id\":\"b\",\"text\":\"Panel (b) demonstrates that increasing pulse time beyond 2 s does not further increase the growth rate, indicating saturation behavior.\"},{\"panel_id\":\"c\",\"text\":\"Panel (c) shows a similar saturation with hydrogen pressure—growth rate reaches a maximum of ~0.045 Å/cycle at 2 atm.\"},{\"panel_id\":\"d\",\"text\":\"Panel (d) highlights the effect of H₂ dose, where a minimum threshold is needed to initiate growth, followed by saturation. These trends collectively illustrate the self-limiting surface reactions characteristic of ALD processes.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Precursor temperature (°C) | Growth rate (Å per cycle) |\\n|---------------------------|----------------------------|\\n| 50                        | 0.025                      |\\n| 55                        | 0.035                      |\\n| 60                        | 0.045                      |\\n| 65                        | 0.045                      |\\n| 75                        | 0.045                      |\"},{\"panel_id\":\"b\",\"text\":\"| Pulse time (s) | Growth rate (Å per cycle) |\\n|----------------|----------------------------|\\n| 1              | 0.25                       |\\n| 2              | 0.35                       |\\n| 3              | 0.35                       |\\n| 4              | 0.35                       |\"},{\"panel_id\":\"c\",\"text\":\"| H₂ pressure (atm) | Growth rate (Å per cycle) |\\n|-------------------|----------------------------|\\n| 0                 | 0.00                       |\\n| 1                 | 0.025                      |\\n| 2                 | 0.045                      |\\n| 3                 | 0.045                      |\"},{\"panel_id\":\"d\",\"text\":\"| H₂ pressure (atm) | Growth rate (Å per cycle) |\\n|-------------------|----------------------------|\\n| 0                 | 0.00                       |\\n| 1                 | 0.025                      |\\n| 2                 | 0.045                      |\\n| 3                 | 0.045                      |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Precursor pulse time, Hydrogen pressure\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.045 Å per cycle.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The plateau observed in each panel reflects the self-limiting nature of atomic layer deposition (ALD). In ALD, once all available reactive sites on the substrate surface are saturated during a precursor pulse, further increases in precursor supply (by time, dose, or temperature) do not result in additional growth. This ensures precise thickness control and uniform film coverage. For example, once sufficient H₂ pressure or dose is supplied to reduce surface intermediates, additional hydrogen does not increase growth, maintaining a constant growth per cycle.\"}]}]","bbox":[{"panel_id":"a","x":24,"y":10,"width":657,"height":450},{"panel_id":"b","x":751,"y":8,"width":665,"height":562},{"panel_id":"c","x":24,"y":491,"width":666,"height":548},{"panel_id":"d","x":749,"y":603,"width":659,"height":538}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/BOOYONG S. LIM et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":1417,"height":1189,"image_format":"jpeg","image_sha256":"d4e57a3a4c610036481dc05c5f75df0942acbf51cbc42bf2da87647548e2e6a1","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_33_figure_4.jpg","caption":"Figure 4 The linear relationship between Co film thickness on  $\\mathrm{SiO}_2$  and the number of deposition cycles at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/33/figure_4","sample_id":"atomic-layer-deposition/experimental-usecase/33/figure_4","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure illustrates a linear relationship between the number of ALD cycles and the resulting film thickness in angstroms (Å). As the number of cycles increases from 0 to 3000, the film thickness increases proportionally from approximately 5 Å to 120 Å. This linear trend, supported by error bars, demonstrates a consistent growth-per-cycle (GPC) rate, which is a key indicator of controlled and repeatable deposition in ALD processes\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle number | Thickness (Å) |\\n|---|---|\\n| 0 | 5 |\\n| 100 | 10 |\\n| 1000 | 40 |\\n| 2000 | 80 |\\n| 3000 | 120 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear increase in film thickness with cycle number indicates a highly controlled atomic layer deposition (ALD) process where the growth-per-cycle (GPC) remains constant over time. This suggests that each ALD cycle deposits a uniform layer with no saturation or deviation, highlighting the process repeatability and reliability, which are essential for precise thickness control in thin-film applications\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"40 Å.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Linear trend between cycle number and thickness, Small error bars across all data points\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":658,"height":531}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/33/BOOYONG S. LIM et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":658,"height":531,"image_format":"jpeg","image_sha256":"5dc54da2d7aed122fd8748ab583bd2d9a92b1e576a906ae93d8b31881627cb6e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_38_figure_11.jpg","caption":"Figure 11. Average oxidation state of manganese determined from the TEY XANES (black squares).","id":"train/atomic-layer-deposition/experimental-usecase/38/figure_11","sample_id":"atomic-layer-deposition/experimental-usecase/38/figure_11","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates a progressive decrease in the average manganese oxidation state from ~3.7 to ~2.5 with increasing lithiation cycles, confirming lithium ion insertion into the lattice. This evolution indicates a transition from the initial MnO₂ phase through the LiMn₂O₄ spinel regime, followed by degradation toward an inactive MnO rock-salt–like surface state at high cycle numbers.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Amount of lithium cycles | Mn average oxidation state |\\n|--------------------------|----------------------------|\\n| MnO₂ (reference) | ~3.8 |\\n| 10               | ~3.7 |\\n| 50               | ~3.6 |\\n| 100              | ~3.25 |\\n| 200              | ~3.3 |\\n| 200 + A          | ~3.1 |\\n| 300              | ~2.9 |\\n| 300 + A          | ~2.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxidation state of Mn decreases as the lithium concentration (the amount of lithium cycles in this case, as is shown in the figure) increases.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The MnO2 sample shows an average oxidation state value close to 3.7, which decreases as the number of lithium cycles increases, as does the value for annealed samples ( after 200 and 300 cycles).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The annealed sample with 300 cycles of lithium.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Depending on the number of lithium cycles and post-treatment (with or without annealing), it is possible to produce different species (i.e. different oxidation states of Mn).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The data suggests that the undoped MnO2 does not have an oxidation state of 4+, indicating that it is not exactly MnO2.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The number of lithium cycles should roughly be around 60.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"More lithium means that oxygen is not only bonded to manganese anymore but also to lithium, lowering the oxidation state of manganese.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both for 200 and 300 cycles of lithium the oxidation state decreases after an anneal.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The oxidation state is extracted from TEY XANES measurements by comparing the Mn absorption edge position and spectral shape with reference compounds of known oxidation state. This allows an average Mn valence to be assigned for each lithium cycling condition.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Mn oxidation state decreases as the number of lithium cycles increases.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It decreases slightly, from about ~3.25 to ~3.1–3.15.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Oxidation states near ~2.5 imply significant Mn³⁺ content, which can drive Jahn–Teller distortion, lattice strain, and eventual capacity fading. This highlights a trade-off between lithiation depth and structural stability.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It decreases from ~+3.7 in pristine MnO₂ to ~+2.5 in 300cLi-MnO₂-600, reflecting progressive reduction of Mn⁴⁺ to Mn³⁺ (and possibly Mn²⁺) upon lithium intercalation.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The XANES edge position is affected by factors beyond simple valence, including bond lengths, coordination geometry, and covalent/ionic character. The presence of Jahn-Teller active Mn³⁺ ions distorts the local structure, contributing to the apparent low oxidation states.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Annealed samples—200+A was annealed at 600°C and 300+A at 600°C for 10 min under N₂.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. Despite having near-stoichiometric Li₁.₁Mn₂O₄ composition, the measured oxidation state (~2.9–3.0) is below the expected +3.5, possibly due to impurity phases or structural distortions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Perform 100 charge-discharge cycles.\\n\\nDisassemble in glovebox.\\n\\nClean and dry electrode.\\n\\nTransfer air-free to beamline.\\n\\nAlign and measure.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Annealed samples show higher Mn oxidation states, indicating better structural recovery and slower degradation than non‑annealed cycled electrodes.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Mn oxidation decreases toward +3 with cycling due to increased Mn³⁺. This leads to Jahn‑Teller distortion, structural instability, and capacity fade.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The steady decline in oxidation state suggests predictable, cycle‑dependent degradation, enabling reliable lifetime estimates for LiMn₂O₄ batteries.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. valence of the absorbing atom\\n2. structural variations\\n3.  ligand type \\n4.  geometry\\n5.  coordination number\\n6.  bond lengths\\n7. covalent or ionic character of the bonds\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300cLi−MnO2 and 300cLi−MnO2-600\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The determination of the oxidation states was also attempted from the average Mn−O bond lengths obtained from the EXAFS results. The analysis was based on the bond valence model50,51 where the oxidation state of an ion is correlated to the nearest neighbor distance in the metal−ligand bond. The oxidation states determined by this method varied in the range of +4 to +3.5, but without any correlation to the lithium concentration.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"+3.75.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"+2.5.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Although the trend is as expected, the measured oxidation states are unrealistically low. However, in addition to the valence of the absorbing atom, structural variations, ligand type, geometry, coordination number, bond lengths, and the covalent or ionic character of the bonds are known to have effects on the chemical shifts in the absorption spectra, which might affect the results and cause the unrealistically low oxidation states. The very low oxidation states of 300cLi−MnO2 and 300cLi−MnO2-600 could arise from some unexpected behavior related to 300cLi− MnO2, which was also noticed earlier in the shape of the XAS spectrum.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"+3.5.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300 + A.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Annealing generally lowers the average oxidation state of manganese compared to the non-annealed samples.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The effect of lithium incorporation into the manganese film is a reduction in the average oxidation state of manganese. As the number of lithium cycles increases from 0 to 300, the manganese oxidation state steadily decreases from approximately +3.7 down to +2.5. This is indicating that incorporation of lithium causes manganese to gain electrons.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"50 cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The average manganese oxidation state decreases steadily as the number of lithium cycles increases, indicating progressive reduction of Mn from values characteristic of MnO₂ toward lower oxidation states at higher lithium loadings.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 2.5\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"More Li uptake, Mn reduction, Shift toward LiMn₂O₄ phase\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately +2.5.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"MnO₂ starts at a high oxidation state (~+3.8), Oxidation state gradually decreases with increasing lithium content, 100cLi−MnO₂ and 200cLi−MnO₂ show intermediate oxidation states (~+3.3 to +3.2), Annealed samples show slightly lower oxidation states than unannealed ones, 300cLi−MnO₂-600 exhibits the lowest oxidation state (~+2.5)\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The manganese oxidation state strongly influences the structural stability, electronic structure, and electrochemical behavior of Li-Mn-O materials. A decreasing oxidation state with increasing lithium content reflects changes in charge compensation and local bonding environments, which can affect lithium storage capacity, phase stability, and degradation mechanisms. Identifying unusually low oxidation states, such as those observed in highly lithiated and annealed samples, helps highlight potential structural distortions or impurity phases that may impact long-term performance in applications like lithium-ion batteries.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The average oxidation state of manganese was determined from TEY XANES measurements.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":517,"height":433}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/Nieminen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":517,"height":433,"image_format":"jpeg","image_sha256":"7c6f3e8c252caef1e0ccf260d1de6272269cdd3fc7ed4119914eddabd5f00985","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_49_figure_5.jpg","caption":"Figure 5. a) Photocurrent density and response of the TiNTs (purple) and  $\\mathrm{MoS_2 / TiNTs}$  (red, with 200 ALD cycles  $\\mathrm{MoS_2}$  deposition). The photocurrent was recorded in  $0.5\\mathrm{M}$ $\\mathrm{Na_2SO_4}$  electrolyte with a  $0.5\\mathrm{V}$  external bias potential in the dark and under  $430~\\mathrm{nm}$  light irradiation. b) Nyquist and c) MottSchottky plots of TiNTs (purple) and  $\\mathrm{MoS_2 / TiNTs}$  heterostructure (red). The Nyquist plots were recorded under  $430~\\mathrm{nm}$  light irradiation.","id":"train/atomic-layer-deposition/experimental-usecase/49/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/49/figure_5","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"spectra chart"},{"panel_id":"c","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure compares the photoelectrochemical (PEC) performance of bare Titanium Oxide Nanotubes (TiNTs, purple) and MoS₂-coated TiNT heterostructures (MoS₂/TiNTs, red). Under chopped light illumination, the MoS₂/TiNT heterostructure exhibits a significantly enhanced photocurrent density of approximately 30 μA/cm², compared to about 10 μA/cm² for bare TiNTs. This nearly threefold increase indicates improved light absorption, charge separation, and carrier transport due to the heterostructure formation.\"},{\"panel_id\":\"b\",\"text\":\"The Nyquist plot shows that the MoS₂/TiNTs sample has a much smaller semicircle radius in the high-frequency region than bare TiNTs. This reflects a lower charge-transfer resistance (R_ct) at the electrode–electrolyte interface, indicating faster interfacial charge transfer.\"},{\"panel_id\":\"c\",\"text\":\"Both samples exhibit positive slopes, confirming n-type semiconducting behavior. The MoS₂/TiNT heterostructure shows a smaller slope than bare TiNTs, corresponding to a higher donor density (N_d), which contributes to enhanced electrical conductivity and improved PEC performance.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Current Density (µA/cm²) |\\n|---------:|--------------------------:|\\n| 100      | 30                       |\\n| 200      | 27                       |\\n| 300      | 26                       |\\n| 400      | 25                       |\\n| 500      | 25                       |\\n| 600      | 24                       |\"},{\"panel_id\":\"b\",\"text\":\"| Z′ (kΩ) | Z″ (kΩ) | Materials |\\n|---|---| -- |\\n| 0.0 | 0.0 | TiNTs |\\n| 0.0 | 00 | MoS2/TiNTs |\\n| 0.1 | 0.4 | TiNTs |\\n| 0.1 |  0.1 | MoS2/TiNTs |\\n| 0.2 | 0.7 | TiNTs |\\n| 0.2 |  0.25 | MoS2/TiNTs |\\n| 0.3 | 0.1 | TiNTs |\\n| 0.3 |  0.3 | MoS2/TiNTs |\\n| 0.4 | -- | TiNTs |\\n| 0.4 |  0.3 | MoS2/TiNTs |\\n| 0.5 | -- | TiNTs |\\n| 0.5 | 0.29 | MoS2/TiNTs |\\n| 0.6 | -- | TiNTs |\\n| 0.6 | 0.26 | MoS2/TiNTs |\\n| 0.7 | -- | TiNTs |\\n| 0.7 | 0.22 | MoS2/TiNTs |\\n| 0.8 | -- | TiNTs |\\n| 0.8 | 3.4 | MoS2/TiNTs |\\n| 0.9 | -- | TiNTs |\\n| 0.9 | 0.57 | MoS2/TiNTs |\\n| 1.0 | -- | TiNTs |\\n| 1.0 |  0.8 | MoS2/TiNTs |\"},{\"panel_id\":\"c\",\"text\":\"| E (V) vs. (Ag/AgCl) | 1/C² (cm⁴µF²) | Material |\\n|---|---|---|\\n| -1.0 | 0.00 | TiNTs |\\n| -1.0 | 0.00 | MoS2/TiNTs |\\n| -0.5 | 0.00 |TiNTs |\\n| -0.5 | 0.00 | MoS2/TiNTs |\\n| -0.53 | 0.00 |TiNTs |\\n| -0.53 | 0.00 | MoS2/TiNTs |\\n| 0.0 | 0.9 |TiNTs |\\n| 0.0 | 0.02 | MoS2/TiNTs |\\n| 0.5 | 0.11 |TiNTs |\\n| 0.5 | 0.05 | MoS2/TiNTs |\\n| 1.0 | 0.13 | TiNTs |\\n| 1.0 | 0.06 | MoS2/TiNTs |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The bare TiNTs exhibit a peak photocurrent of approximately 10 μA/cm², whereas the MoS₂/TiNTs heterostructure reaches about 30 μA/cm², corresponding to a threefold enhancement. The immediate drop of current to zero when the light is switched off confirms that the current is purely photo-generated. This behavior demonstrates effective photo-switching and confirms the device functions as a photo-anode driven entirely by illumination.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The peak photocurrent of the MoS₂/TiNTs decreases slightly from approximately 32 to 20 µA/cm² over the measurement period. This gradual photocurrent decay indicates limited long-term stability, suggesting that the MoS₂ layer may undergo slow degradation, oxidation, or partial delamination under sustained illumination and applied bias in the Na₂SO₄ electrolyte.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The MoS₂/TiNTs sample (red) exhibits higher conductivity. This is evidenced by the flatter (smaller) slope of the red curve compared to the steeper purple curve. Since the carrier density (Nₙ) is inversely proportional to the slope of the Mott–Schottky plot, a smaller slope indicates a higher carrier density. The increased carrier density directly leads to improved electrical conductivity and enhanced photoelectrochemical performance.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The heterojunction facilitates charge transfer. The bare TiNTs show a large semicircle, indicating high charge-transfer resistance at the electrode–electrolyte interface. The MoS₂ coating forms a favorable band alignment with TiO₂, providing active sites that enhance interfacial charge transfer and reduce the resistance, resulting in a much smaller semicircle.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":3,"width":336,"height":312},{"panel_id":"b","x":348,"y":9,"width":328,"height":304},{"panel_id":"c","x":673,"y":6,"width":333,"height":308}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/49/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/49/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/49/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/49/Shen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"49","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":1008,"height":317,"image_format":"jpeg","image_sha256":"a650741ae4f81120e8cb27883c7e159bc32f64389ca77c7a9a2b888d8753496e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_52_figure_3.jpg","caption":"Figure 3. Saturation curves for remote plasma ALD of  $\\mathrm{LiCOO_2}$  at  $325^{\\circ}C$  (a) Growth rate as a function of  $\\mathrm{LiO^4}$  Bu dosing time, while the  $\\mathrm{CoCp_2}$  dosing time and the plasma exposure time are kept constant at  $2\\mathrm{s}$  and  $5\\mathrm{s}$  respectively. (b) Growth rate as a function of  $\\mathrm{CoCp_2}$  dosing time, while the  $\\mathrm{LiO^4}$  Bu dosing time and the plasma exposure time are kept constant at  $2\\mathrm{s}$  and  $5\\mathrm{s}$  respectively. The lines serve as guides to the eye.","id":"train/atomic-layer-deposition/experimental-usecase/52/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/52/figure_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The growth per cycle increases with increasing Li precursor dosing time until it plateaus.\"},{\"panel_id\":\"b\",\"text\":\"The growth per cycle shows a slight increase with increasing Co precursor dosing time, but remains relatively stable\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Li precursor dosing time (s) | Growth per cycle (nm) |\\n|---|---|\\n| 0 | 0.00 |\\n| 0.5 | 0.08 |\\n| 1 | 0.10 |\\n| 2 | 0.12 |\\n| 3 | 0.12 |\"},{\"panel_id\":\"b\",\"text\":\"| Co precursor dosing time (s) | Growth per cycle (nm) |\\n|---|---|\\n| 0 | 0.00 |\\n| 0.5 | 0.08 |\\n| 1 | 0.09 |\\n| 2 | 0.08 |\\n| 3 | 0.10 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A plateau indicates that increasing precursor dosing time no longer increases the growth per cycle, which is consistent with surface saturation. In ALD terms, this suggests the surface reactions become self limiting once enough precursor has been supplied. Operating at or beyond the saturation point helps ensure reproducible growth and composition control.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"CoCp2 reaches saturation faster, because panel b levels off at a shorter dosing time than panel a. In panel b, the growth per cycle increases sharply and then becomes nearly constant by around the first second or less. In panel a, the curve continues rising over a broader dosing time range before leveling off.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The x axis in panel b is Co precursor dosing time (s), and the y axis in both panels is growth per cycle (nm).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"To avoid dose limited growth that would change the growth per cycle.\\n\\nTo improve run to run reproducibility by ensuring surface saturation.\\n\\nTo reduce sensitivity to small dosing time fluctuations during long depositions.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":5,"width":603,"height":427},{"panel_id":"b","x":1,"y":451,"width":603,"height":432}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/Donders et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"52","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":609,"height":886,"image_format":"jpeg","image_sha256":"c39addbbfa30afacec08423f0fbbfbdb91faaa1b68ef6328e5c644fe9e787420","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_52_figure_4.jpg","caption":"Figure 4. Thickness of  $\\mathrm{LiCoO_2}$  on  $\\mathrm{Si(100)}$  as a function of the number of ALD super-cycles, as measured with in situ spectroscopic ellipsometry. The used Co:Li dosing ratio is 4:1.","id":"train/atomic-layer-deposition/experimental-usecase/52/figure_4","sample_id":"atomic-layer-deposition/experimental-usecase/52/figure_4","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A line chart showing a linear relationship between the number of supercycles and thickness in nanometers\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of supercycles | Thickness (nm) |\\n|-----------------------|---------------|\\n| 0                     | 0             |\\n| 50                    | 20            |\\n| 100                   | 40            |\\n| 150                   | 60            |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"In situ spectroscopic ellipsometry\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Thickness increases roughly linearly with super cycle number. This indicates that the average thickness added per super cycle is approximately constant over the range shown. Such linearity is consistent with stable, repeatable growth during the super cycle process under the reported conditions.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Si(100).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ratio refers to how many cycles of the Co containing sub process are applied relative to cycles of the Li containing sub process within the super cycle sequence. A 4:1 ratio means the Co process is applied more frequently than the Li process during growth. This is a process control parameter used to tune the film composition while maintaining stable thickness increase per super cycle.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":550,"height":364}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/Donders et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"52","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":556,"height":370,"image_format":"jpeg","image_sha256":"8a0473aba0d042fea8857b72f7093e918acd0f0679131c1deb078b0ef8515356","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_59_fig_9.jpg","caption":"FIG. 9. (Color online) The time-integrated intensity of the OES signal during the plasma step plotted as a function of the PDMAT dosing time in the ALD cycle. The OES signal at  $388~\\mathrm{nm}$  is integrated over the  $10\\mathrm{~s~H}_2 - \\mathrm{N}_2$  (98:2) plasma exposure time in the ALD cycle. The line serves as a guide to the eyes.","id":"train/atomic-layer-deposition/experimental-usecase/59/fig_9","sample_id":"atomic-layer-deposition/experimental-usecase/59/fig_9","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between the integrated OES intensity (a.u.) versus the PDMAT dosing (s). Intensity increases with increasing dosing time up to approximately three second, after which it plateaus.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| PDMAT dosing (s) | Integrated OES intensity (a.u.) |\\n|------------------|--------------------------------|\\n| 0                | 760                            |\\n| 1                | 825                            |\\n| 2                | 875                            |\\n| 3                | 900                            |\\n| 4                | 860                            |\\n| 5                | 875                            |\\n| 7                | 880                            |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The OES signal at 388 is integrated for this plot.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The integrated OES intensity increases with the PDMAT dosing time, until it reached a plateau at around 4 s.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The highest integrated OES intensity values is 900 a.u., which is reached after 3 s od PDMAT dosing.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes, because after 3 s, it is possible to observe that the curve remains constant and the integrated OES intensity is based at 388 nm, where that bond can be found.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":0,"width":604,"height":460}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/Langereis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"59","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":608,"height":464,"image_format":"jpeg","image_sha256":"056617f5ac05c09e09744731b5e8e681e5d820f4af8a18e7862eed715035f054","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_66_figure_7.jpg","caption":"Figure 7. Frequency dependence of (a) dielectric constant  $\\epsilon_{\\mathrm{r}}$  and (b) dielectric loss tan  $\\delta$  and (c) capacitance-voltage characteristics of TM-doped  $\\mathrm{ZnO}$  films.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_7","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_7","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This plot shows dielectric constant as a function of frequency. Dielectric constant (ε<sub>r</sub>) decreases with increasing frequency for all materials (Co, Fe, Ni).\"},{\"panel_id\":\"b\",\"text\":\"Plot shows the tangent delta (tanδ) as a function of frequency where tangent decreases with increasing frequency for all materials (Co, Fe, Ni).\"},{\"panel_id\":\"c\",\"text\":\"Plot shows variation of capacitance (C) as a function of  applied voltage for different frequencies (100kHz, 1MHz) and materials (Co, Fe, Ni).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| frequency (Hz) | dielectric constant, ε<sub>r</sub>  of Co | Fe | Ni |\\n| --- | --- | --- | --- | \\n| 10<sup>3</sup>  | 50 | 50 | 75 |\\n| 10<sup>4</sup>  | 50 | 50 | 50 |\\n| 10<sup>5</sup>  | 50 | 40 | 30 |\\n| 10<sup>6</sup>  | 50 | 20 | 10 |\"},{\"panel_id\":\"b\",\"text\":\"| frequency (Hz) | tanδ of Co | Fe | Ni |\\n| --- | --- | --- | --- | \\n| 10<sup>3</sup> | 800 | 200 | 200 | \\n| 10<sup>4</sup> | 400 | 100 | 100 | \\n| 10<sup>5</sup> | 200 | 50 | 50 | \\n| 10<sup>6</sup> |  80 | 20 | 20 |\"},{\"panel_id\":\"c\",\"text\":\"| Applied Voltage (V) | C (Co, 100kHz) | C (Fe, 100kHz) | C (Ni, 100kHz) | C (Co, 1MHz) | C (Fe, 1MHz) | C (Ni, 1MHz) |\\n|---|---|---|---|---|---|---|\\n| -4 | 220 | 230 | 480 | 40 | 50 | 70 |\\n| -2 | 230 | 260 | 500 | 45 | 55 | 75 |\\n| 0  | 200 | 320 | 520 | 50 | 60| 100 |\\n| 2  | 200 | 320 | 520 | 50 | 60 | 80 |\\n| 2  | 260 | 280 | 500 | 45 | 55 | 75 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the ZnO/Ni film, it is seen that there is another polarization mechanism, which is involved at f = 10^3−10^4 Hz and significantly increases εr in this frequency range. The most likely polarization mechanism, which adds to the net polarizability in the range 10^3−10^5 Hz, is the orientation (dipolar) polarization\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The dielectric response of the ZnO/Co film is significantly weaker than that of the other two films, εr is substantially smaller than that of ZnO/Ni and ZnO/Fe films; it is also smaller than that of pure ZnO layers. However, it should be noticed that the measurement of C−f curves is performed at a small voltage (0.3−0.5 V) where the capacitance of the ZnO/Co film has the minimum in the C−V curves, while that of the other two films reaches its maximum\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A strong frequency dependence of the dielectric constant is observed, especially for Fe- and Ni-doped ZnO. Various polarization mechanisms (ionic, orientation, and space\\ncharge) can exist in the layers and give rise to an increased polarization (hence εr) at a certain frequency\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"All layers have a relatively high dielectric loss  due to the increased conductivity of the layers.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":321,"height":280},{"panel_id":"b","x":332,"y":3,"width":330,"height":272},{"panel_id":"c","x":135,"y":287,"width":390,"height":307}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_7.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":"line chart","caption_source":"content.json:image_caption"},"width":663,"height":595,"image_format":"jpeg","image_sha256":"fc36d829be81d580b533e6685f94fb75d6a68c64284efbdff1203a27ef214194","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_12_figure_8.jpg","caption":"Figure 8. BOMD simulations performed on adsorbed precursors in the presence of five layers of water.","id":"train/atomic-layer-deposition/simulation-usecase/12/figure_8","sample_id":"atomic-layer-deposition/simulation-usecase/12/figure_8","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows a stepwise increase in temperature over time, starting at 100 K and reaching a plateau at 300 K and 580 K.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (ps) | Temperature (K) |\\n|---|---|\\n| 0 | 100 |\\n| 2 | 250 |\\n| 4 | 300 |\\n| 6 | 480 |\\n| 8 | 580 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Born-Oppenheimer Molecular Dynamics (BOMD)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The simulations start at a temperature of 100 K, and then increase over time until they reach approximately 580 K.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2 fs for all the MD runs.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The temperature starts to increase at 5 ps and reaches a plateau of 580 K at approximately 7 ps.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":1,"width":665,"height":479}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/12/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/12/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/12/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/12/Atashi B. Mukhopadhyay et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"12","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":670,"height":486,"image_format":"jpeg","image_sha256":"ce80e9c64e91c5bd0a26d36be500e55fe2062a2f576ff10a2fb8d0fc2ede1327","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_17_figure_12.jpg","caption":"Figure 12: Computed band gap with regards to the Pb content in the compound [94].","id":"train/atomic-layer-deposition/simulation-usecase/17/figure_12","sample_id":"atomic-layer-deposition/simulation-usecase/17/figure_12","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line graph shows the variation in computed bandgap (eV) as a function of  lead (Pbₓ)  in Sn₁₋ₓPbₓO. The band gap seems to increase with increase in (Pbₓ)  content.  This trend, calculated using Density Functional Theory (DFT), demonstrates that lead doping is an effective method for tuning the electronic properties of tin oxide without creating defect states within the band gap.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| x: Sn_{1-x}Pb_xO | band gap (eV) |\\n|---|---|\\n| 0 | 0.4 |\\n| 0.2 | 0.55 |\\n| 0.4 | 0.8 |\\n| 0.6 | 1.1 |\\n| 0.8 | 1.25 |\\n| 1 | 1.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Density Functional Theory (DFT) was used for calculations and investigate the electronic properties such as band gap here.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. The band gap increases linearly as the Pb content (x) increases from 0 to 1.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Introducing 2% Sn vacancies increases SnO's band gap by 0.5 eV. This shows that defects can tune properties but often introduce disruptive gap states. In contrast, Pb alloying tunes the band gap linearly from 0.4 to 1.4 eV without creating mid-gap defect states, offering a cleaner method for electronic band engineering.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No Defect States: Alloying avoids introducing electronic trap states within the band gap, which can degrade charge carrier mobility and device efficiency.\\n\\nPredictable Tuning: The band gap changes linearly with composition, allowing for precise and predictable control over optical absorption and emission wavelengths.\"}]}]","bbox":[{"panel_id":"a","x":7,"y":1,"width":562,"height":474}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/David Sibanda et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":573,"height":481,"image_format":"jpeg","image_sha256":"966769b0c24ea10bffea8355aee527c12d8e7edb17c9b1eb5f0d2dbb6f86a9aa","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_17_figure_20.jpg","caption":"Figure 20: Fraction of nitrogen of deposited films at a temperature of  $1,800\\mathrm{K}$  with various nitrogen-to-aluminium flux ratios starting at 0.8–2.8 [195].","id":"train/atomic-layer-deposition/simulation-usecase/17/figure_20","sample_id":"atomic-layer-deposition/simulation-usecase/17/figure_20","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart image represents the fraction of Nitrogen(N) in terms of percentage as a function of various nitrogen-to-aluminium flux ratios (N : Al), values of a range from 0.8 to 2.8. The values of N fraction steadily seems to increase till 2 and then saturates at close to 50% of N fraction. The image also shows the molecular representation computed using Molecular dynamics at temperature of 1800K\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| N : Al flux ratio | N fraction (%) |\\n|---|---|\\n| 0.8 | 42 |\\n| 1.2 | 44 |\\n| 1.6 | 49 |\\n| 2.0 | 51 |\\n| 2.4 | 51 |\\n| 2.8 | 52 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Growth Rate: Decreases as temperature increases.\\n\\nCrystallinity: Enhances significantly up to 1,800 K, then saturates between 1,800 K and 2,000 K.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The N-deficient route (0.8) is more detrimental. The text explicitly links good crystallinity to a \\\"near 50% nitrogen fraction,\\\" and the visuals confirm that significant N-deficiency destroys the lattice order.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A higher N:Al flux ratio delivers more N atoms per Al atom to the growth surface, increasing the N fraction in the film toward the ideal 50%. The caption states that good crystallinity is linked to this near-stoichiometric composition. The cause is that a stoichiometric film minimizes point defects (Al or N vacancies) and anti-site defects, allowing atoms to find proper lattice positions and form a well-ordered crystal.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies the reactor must be capable of delivering a highly N-rich environment (flux ratio >2:1) to achieve stoichiometric films. For a deep-UV LED, the near-perfect crystallinity resulting from this stoichiometry is critical to minimize defect-related non-radiative recombination and ensure high internal quantum efficiency for efficient light emission.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":650,"height":486}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_20.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_20.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/David Sibanda et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":653,"height":495,"image_format":"jpeg","image_sha256":"d1edb38267ec92b1df9adbec0c890c9c9365ab2ea070a48275e1622d5d31013f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_17_figure_5.jpg","caption":"Figure 5: Density functional theory (GGA/PBE-D3) [96] of minimum energy paths (MEPs) needed to adsorb Hacac on (a) aluminium oxide  $(\\mathrm{Al}_2\\mathrm{O}_3)$  and (b) silicon dioxide  $(\\mathrm{SiO}_2)$  surfaces [52].","id":"train/atomic-layer-deposition/simulation-usecase/17/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/17/figure_5","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The diagram represent the Density Functional Theory (DFT)-calculated reaction energy values comparing the chemisorption of Hacac on Al₂O₃ surface. The diagrams show the activation energy (Eₐ) and reaction energy (ΔE) for the ligand-exchange reaction where Hacac replaces a surface hydroxyl group.\"},{\"panel_id\":\"b\",\"text\":\"The diagram represent the Density Functional Theory (DFT)-calculated reaction energy values comparing the chemisorption of Hacac on SiO₂surfaces. The diagrams show the activation energy (Eₐ) and reaction energy (ΔE) for the ligand-exchange reaction where Hacac replaces a surface hydroxyl group. The key difference is that the reaction is exothermic and low-barrier on Al₂O₃ (ΔE = -0.49 eV, Eₐ = 0.01/0.25 eV) but endothermic and high-barrier on SiO₂ (ΔE = +0.98 eV, Eₐ = 2.35 eV).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Reaction step | Energy (eV) |\\n| ------ | ------- |\\n| Physisorbed state | 0.05  |\\n| Transition State  | 0.25  |\\n|  Chemisorbed Product State | -0.49 |\"},{\"panel_id\":\"b\",\"text\":\"| Reaction step | Energy (eV) |\\n| ------ | ------- |\\n| Physisorbed state | 0.05  |\\n|  Transition State  | 2.25  |\\n|  Chemisorbed Product State | 0.98 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It implies the Al–O(acac) bond is significantly stronger than the Si–O(acac) bond relative to their respective starting surface OH bonds.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It suggests performing DFT calculations of the inhibitor's adsorption reaction energy (ΔE) and activation energy (Eₐ) on both target and non-target surfaces. A successful selective process will likely require a large thermodynamic and kinetic preference (exothermic/low-barrier on target, endothermic/high-barrier on non-target) similar to the Al₂O₃/SiO₂ case shown here. This computational screen can rapidly identify promising inhibitor candidates before costly experimentation.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Physisorbed State: Hacac near the surface (0 eV reference).\\n\\n, Transition State: Peak at ~0.25 eV (Eₐ).\\n\\n, Chemisorbed Product State: Al(acac)⁺ + H₂O(g) at -0.49 eV (ΔE).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The very low Eₐ on Al₂O₃ (0.25 eV) means Hacac reacts easily at low temperature, while the much higher Eₐ on SiO₂ (2.35 eV) means it barely reacts at all under the same conditions. This large gap ensures that at typical ALD temperatures (~250–300 °C), Hacac will strongly chemisorb on Al₂O₃ but remain essentially inactive on SiO₂. As a result, the process naturally becomes highly selective without needing extreme temperatures or additional steps.\"}]}]","bbox":[{"panel_id":"b","x":606,"y":3,"width":569,"height":583},{"panel_id":"a","x":2,"y":3,"width":590,"height":580}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/David Sibanda et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":1178,"height":595,"image_format":"jpeg","image_sha256":"b7acc8d9492b66000b6a4b822fdc5aae126e8272a594c40c83d99b7ac1c674b2","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_27_figure_6.jpg","caption":"Figure 6. Simulated (a) growth per cycle and (b) mass density of deposited  $\\mathrm{SiO_2}$  films as a function of the probability  $p_{\\mathrm{P2}}$ .","id":"train/atomic-layer-deposition/simulation-usecase/27/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/27/figure_6","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"shows a non-linear trend where the growth per cycle increases to a maximum around pP₂ = 0.25–0.5 and then decreases at higher probabilities.\"},{\"panel_id\":\"b\",\"text\":\"shows density steadily rising with increasing pP₂, indicating denser films when P₂-type surface terminations dominate.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| occupation probability p<sub>P2</sub>|growth per cycle [Å]|\\n|--------------------------------------|--------------------|\\n|0|2.2|\\n|0.25|2.7|\\n|0.5|2.6|\\n|0.75|2.5|\\n|1|2.0|\"},{\"panel_id\":\"b\",\"text\":\"| occupation probability p<sub>P2</sub>|density [g/cm<sup>3</sup>]|\\n|--------------------------------------|--------------------------|\\n|0|1.6|\\n|0.25|2.3|\\n|0.5|2.4|\\n|0.75|2.5|\\n|1|2.7|\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The trends show that lower to mid-range occupation probabilities (around pP₂ = 0.25–0.50) yield higher growth per cycle, which can improve deposition throughput. Meanwhile, the steady increase in density with higher pP₂ suggests that P₂-rich growth pathways produce more compact and potentially higher-quality SiO₂ films. By balancing these opposing trends, researchers can tune PEALD process conditions to achieve films with optimal thickness, uniformity, and mechanical stability, leading to better performance in microelectronic and optical applications\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The maximum growth per cycle occurs at pP₂ ≈ 0.25.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, the growth per cycle changes significantly with varying pP₂.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Growth per cycle (Å/cycle), Film density (g/cm³)\"}]}]","bbox":[{"panel_id":"a","x":38,"y":11,"width":562,"height":372},{"panel_id":"b","x":637,"y":9,"width":564,"height":373}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/27/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/27/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/27/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/27/Martin Becker et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"27","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":1203,"height":383,"image_format":"jpeg","image_sha256":"18fb91c64e56d826023d2f80411721f184ee6923ae3c39ec5a9c84b3b902b983","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_32_41507260afda38f50f2d4921e128ba2a885bf7c74e3e0c35c21b3df5b9ecfbf7.jpg","caption":"Fig. B1 - Experimental outlet pressure monitoring during various ALD cycles, as measured by the Pirani gauge.","id":"train/atomic-layer-deposition/simulation-usecase/32/41507260afda38f50f2d4921e128ba2a885bf7c74e3e0c35c21b3df5b9ecfbf7","sample_id":"atomic-layer-deposition/simulation-usecase/32/41507260afda38f50f2d4921e128ba2a885bf7c74e3e0c35c21b3df5b9ecfbf7","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the experimental outlet pressure over time for two substances, H₂O and TMA, with distinct peaks and troughs monitored using pirani gauge.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Outlet Pressure (Torr) |\\n|----------|--------------------------|\\n| 0        | 0.049                   |\\n| 5        | 0.075                   |\\n| 10       | 0.058                   |\\n| 15       | 0.075                   |\\n| 20       | 0.049                   |\\n| 25       | 0.075                   |\\n| 30       | 0.058                   |\\n| 35       | 0.075                   |\\n| 40       | 0.049                  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The experimental reactor outlet pressure was monitored via the Pirani gauge during various ALD cycles, as deposition took place. The base pressure of the process i.e the system pressure during the purge step of the ALD cycle, is set at 50 mTorr, measured via the Pirani gauge.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Pirani gauge is calibrated for nitrogen (N2), while the inert gas used for the ALD process is Ar. Typically, the pressure reading by a Pirani gauge is multiplied by a correction factor, different for each gas, in order to calibrate the measurements to the real pressure value. In this study, these correction factors are unknown\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In order to have a value for the correction factor of Ar, the reactor outlet pressure was measured, for a varying Ar inlet, using a constant butterfly valve opening at the outlet.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":1012,"height":743}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/41507260afda38f50f2d4921e128ba2a885bf7c74e3e0c35c21b3df5b9ecfbf7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/41507260afda38f50f2d4921e128ba2a885bf7c74e3e0c35c21b3df5b9ecfbf7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/reactorGP Gakis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"32","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":1017,"height":748,"image_format":"jpeg","image_sha256":"0b2e8e6803cd45447f8be26a9ac7072b34f50267ae33c4c742784898e811af55","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_38_figure_13.jpg","caption":"Figure 13. Multiscale CFD plots illustrating the surface coverage (\\%) progression as a function of time for (a) 0.1 mole fraction Hacac adsorption, (b) 0.5 mole fraction BDEAS adsorption, (c) 0.1 mole fraction  $\\mathrm{O_3}$  adsorption. Evaluated multiscale process times to reach full surface coverage were longer than observed mesoscopic model process times conducted from prior work [34].","id":"train/atomic-layer-deposition/simulation-usecase/38/figure_13","sample_id":"atomic-layer-deposition/simulation-usecase/38/figure_13","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Hacac surface coverage (%) over time during adsorption (0.1 mole fraction), reaching full coverage at ~3.77 s. The progression reflects non-linear adsorption behavior due to integration of macroscopic fluid dynamics with mesoscopic surface kinetics.\"},{\"panel_id\":\"b\",\"text\":\"BDEAS surface coverage (%) over time during adsorption (0.5 mole fraction), reaching full coverage at ~6.99 s. Slower adsorption is influenced by pressure depletion zones and reagent interactions in the fluid phase.\"},{\"panel_id\":\"c\",\"text\":\"O₃ surface coverage (%) over time during adsorption (0.1 mole fraction), reaching full coverage at ~3.72 s. The curves indicate non-linear adsorption behavior and are consistent with experimental dosage times reported in the literature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s)|Hacac Coverage (%)|\\n|---------|------------------|\\n|0.0|0|\\n|1.0|25|\\n|2.0|60|\\n|3.0|90|\\n|3.77|100|\"},{\"panel_id\":\"b\",\"text\":\"| Time (s)|BDEAS Coverage (%)|\\n|---------|------------------|\\n|0.0|0|\\n|2.0|20|\\n|4.0|50|\\n|6.0|85|\\n|6.99|100|\"},{\"panel_id\":\"c\",\"text\":\"| Time (s)|O₃ Coverage (%)|\\n|---------|---------------|\\n|0.0|0|\\n|1.0|40|\\n|2.0|80|\\n|3.0|95|\\n|3.72|100|\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"By analyzing the coverage progression over time, researchers can adjust operating conditions such as reagent dosage, surface pressure, and wafer exposure time to ensure full coverage without overexposure. Understanding that BDEAS requires longer process times due to slower adsorption allows for precise control of deposition thickness and prevents oversaturation. Additionally, recognizing the impact of pressure depletion zones and species interactions informs reactor design modifications, such as improved purge streams, to maintain reagent availability and optimize surface coverage kinetics for all adsorption steps.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The BDEAS step requires the longest process time, reaching full coverage at 6.99 seconds.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The integration introduces pressure depletion zones and species interactions that reduce adsorption rates, leading to longer process times in the multiscale CFD simulations compared to prior mesoscopic models.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Pressure depletion zones in the reactor, Reagent dilution and reduced surface exposure, Repulsive interactions between species in the fluid phase, Partitioning of the wafer into discrete sections for kMC simulations, which reduces spatial resolution\"}]}]","bbox":[{"panel_id":"a","x":4,"y":9,"width":498,"height":385},{"panel_id":"b","x":547,"y":9,"width":513,"height":383},{"panel_id":"c","x":273,"y":488,"width":504,"height":387}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/images/figure_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/images/figure_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/Sungil Yun et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":1061,"height":930,"image_format":"jpeg","image_sha256":"4626101f5fdf3e62753d1c2996be95e3d9416f8992e6463026ac598c8126020b","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_38_figure_7.jpg","caption":"Figure 7. Hacac partial pressure over time on a node in the mesh. As the wafer rotates, the pressure rises around  $0.8\\mathrm{s}$  to reach the plateau and goes to 0 around  $3.6\\mathrm{s}$ . The length of the straight line in red is defined as the exposure time.","id":"train/atomic-layer-deposition/simulation-usecase/38/figure_7","sample_id":"atomic-layer-deposition/simulation-usecase/38/figure_7","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Figure 7 shows a line chart of Hacac partial pressure at a mesh node over time as a wafer rotates. The pressure rises around 0.8 s, reaches a plateau, and falls to zero around 3.6 s. The red line segment highlights the exposure time, representing the duration a wafer section is exposed to the reactant. This figure demonstrates how wafer rotation and reaction zone geometry determine exposure time distribution, which is crucial for achieving uniform surface coverage.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s)|Hacac Pressure (Pa)|\\n|---------|-------------------|\\n|0|0|\\n|0.8|60|\\n|1.5|60|\\n|3.6|0|\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Wafer rotation speed (angular velocity), Geometry of the reaction zone, Position of the wafer in the reactor, Distribution and flow of reactant gases\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. Increasing the angular velocity reduces the exposure time for each wafer section, while decreasing it extends the exposure time, affecting reactant coverage uniformity.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The exposure time is determined by the duration a wafer section is exposed to the reactant as it passes through a reaction zone while mounted on the rotating plate. The Hacac partial pressure rises to a plateau as the section enters the reaction zone and drops to zero as it exits. This timing depends on the wafer’s angular velocity and the shape of the reaction zone, ensuring each section receives controlled exposure for proper surface coverage.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Hacac partial pressure rises to the plateau around 0.8 s and drops back to zero around 3.6 s.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":495,"height":392}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/Sungil Yun et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":495,"height":392,"image_format":"jpeg","image_sha256":"15c0a8949f2ad8e91a8f7a25fa06e0780a90994c205cc27dfbf2032da9356568","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_43_figure_3.jpg","caption":"Figure 3. (a) In situ SE monitoring of  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD on the  $\\mathrm{Ge:H}$  surface; the inset enlarges the segment in the green box. (b) TMA adsorption and growth per cycle as a function of ALD cycles; the TMA adsorption baseline (red dash line) is  $\\sim 2.2\\mathrm{\\AA}$ , and the GPC baseline (blue dash line) is  $\\sim 0.86\\mathrm{\\AA}$ . (c) AFM images at different  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD stages.","id":"train/atomic-layer-deposition/simulation-usecase/43/figure_3","sample_id":"atomic-layer-deposition/simulation-usecase/43/figure_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"multi-axis chart"},{"panel_id":"c","label":"image panel"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The plots show how an Al₂O₃ film grows on Ge during atomic layer deposition (ALD) using TMA and H₂O. The chart a tracks film thickness vs time, Film thickness increases nearly linearly with time, with small periodic steps corresponding to individual ALD cycles driven by alternating TMA and H₂O pulses on Ge(100). The inset highlights that each TMA/H₂O sequence produces a distinct thickness increment, consistent with self‑limiting layer‑by‑layer ALD growth.\"},{\"panel_id\":\"b\",\"text\":\"the right multi axis chart  tracks TMA adsorption and growth per cycle vs ALD cycles. TMA adsorption (red circles, left axis) steadily increases with cycle number but remains below a marked saturation baseline, indicating sub‑monolayer adsorption per cycle. Growth per cycle (blue squares, right axis) also increases with cycle number yet stays below its baseline, implying that while the process becomes more efficient over time, it does not reach the ideal saturated growth rate within 30 cycles.\"},{\"panel_id\":\"c\",\"text\":\"Microscopy images of Ge:H surface at different ALD cycle counts, showing increasing roughness with more cycles.{ Ge:H -> 0.294 nm , 15 ALD Cycles -> 0.439 nm , 27 ALD Cycles -> 0.626 nm }\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (min) | Thickness (Å) |\\n|---|---|\\n| 0 | 0.8 |\\n| 3 | 1.5|\\n| 6 | 2.3 |\\n| 9 | 4.3 |\\n| 12 | 7|\\n| 15 | 10 |\"},{\"panel_id\":\"b\",\"text\":\"| ALD cycles | y | TMA adsorption (Å)/Growth per cycle (Å) |\\n|---|---|---|\\n| 0 | 0 |TMA adsorption (Å)  |\\n| 0 | 0 |Growth per cycle (Å) |\\n| 5 | 0.2 | TMA adsorption (Å) |\\n| 5 | 0.0076 | Growth per cycle (Å) |\\n| 10 | 0.5 | TMA adsorption (Å)|\\n| 10 | 0.017 | Growth per cycle (Å) |\\n| 15 | 0.9 | TMA adsorption (Å) |\\n| 15 | 0.2 | Growth per cycle (Å) |\\n| 20 | 1.5 | TMA adsorption (Å) |\\n| 20 | 0.4 | Growth per cycle (Å) |\\n| 25 | 2.0 | TMA adsorption (Å) |\\n| 25 | 0.60 | Growth per cycle (Å) |\\n| 30 | 2.0 | TMA adsorption (Å) |\\n| 30 | 0.8 | Growth per cycle (Å) |\"},{\"panel_id\":\"c\",\"text\":\"| Image | RMS (nm) |\\n|---|---|\\n| Ge:H | 0.294 |\\n| 15 ALD Cycles | 0.439 |\\n| 27 ALD Cycles | 0.626 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A trimethylaluminum (TMA) pulse and a water (H₂O) pulse are alternated. An inert gas purge after each pulse removes excess precursor, ensuring the surface reactions are self-limiting.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The incubation delay suggests non-uniform nucleation, which often results in a rough interface and high Interface Trap Density . This scattering can degrade carrier mobility and increase leakage current in the final device.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1   Ge:H -> 0.294 nm\\n2   15 ALD Cycles  ->0.439 nm\\n3   27 ALD Cycles ->  0.626 nm \\nAs ALD cycles increases clearly surface roughness seems to increase\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The near-zero initial growth indicates substrate inhibition, where TMA reacts poorly with the Ge surface compared to an Al2O3 surface. The rising trend implies the transition from isolated \\\"island\\\" nucleation to the coalescence of a continuous film layer.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":29,"width":261,"height":218},{"panel_id":"b","x":306,"y":24,"width":314,"height":227},{"panel_id":"c","x":1,"y":276,"width":622,"height":233}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/Yuanxia Zheng et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"43","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":625,"height":511,"image_format":"jpeg","image_sha256":"a141efa52ff72801217a42650663d202e6d5b97c5b36a361571a41178f37f66e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_43_figure_6.jpg","caption":"Figure 6.  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD  $(T = 270^{\\circ}\\mathrm{C})$  on  $\\mathrm{GeO}_x(5\\mathrm{\\AA}) / \\mathrm{Ge}(100)$ . (a) In situ SE monitoring of  $\\mathrm{Al}_2\\mathrm{O}_3$  thickness. (b) TMA adsorption and growth per cycle (GPC) as a function of ALD cycles extracted from SE measurement; the TMA adsorption baseline (red dash line) is  $\\sim 2.2\\mathrm{\\AA}$  and the GPC baseline (blue dash line) is  $\\sim 0.86\\mathrm{\\AA}$ . (c) GPC/TMA-adsorption ratio extracted from SE; the ratio baseline (green dash line) is  $\\sim 0.4$","id":"train/atomic-layer-deposition/simulation-usecase/43/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/43/figure_6","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"multi-axis chart"},{"panel_id":"c","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This line chart shows the stepwise increase in film thickness over time during Al₂O₃ Atomic Layer Deposition (ALD) using TMA and H₂O on a GeOₓ/Ge substrate, with each thickness jump corresponding to one ALD cycle.\"},{\"panel_id\":\"b\",\"text\":\"This multi axis chart shows the TMA dose peak and growth per cycle over ALD cycles. The values seems to dip initally till 3 ALD cycles and increased slowly.\"},{\"panel_id\":\"c\",\"text\":\"This  line chart shows the evolution of Growth Per Cycle (GPC) for TMA adsorption over the first 15 ALD cycles, starting from a baseline. The GPC increases from an initial low value before stabilizing, indicating a nucleation period before steady-state growth is achieved.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (min) | Thickness (Å) |\\n|------------|----------------|\\n| 0  | 0   |\\n| 2  | 1.8 |\\n| 4  | 2.7 |\\n| 6  | 5   |\\n| 8  | 8.2 |\\n| 10 | 10  |\"},{\"panel_id\":\"b\",\"text\":\"| ALD Cycles | Y (Å) | Parameter |\\n|------------|--------|------------------------|\\n| 0  | 0    | TMA dose peak (Å) |\\n| 0  | 0    | Growth per cycle (Å) |\\n| 3  | 1.32 | TMA dose peak (Å) |\\n| 3  | 0.04 | Growth per cycle (Å) |\\n| 6  | 1.29 | TMA dose peak (Å) |\\n| 6  | 0.29 | Growth per cycle (Å) |\\n| 9  | 1.9  | TMA dose peak (Å) |\\n| 9  | 0.5  | Growth per cycle (Å) |\\n| 12 | 2.21 | TMA dose peak (Å) |\\n| 12 | 0.86 | Growth per cycle (Å) |\\n| 15 | 2.2  | TMA dose peak (Å) |\\n| 15 | 0.86 | Growth per cycle (Å) |\"},{\"panel_id\":\"c\",\"text\":\"| ALD Cycles | GPC / TMA-Adsorp. |\\n|------------|--------------------|\\n| 0  | 0    |\\n| 3  | 0.04 |\\n| 6  | 0.23 |\\n| 9  | 0.30 |\\n| 12 | 0.40 |\\n| 15 | 0.37 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The sequence is a TMA precursor pulse followed by a H₂O oxidant pulse. A purge step with inert gas separates each pulse to remove excess precursor and byproducts, enabling self-limiting surface reactions.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The reaction on the dangling bond is significantly more exothermic (-1.74 eV) than on the H-terminated surface (-1.09 eV).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It demonstrates linear, layer-by-layer growth with a constant growth per cycle (GPC). This implies the surface reactions are self-limiting and saturate each cycle, leading to precise thickness control.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This selectivity leads to island growth rather than layer-by-layer deposition, as nucleation occurs exclusively at defect sites first. This non-uniform nucleation creates pinholes and a rough interface, which typically results in high interface trap densities and increased gate leakage current in the final device.\"}]}]","bbox":[{"panel_id":"a","x":7,"y":35,"width":199,"height":248},{"panel_id":"b","x":216,"y":36,"width":248,"height":248},{"panel_id":"c","x":475,"y":40,"width":191,"height":245}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/43/Yuanxia Zheng et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"43","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":667,"height":284,"image_format":"jpeg","image_sha256":"404e48b113630df932745092c5f8809dfbdc46de4307b604d14a07c5efcd53a4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_8_figure_10.jpg","caption":"Figure 10. Simulated mass change as a function of time from the kMC model (left) and the corresponding experimental QCM measurement adopted from ref 25 (right). The different phases (DEZ pulse/purge/  $\\mathrm{H}_2\\mathrm{O}$  pulse/purge) within the ALD cycle are indicated by the vertical dashed lines. The  $\\mathrm{ZnO}$  thin films are deposited at temperature  $177^{\\circ}\\mathrm{C}$  in both the simulation and the experiment.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_10","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_10","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the mass deposited per unit area over time during a ZnO ALD process according to simulations. It includes labeled phases such as 'Dose DEZ', 'Purge', 'Dose H<sub>2</sub>O', and 'Purge'.\"},{\"panel_id\":\"b\",\"text\":\"The chart illustrates the mass deposited over time during a ZnO ALD process according to experimental data. It includes labeled phases such as 'Dose DEZ', 'Purge', 'Dose H<sub>2</sub>O', and 'Purge'.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"|Time (s)| Mass deposited / ng cm<sup>-2</sup> | \\n|---|---|\\n| 0.0 | 155 |\\n|0.4| 155 |\\n|0.8|160|\\n|1.2|155|\\n|1.6|155|\"},{\"panel_id\":\"b\",\"text\":\"|Time (s)| Mass deposited / ng cm<sup>-2</sup> | \\n|---|---|\\n| 0 | 120 |\\n|2| 115 |\\n|4|115|\\n|6|110|\\n|7|120|\\n|8|115|\\n|10|115|\\n|12|115|\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The gradual increase is related to the reordering at the surface to reach perfect saturation, which will not happen right from the start. It could be that this is not incorporated in the model.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The duration of a single cycle is 12 s. It seems as if the purge after the water dose is longer than required.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.7 s\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It could be that water initially physisorbs at the surface after which the ligands are removed. Leading to a mass increase first.\"}]}]","bbox":[{"panel_id":"b","x":554,"y":7,"width":426,"height":427},{"panel_id":"a","x":1,"y":4,"width":515,"height":405}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/Timo Weckman et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":980,"height":434,"image_format":"jpeg","image_sha256":"13963c880061defe2c95cdb0d7b8591c7a08956312484133ce65dcac2e326738","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_2.jpg","caption":"Figure 2. DC self-bias under varying bias powers at the (a)  $\\mathrm{O_2}$  and (b)  $\\mathrm{BCl_3}$  plasma steps, and under varying ICP powers at the (c)  $\\mathrm{O_2}$  and (d)  $\\mathrm{BCl_3}$  plasma steps.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_2","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"line chart"},{"panel_id":"d","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between bias power and DC self-bias for O2 plasma step at ICP power of 400 W.\"},{\"panel_id\":\"b\",\"text\":\"The chart shows the relationship between bias power and DC self-bias for BCl3 plasma step at ICP power of 400 W.\"},{\"panel_id\":\"c\",\"text\":\"The chart shows the relationship between ICP power and DC self-bias for O2 plasma step at bias power of 3 W.\"},{\"panel_id\":\"d\",\"text\":\"The chart shows the relationship between ICP power and DC self-bias for BCl3 plasma step at bias power of 2 W.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Bias Power [W] | DC self-bias [V] |\\n|---|---|\\n| 0 |   |\\n| 1 |   |\\n| 2 |   |\\n| 3 | 5 |\\n| 4 | 10 |\\n| 5 | 15 |\\n| 6 | 20 |\"},{\"panel_id\":\"b\",\"text\":\"| Bias Power [W] | DC self-bias [V] |\\n|---|---|\\n| 0 |   |\\n| 1 |   |\\n| 2 | 5 |\\n| 3 | 10 |\\n| 4 | 13 |\\n| 5 | 17 |\\n| 6 | 21 |\"},{\"panel_id\":\"c\",\"text\":\"| ICP power [W] | DC Self-bias [V] |\\n|---|---|\\n| 250 |   |\\n| 300 |   |\\n| 350 |   |\\n| 400 | 4.0 |\\n| 450 | 4.1 |\\n| 500 | 4.2 |\"},{\"panel_id\":\"d\",\"text\":\"| ICP power [W] | DC Self-bias [V] |\\n|---|---|\\n| 150 | 11 |\\n| 200 | 10 |\\n| 250 | 8 |\\n| 300 | 6 |\\n| 350 | 4 |\\n| 400 | 4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"400 W.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It increases linearly.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It decreases linearly.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3 W.\"}]}]","bbox":[{"panel_id":"a","x":21,"y":8,"width":510,"height":407},{"panel_id":"b","x":578,"y":8,"width":526,"height":405},{"panel_id":"c","x":3,"y":455,"width":543,"height":404},{"panel_id":"d","x":573,"y":462,"width":540,"height":400}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_2.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":"line chart","caption_source":"content.json:img_caption"},"width":1114,"height":867,"image_format":"jpeg","image_sha256":"e7c6346018864fd107a2d0b7eefa8b491ecf0215d4cc941fa534113595868c16","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_6.jpg","caption":"Figure 6. Etch rate per cycle of the digital etching performed under (a) varying  $\\mathrm{O_2}$  plasma times and fixed HCl dipping time and (b) varying HCl dipping times and fixed  $\\mathrm{O_2}$  plasma time.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_6","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_6","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch rate increases with increasing O₂ plasma time up to 2 minutes, after which it plateaus.\"},{\"panel_id\":\"b\",\"text\":\"The etch rate increases with increasing dipping time in HCl up to 4 minutes, after which it plateaus.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| O₂ plasma time [min] | Etch rate [nm/cycle] |\\n|---|---|\\n| 0 | 0.0 |\\n| 1 | 0.3 |\\n| 2 | 0.5 |\\n| 3 | 0.5 |\\n| 4 | 0.5 |\"},{\"panel_id\":\"b\",\"text\":\"| Dipping time in HCl at 80 °C [min] | Etch rate [nm/cycle] |\\n|---|---|\\n| 0 | 0.0 |\\n| 2 | 0.4 |\\n| 4 | 0.5 |\\n| 5 | 0.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"5 minutes.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.45 nm/cycle.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"30 W.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.5 nm/cycle.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":572,"height":427},{"panel_id":"b","x":667,"y":7,"width":552,"height":426}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_6.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":"line chart","caption_source":"content.json:img_caption"},"width":1222,"height":434,"image_format":"jpeg","image_sha256":"d8152c77fe901418877268e7d5cf95e7a30cf31e6a708d55470ac4986541184d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_14_figure_2.jpg","caption":"Figure 2. Mass change versus time during  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE at  $200^{\\circ}\\mathrm{C}$  using HF and  $\\mathrm{Sn(acac)}_2$  as the reactants. Reproduced with permission from ref 5. Copyright 2015 American Chemical Society.","id":"train/atomic-layer-etching/experimental-usecase/14/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/14/figure_2","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the change in mass over time. A linear decrease in mass change can be seen over time at 200 °C on Al2O3. The process was performed with an exposure time of 1 s for Sn(acac)2 and 1 s of HF, with a 30 s purge between each step.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Change (ng/cm²) |\\n|---|---|\\n| 0 | 80 |\\n| 2000 | -200 |\\n| 4000 | -450 |\\n| 6000 | -700 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Hf and Sn(acac)2.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The mass change decreases linearly with increasing time.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experiment was performed at 200 °C.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"-8.4 ng/(cm2 cycle)\"}]}]","bbox":[{"panel_id":"a","x":5,"y":-1,"width":660,"height":517}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/Mechanisms of Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"14","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":669,"height":525,"image_format":"jpeg","image_sha256":"ae3dd52e2be2fdcde7bbdc16286ced01d799e0d9d5bd9489c9d99f3dc9964af0","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_11.jpg","caption":"Figure 11. Mass change versus time during  $\\mathrm{VO}_2$  ALE using  $\\mathrm{SF}_4$  and  $\\mathrm{Sn(acac)}_2$  as the reactants at  $200^{\\circ}\\mathrm{C}$ . Etch rate is  $0.30\\mathrm{\\AA}$  /cycle.","id":"train/atomic-layer-etching/experimental-usecase/17/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_11","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the mass change during atomic layer etching (ALE) of VO₂ at 200 °C over 50 cycles. A steady linear decrease in mass is observed, reaching approximately –600 ng/cm² by 5000 seconds. The etch rate is specified as 0.30 Å/cycle, and the process involves alternating pulses of SF₄ and Sn(acac)₂, each with pulse and purge times indicated as \\\"3–55–3–55\\\". The linear profile confirms self-limiting and uniform removal per cycle, consistent with ideal ALE behaviour.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Change (ng/cm²) |\\n|----------|----------------------|\\n| 0        | 150                  |\\n| 1000     | 0                    |\\n| 2000     | -150                 |\\n| 3000     | -300                 |\\n| 4000     | -450                 |\\n| 5000     | -600                 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.30 Å/cycle.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SF₄: 3 s pulse, 55 s purge , Sn(acac)₂: 3 s pulse, 55 s purge\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear decrease in mass over time suggests a consistent and self-limiting etching mechanism, which is characteristic of atomic layer etching. The use of alternating SF₄ and Sn(acac)₂ exposures, each followed by a purge, appears to produce uniform mass loss per cycle. The defined etch rate of 0.30 Å/cycle confirms precise and repeatable removal of material. This behavior is essential for achieving nanoscale control in ALE processes, particularly in applications requiring high fidelity and minimal damage to underlying layers.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":664,"height":527}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":664,"height":527,"image_format":"jpeg","image_sha256":"899157dcc9797dae1c63eb1b0a8ca737b57b06a6d0ad9a39738c64f2606ee212","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_12.jpg","caption":"Figure 12. Expanded view of three  $\\mathrm{VO}_2$  ALE cycles from Figure 11 showing mass gain during the  $\\mathrm{SF}_4$  exposure and mass loss during  $\\mathrm{Sn(acac)}_2$  exposure.","id":"train/atomic-layer-etching/experimental-usecase/17/figure_12","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_12","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates mass change (in ng/cm²) as a function of time during the atomic layer etching (ALE) of VO₂ at 200°C. The etching process involves alternating exposures to SF₄ and Sn(acac)₂, as labeled by the arrows on the plot. Distinct mass loss and gain events are observed during the SF₄ and Sn(acac)₂ steps, respectively. The graph captures both the etching half-cycle (mass loss from SF₄) and the subsequent precursor adsorption step (mass gain from Sn(acac)₂), culminating in a repeatable etch pattern. The label \\\"MCPC\\\" denotes the minimum mass point per cycle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time(s) | Mass Change (ng/cm²) |\\n|---|---|\\n| 1150 | -30 |\\n| 1200 | -40 |\\n| 1250 | -50 |\\n| 1300 | -60 |\\n| 1350 | -70 |\\n| 1400 | -80 |\\n| 1450 | -90 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The sharp drop in mass during SF₄ exposure suggests that SF₄ acts as the primary etching agent in the ALE cycle. Its interaction with the VO₂ surface likely results in the formation and removal of volatile byproducts, leading to a measurable decrease in surface mass. This aligns with the expected function of SF₄ in selectively breaking bonds and removing material from the surface layer.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Mass increases during Sn(acac)₂ exposure, confirming its adsorption onto the surface.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SF₄, Sn(acac)₂\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":670,"height":553}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":670,"height":553,"image_format":"jpeg","image_sha256":"3bba86be54d1c4aaf2f9db6db38aa0e3550784fb44a2bcee0af8eaca3fb4a8b8","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_13.jpg","caption":"Figure 13. First three cycles of  $\\mathrm{VO}_2$  ALE using  $\\mathrm{SF}_4$  and  $\\mathrm{Sn(acac)}_2$  at  $200^{\\circ}\\mathrm{C}$  showing initial mass gain from fluorination during  $\\mathrm{SF}_4$  exposure and mass changes during the first  $\\mathrm{Sn(acac)}_2$  exposure and subsequent  $\\mathrm{SF}_4$  and  $\\mathrm{Sn(acac)}_2$  exposures.","id":"train/atomic-layer-etching/experimental-usecase/17/figure_13","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_13","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents a mass change profile over time during a VO₂ atomic layer etching (ALE) process conducted at 200 °C. The mass oscillations correspond to alternating exposures of SF₄ and Sn(acac)₂, clearly indicated by arrows. Each cycle begins with a mass loss during SF₄ exposure, suggesting etching, followed by a mass gain during Sn(acac)₂ dosing, indicative of precursor adsorption. The overall trend highlights the stepwise nature of the ALE process, where the mass change remains tightly controlled and periodic, showcasing the self-limiting behavior typical of atomic layer processing.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time(s) | Mass Change (ng/cm²) |\\n|---|---|\\n| 0 | 0 |\\n| 50 | 27 |\\n| 100 | 56 |\\n| 150 | 28 |\\n| 200 | 27 |\\n| 250 | 0 |\\n| 300 | -27 |\\n| 350 | -56 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"200°C\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SF₄  , Sn(acac)₂\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The periodic mass increases and decreases suggest a well-defined ALE process, where SF₄ causes mass loss during etching and Sn(acac)₂ leads to a mass gain through precursor adsorption. The regularity of the mass changes with each cycle indicates that the reactions are self-limiting and repeatable, hallmarks of a successful atomic layer etching mechanism. This confirms that the ALE process is tightly controlled at 200°C, producing predictable material removal and surface renewal with each precursor pulse.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":667,"height":539}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":667,"height":539,"image_format":"jpeg","image_sha256":"eb136fa43deb59d03ecbbb8a561ab5073604eb33f43f15f6a97819fc7d722bac","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_14.jpg","caption":"Figure 14. Mass gains during fluorination of the initial  $\\mathrm{VO}_2$  ALD film with either  $\\mathrm{SF}_4$  or HF.  $\\mathrm{SF}_4$  exposure at  $500 \\mathrm{mTorr}$  leads to a higher mass gain than HF exposures at either 60 or  $500 \\mathrm{mTorr}$ .","id":"train/atomic-layer-etching/experimental-usecase/17/figure_14","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_14","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows mass gain over time during fluorination of VO₂ at 200 °C using three different conditions: SF₄ at ~500 mTorr, HF at ~500 mTorr, and HF at ~60 mTorr. All three conditions exhibit a sharp increase in mass at around 30 s, indicating fluorine uptake, followed by a plateau, suggesting saturation. SF₄ leads to the highest mass increase (~32 ng/cm²), followed by HF at higher pressure (~18 ng/cm²) and HF at lower pressure (~10 ng/cm²). These differences highlight the influence of both fluorine source and pressure on the extent of fluorination.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | SF₄ ~500 mTorr | HF ~500 mTorr | HF ~60 mTorr |\\n|----------|----------------|----------------|---------------|\\n| 0        | 0              | 0              | 0             |\\n| 15       | 0              | 0              | 0             |\\n| 30       | 32             | 18             | 10            |\\n| 45       | 32             | 18             | 10            |\\n| 60       | 32             | 18             | 10            |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SF₄ at ~500 mTorr.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SF₄ at ~500 mTorr  , HF at ~500 mTorr  , HF at ~60 mTorr\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure shows that higher pressure significantly enhances the effectiveness of HF as a fluorination agent. At 500 mTorr, HF leads to a mass increase of approximately 18 ng/cm², while at 60 mTorr, the mass gain drops to about 10 ng/cm². This suggests that greater precursor availability at higher pressures improves surface fluorination. The consistent plateau observed in both HF traces further indicates that surface saturation is still achieved, but the extent of uptake is pressure-dependent.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":667,"height":541}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":667,"height":541,"image_format":"jpeg","image_sha256":"da27dbdb34182370199d1867239118d24fc924c9fcbc3641db5eaf671e12ab25","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_2.jpg","caption":"Figure 2. Expanded view of three  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE cycles from Figure 1 showing mass gain during the  $\\mathrm{SF}_4$  exposure and mass loss during  $\\mathrm{Sn(acac)}_2$  exposure.","id":"train/atomic-layer-etching/experimental-usecase/17/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_2","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This line chart presents mass change data during Al₂O₃ atomic layer etching (ALE) at 200°C, showing stepwise decreases in mass following alternating exposures to SF₄ and Sn(acac)₂. Distinct mass loss events (ΔM) occur upon each SF₄ pulse, indicating volatile byproduct formation from fluorination. The subsequent Sn(acac)₂ exposures yield mass gains, consistent with precursor adsorption. The black dashed arrow labeled MCPC highlights a stable mass plateau following precursor saturation, marking the point of maximum chemisorbed precursor coverage.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Change (ng/cm²) |\\n|----------|----------------------|\\n| 13700    | -250                 |\\n| 13720    | -255                 |\\n| 13740    | -257                 |\\n| 13760    | -258                 |\\n| 13780    | -260                 |\\n| 13800    | -263                 |\\n| 13820    | -265                 |\\n| 13840    | -266                 |\\n| 13860    | -268                 |\\n| 13880    | -269                 |\\n| 13900    | -270                 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"MCPC refers to the Maximum Chemisorbed Precursor Coverage.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SF₄ removes surface oxygen , Sn(acac)₂ replenishes surface metal sites\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The mass profile displays repeated and saturating steps, where each pulse of SF₄ causes a predictable mass loss (etch step), followed by a stable plateau after Sn(acac)₂ adsorption. The return to similar mass levels after each Sn(acac)₂ exposure indicates that surface sites become fully occupied and no further mass uptake occurs until the next etch step. This cyclical pattern, along with the MCPC plateau, supports the self-limiting mechanism characteristic of ALE, ensuring precise and controlled etch per cycle behaviour.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":669,"height":539}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":669,"height":539,"image_format":"jpeg","image_sha256":"52fc7227f4e63ab18cdcfa1be1e6ed28670d6bdffec458144c80f86bc7965ed2","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_5.jpg","caption":"Figure 5. First three cycles of  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE using  $\\mathrm{SF}_4$  and  $\\mathrm{Sn(acac)}_2$  at  $200^{\\circ}\\mathrm{C}$  showing initial fluorination mass gain during  $\\mathrm{SF}_4$  exposure, subsequent mass changes during the first  $\\mathrm{Sn(acac)}_2$  exposure, and subsequent  $\\mathrm{SF}_4$  and  $\\mathrm{Sn(acac)}_2$  exposures.","id":"train/atomic-layer-etching/experimental-usecase/17/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_5","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart presents the mass change over time during atomic layer etching (ALE) of Al₂O₃ at 200 °C. The process alternates between SF₄ and Sn(acac)₂ exposures, with each reagent inducing distinct mass gain or loss steps. The mass increases sharply following each Sn(acac)₂ dose and decreases following SF₄ exposure, consistent with a ligand-exchange-driven etch mechanism. The repeatability of these mass shifts across cycles demonstrates the controlled and cyclical nature of the ALE process.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Change (ng/cm²) |\\n|----------|----------------------|\\n| 0        | 0                    |\\n| 50       | 40                   |\\n| 100      | 80                   |\\n| 150      | 60                   |\\n| 200      | 100                  |\\n| 250      | 60                   |\\n| 300      | 100                  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The mass change profile shows distinct, repeatable mass gains upon Sn(acac)₂ exposure and mass losses upon SF₄ exposure, consistent with a two-step self-limiting reaction mechanism. This cyclic pattern confirms that Al₂O₃ undergoes a controlled etching process. Sn(acac)₂ deposits organic ligands or intermediate complexes on the surface, increasing mass, while SF₄ likely facilitates removal of fluorinated byproducts, reducing mass. The regularity of these changes demonstrates a stable and reproducible ALE process.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"During Sn(acac)₂ exposure.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Reproducible cyclic mass change, distinct etch and deposition phases, stable magnitude across cycles, consistent thermal conditions at 200 °C\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":669,"height":533}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":669,"height":533,"image_format":"jpeg","image_sha256":"552079d405aa98307b4215a07a6866659fd2f37b3d52f1b334941ff7ac51299e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_8.jpg","caption":"Figure 8. Mass change versus time during 125 cycles of  $\\mathrm{VO}_2$  ALD at  $150^{\\circ}\\mathrm{C}$  using TEMAV and  $\\mathrm{H}_2\\mathrm{O}$  as the reactants. Growth rate is  $0.63 \\mathrm{Å / cycle}$ .","id":"train/atomic-layer-etching/experimental-usecase/17/figure_8","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_8","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This line graph displays the mass change of a VO₂ film during atomic layer deposition (ALD) at 150 °C over the course of 125 cycles. The increase in mass is linear with respect to time, indicating a highly consistent and reproducible deposition process. The graph annotates a growth rate of 0.63 Å/cycle, and the pulse sequence for the precursors (TEMAV and H₂O) is noted as part of the experimental procedure. The steady mass gain confirms the self-limiting, layer-by-layer nature of ALD and supports the reliability of the chosen conditions for high-quality VO₂ film growth.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Change (ng/cm²) |\\n|----------|-----------------------|\\n| 0        | 0                     |\\n| 5000     | 1400                  |\\n| 10000    | 2100                  |\\n| 15000    | 2800                  |\\n| 20000    | 3500                  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear mass gain over time confirms that the VO₂ ALD process at 150 °C is self-limiting and highly reproducible. Each ALD cycle contributes a consistent amount of mass, as shown by the steady slope, which reflects a uniform deposition rate across the 125 cycles. This suggests that both precursor pulses, TEMAV and H₂O, are saturating the surface reactions efficiently, enabling the controlled growth of conformal VO₂ films. The annotated growth rate of 0.63 Å/cycle further validates the precision and scalability of the process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.63 Å/cycle.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Linear, stable mass gain, constant growth rate, efficient precursor saturation\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":672,"height":564}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":672,"height":564,"image_format":"jpeg","image_sha256":"673fc6390596014339d486ed850d3d3b6802afe33dd5c6f1a5c9eac31230f00c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_22_figure_1.jpg","caption":"Figure 1. (a) Mass changes observed by in situ QCM during growth and subsequent etching of  $\\mathrm{MoS}_2$  films on alumina at  $200^{\\circ}\\mathrm{C}$ . The blue arrow indicates the region of  $\\mathrm{MoS}_2$  ALD, and the orange arrow indicates the region of  $\\mathrm{MoS}_2$  ALE. (b) In situ QCM data during two cycles of  $\\mathrm{MoS}_2$  ALE. A mass gain of  $\\sim 42\\mathrm{ng} / \\mathrm{cm}^2$  can be measured after the  $\\mathrm{MoF}_6$  dose, followed by a net mass loss of  $19\\mathrm{ng} / \\mathrm{cm}^2$  following the  $\\mathrm{H}_2\\mathrm{O}$  dose. Orange and blue markers at the bottom of the figure indicate  $\\mathrm{MoF}_6$  and  $\\mathrm{H}_2\\mathrm{O}$  doses, respectively.","id":"train/atomic-layer-etching/experimental-usecase/22/figure_1","sample_id":"atomic-layer-etching/experimental-usecase/22/figure_1","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart depicts mass change during sequential deposition and etching stages over elapsed time. The deposition phase (blue arrow) shows a steady increase in mass, while the etching phase (orange arrow) leads to a gradual decrease, indicating material removal.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates real-time mass changes during alternating exposures to H₂O and MoF₆, visualised through stepwise increases and decreases. Annotated bars indicate the timing of each precursor pulse, and labeled Δ values quantify the mass gain and loss per cycle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Elapsed Time (s) | Mass Change (ng/cm²) |\\n|------------------|----------------------|\\n| 0                | 0                    |\\n| 5                | ~13                  |\\n| 10               | ~26                  |\\n| 15               | ~39                  |\\n| 17               | ~42 (peak)           |\\n| 20               | ~37                  |\\n| 25               | ~25                  |\\n| 30               | ~12                  |\\n| 35               | ~0                   |\"},{\"panel_id\":\"b\",\"text\":\"| Elapsed Time (s) | Mass Change (ng/cm²) |\\n|------------------|----------------------|\\n| 0                | 0                    |\\n| 10               | ~42                  |\\n| 12               | ~23                  |\\n| 20               | ~65                  |\\n| 22               | ~46                  |\\n| 30               | ~88                  |\\n| 32               | ~69                  |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"H₂O, MoF₆\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Panel (b) shows mass gain and loss in response to alternating pulses of H₂O and MoF₆. The plateaus between each pulse and the consistent magnitude of change suggest that the surface reactions saturate at each step, characteristic of self-limiting behavior in atomic layer etching (ALE). The consistent Δ mass values following each precursor exposure reinforce the interpretation that a stoichiometric, repeatable reaction occurs with each cycle, indicative of controlled surface etching rather than continuous etching or deposition.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 42 ng/ cm²\"}]}]","bbox":[{"panel_id":"a","x":38,"y":16,"width":469,"height":358},{"panel_id":"b","x":35,"y":402,"width":471,"height":351}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/Thermal Atomic Layer Etching of MoS2 Using MoF6 and H2O.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":509,"height":756,"image_format":"jpeg","image_sha256":"ddbeac8b76fa32075689025da3ed985686d7710c74b976a717f706dc8e3409ce","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_22_figure_2.jpg","caption":"Figure 2. QCM saturation curves for precursors (a)  $\\mathrm{MoF}_6$  and (b)  $\\mathrm{H}_2\\mathrm{O}$  during ALE at  $200^{\\circ}\\mathrm{C}$ . ALE timing sequence followed  $X / 20 / 3.5 / 20$  and  $1 / 20 / X / 20$  s for  $\\mathrm{MoF}_6$  and  $\\mathrm{H}_2\\mathrm{O}$ , respectively, where  $X$  was 0.5, 1.0, 1.5, 2.0, and 3.0 (and 4.0 for  $\\mathrm{H}_2\\mathrm{O}$ ). Figure 3. QCM measurements of averaged MCPC between etching temperatures of  $100 - 300^{\\circ}\\mathrm{C}$  using the timing sequence  $1 - 20 - 2.5 - 20^{\\circ}\\mathrm{C}$ . The mass loss increases dramatically with the etching temperature.","id":"train/atomic-layer-etching/experimental-usecase/22/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/22/figure_2","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart displays the effect of temperature on the mass change per cycle during a process governed by the recipe \\\"1/20/2.5/20\\\". As the temperature increases from 100°C to 300°C, the mass change becomes increasingly negative, indicating greater mass loss per cycle at higher temperatures. This trend suggests that etching becomes more efficient or aggressive as the temperature rises, especially beyond 200°C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Mass Change Per Cycle (ng/cm²) |\\n|------------------|-------------------------------|\\n| 100              | 0                             |\\n| 150              | -10                           |\\n| 200              | -20                           |\\n| 250              | -100                          |\\n| 300              | -250                          |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"-100 ng/cm²\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"100°C, 150°C, 200°C\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure demonstrates that increasing temperature leads to a more significant negative mass change per cycle, indicating enhanced etching. From 100°C to 200°C, the mass loss increases gradually. However, beyond 200°C, the rate of mass loss accelerates sharply, with the most substantial decrease observed at 300°C. This suggests a temperature threshold beyond which the etching process becomes markedly more aggressive, possibly due to increased reactivity or volatilisation of etch products.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":567,"height":433}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/Thermal Atomic Layer Etching of MoS2 Using MoF6 and H2O.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":567,"height":433,"image_format":"jpeg","image_sha256":"19718f292ac0524baf94ee816bc16fd8c981698c9d20f3cefde6e95924f3c0b3","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_23_figure_16.jpg","caption":"Figure 16. Average RMS roughness of  $\\mathrm{ZnS}$  ALD films grown at  $100^{\\circ}\\mathrm{C}$  and after  $\\mathrm{ZnS}$  ALD at  $300^{\\circ}\\mathrm{C}$ . The  $\\mathrm{ZnS}$  film thickness before  $\\mathrm{ZnS}$  ALD was  $30 \\mathrm{nm}$ .","id":"train/atomic-layer-etching/experimental-usecase/23/figure_16","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_16","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between the number of ALE cycles and the RMS roughness of a material. As the number of ALE cycles increases, the RMS roughness initially decreases slightly and then sharply increases.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of ALE Cycles | RMS Roughness (Å) |\\n|---|---|\\n| 0 | 7.26 |\\n| 20 | 6.05 |\\n| 40 | 7.74 |\\n| 60 | 10.77 |\\n| 80 | 17.94 |\\n| 100 | 25 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1.  Tracking roughness is critical because it is a direct indicator of etch uniformity and mechanism; increasing roughness signals loss of layer-by-layer control and the onset of undesirable, non-ideal etching that can degrade device performance. Perform 20 ALE cycles on the sample.\\n\\n2. Transfer the sample to an atomic force microscope (AFM).\\n\\n3. Measure and calculate the RMS roughness over a defined area.\\n\\n4. Repeat steps 1-3 for the next set of cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For an atomically smooth final surface, this ALE process is fundamentally limited. The data shows it cannot be run indefinitely; the roughening after ~60 cycles defines a hard limit on how much material can be removed while maintaining smoothness, unlike an ideal isotropic etch that may maintain a constant roughness.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The trend shows an initial smoothing stage (roughness decreases slightly) followed by a severe roughening stage. The sharp increase likely occurs when the ALE process breaks down, transitioning from a self-limiting, uniform reaction to a non-uniform one where etchant preferentially attacks defects, grain boundaries, or chemically distinct regions, creating a pitted and rough morphology.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies the recipe is only suitable for shallow etching. To etch a deep trench, you would need many more than 60 cycles, but the severe post-60-cycle roughening would make the trench sidewalls and bottom unacceptably rough, increasing scattering loss and harming device performance. The process must be modified or a different chemistry found for deep etches.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":669,"height":560}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_16.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_16.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/Thermal Atomic Layer Etching of Zinc Sulfide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"23","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":672,"height":566,"image_format":"jpeg","image_sha256":"98f2b4f9aa991c95dcfbedd66b27f4b32165fe72dedbdfafb912e48121e88f6e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_23_figure_3.jpg","caption":"Figure 3. Thickness of  $\\mathrm{ZnS}$  film versus number of  $\\mathrm{ZnS}$  ALE cycles using HF and TMA as reactants at  $300^{\\circ}\\mathrm{C}$ . Initial  $\\mathrm{ZnS}$  ALD films were grown at  $100^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/23/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line plot represents the  thickness of ZnS film measured against number of ZnS ALE cycles using HF and TMA as reactants at 300 °C. Initial ZnS ALD films in this study were grown at 100 °C. The etch rate is approximately 2 Å per cycle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of ALE Cycles | ZnS Thickness (Å) |\\n|---|---|\\n| 0 | 200 |\\n| 10 | 180 |\\n| 20 | 160 |\\n| 30 | 140 |\\n| 40 | 120 |\\n| 50 | 100 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. HF Pulse: Reacts with the ZnS surface to convert it to a volatile zinc fluoride (e.g., ZnF₂) and a sulfur-containing layer, or to form a modified, fluorinated surface.\\n\\n2.  TMA Pulse: Reacts with the modified surface (e.g., with sulfur or fluorine) to produce volatile byproducts (like AlF₃, (CH₃)₂S), completing the removal of a thin layer. The TMA step acts as the co-reactant that removes the modified surface species left after HF exposure, regenerating a reactive surface for the next HF pulse and enabling the cyclic, self-limiting removal.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For damage-free etching over sensitive materials, the thermal ALE process is strongly preferable. The trade-off is potentially slower etch rates and more specific chemistry development, but it preserves the integrity of delicate sub-layers.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ALD at 100°C typically produces a less dense, more disordered, and potentially impurity-containing ZnS film. This porous, defective microstructure presents a higher surface area and more reactive sites for the HF and TMA etchants to attack, leading to a higher removal per cycle compared to a denser film deposited at a higher temperature.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This enables the fabrication of high-precision optical nanostructures like waveguides, photonic crystals, or etch-stop layers. The atomic-level control allows engineers to define features with vertical sidewalls and exact depths, which is critical for controlling light propagation, resonant modes, and interference effects in miniaturized photonic devices.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":662,"height":526}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/Thermal Atomic Layer Etching of Zinc Sulfide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"23","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":666,"height":533,"image_format":"jpeg","image_sha256":"21e88c0c955e9513fa45211bfe85798a11029cc2ff351b583b1181111da626aa","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_26_fig_3.jpg","caption":"Fig. 3. (a) Etching depth of the ALE process. (b) Etching level for etch step and ALE process. (c) Roughness image of MOCVD reference, (d)  $\\mathrm{Cl}_2$  plasma etching, and (e) ALE.","id":"train/atomic-layer-etching/experimental-usecase/26/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/26/fig_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"bar chart"},{"panel_id":"c","label":"image panel"},{"panel_id":"d","label":"image panel"},{"panel_id":"e","label":"image panel"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":0,"y":0,"width":259,"height":207},{"panel_id":"b","x":292,"y":16,"width":275,"height":195},{"panel_id":"c","x":0,"y":226,"width":190,"height":201},{"panel_id":"d","x":207,"y":229,"width":169,"height":199},{"panel_id":"e","x":394,"y":230,"width":173,"height":198}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/26/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/26/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/26/A Normally-Off GaN MIS-HEMT Fabricated.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":567,"height":428,"image_format":"jpeg","image_sha256":"4a913e299febca04f4c754faf33cdfb41f2eaeea896a77bbdc69472d2cebb17c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_29_fig_1.jpg","caption":"FIG. 1. EPC of ALE for the AlGaN film as a function of self-bias voltage for a fixed  $30\\mathrm{s}$  of Ar plasma exposure in the removal step: (i) incomplete removal, (ii) ALE window, and (iii) sputtering.","id":"train/atomic-layer-etching/experimental-usecase/29/fig_1","sample_id":"atomic-layer-etching/experimental-usecase/29/fig_1","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between self-bias voltage and etch depth per cycle. There is a linear increase in etch depth with increasing self-bias voltage after the ALE window.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Self-Bias Voltage (V) | Etch Depth per Cycle (nm) |\\n|---|---|\\n| 0 | 0 |\\n| 25 | 0 |\\n| 30 | 0.3 |\\n| 35 | 0.5 |\\n| 40 | 0.6 |\\n| 45 | 0.6 |\\n| 50 | 0.6 |\\n| 55 | 0.6 |\\n| 60 | 0.8 |\\n| 65 | 0.9 |\\n| 70 | 1.0 |\\n| 75 | 1.2 |\\n| 80 | 1.5 |\\n| 85 | 1.7 |\\n| 90 | 2.0 |\\n| 95 | 2.2 |\\n| 100 | 2.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"With the self-bias voltage the accelaration of ions is influenced, which are affected the most.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In that part of the bias range, sputtering is detected, this looks more like continuous reactive ion etching than like ALE.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the EPC, the ALE window should be looked at which is in panel 2. The EPC is then around 0.6 nm/cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The thresshold is 25 V.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":537,"height":462}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/Atomic layer etching of AlGaN using Cl2 and Ar gas.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":539,"height":467,"image_format":"jpeg","image_sha256":"3fe566e16ec5132e9c09ea93f7948143d4d9550faf0cc0f6e9524f6f66e0facd","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_33_fig_8.jpg","caption":"FIG. 8. Change in N/Ti ratio during an ALE cycle. The atomic concentrations of N and Ti were measured for four TiN films: (1) as-grown with native oxide; (2) before surface modification (after removal of native oxide); (3) after surface modification; (4) after surface modification and IR annealing.","id":"train/atomic-layer-etching/experimental-usecase/33/fig_8","sample_id":"atomic-layer-etching/experimental-usecase/33/fig_8","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the variation of N/Ti ratio across different conditions: 'With native oxide', 'Before mod.', 'After mod.', and 'IR Annealing'.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"|Process step| N/Ti Ratio | \\n|---|---|\\n| With native oxide| 0.95 | \\n| Before mod. |0.98 | \\n| After mod. |0.75 | \\n| IR Annealing | 1.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Oxygen atoms are also bonded to Ti atoms, resulting in less N-Ti bonds so lower N/Ti ratio.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"With the modification chlorine is added to the surface, reducing the number of N-Ti bonds at the surface for which the XPS is sensitive.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No the stoichiometry of the bulk does not change as the bulk ratio is 1 and after IR anenealing as wel.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In that case the surface will be oxidized again, meaning a native oxide layer will have formed so  a N/Ti ratio of 0.95 will be measured.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":664,"height":508}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/Atomic layer etching of titanium nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":669,"height":511,"image_format":"jpeg","image_sha256":"9c95a4e8eb718a376913e3f790a899dbcac9d685646f3be198aa1e19903f0498","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_38_fig_3.jpg","caption":"FIG. 3. Etch depth as a function of RF power for a fixed Reaction B step time of 5 s, 120 cycles.","id":"train/atomic-layer-etching/experimental-usecase/38/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/38/fig_3","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the relationship between Ar Plasma RF power and depth, indicating an increase in depth with increasing power. The depth plateaus in the region 15-30W, labeled as ALE Window.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ar Plasma RF power (W) | Depth (nm) |\\n|---|---|\\n| 0 | 0 |\\n| 5 | 2 |\\n| 10 | 5 |\\n| 15 | 13 |\\n| 20 | 12 |\\n| 25 | 14 |\\n| 30 | 14 |\\n| 35 | 28 |\\n| 40 | 38 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ALE window is the range of settings where the etch is self-limited and controllable per cycle. In this plot, it’s about 15–30 W Ar plasma RF power. It’s identified by the plateau where etch depth changes only slightly with increasing power.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3 cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approx. 0.117 nm per cycle (14 nm/120 cycles).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At RF powers below ~15 W and above ~30 W.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":4,"width":671,"height":408}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/High synergy atomic layer etching of AlGaNGaN.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":675,"height":414,"image_format":"jpeg","image_sha256":"f184de7756193e9e63fee05ee0a637981ab9daef29c9d5c07a6f8b7c08cc89e5","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_45_figure_7.jpg","caption":"Figure 7. Etch rates for TiN ALE versus temperature using  $\\mathrm{O}_3$  and HF as the reactants.","id":"train/atomic-layer-etching/experimental-usecase/45/figure_7","sample_id":"atomic-layer-etching/experimental-usecase/45/figure_7","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch rate increases with temperature up to 250°C and then plateaus\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature(°C) | Etch Rate (Å/cycle) |\\n|---|---|\\n| 150 | 0.06 |\\n| 200 | 0.14 |\\n| 250 | 0.21 |\\n| 300 | 0.20 |\\n| 350 | 0.21 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 250 C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The paper suggests that the oxidation step forms a similar thickness TiO2 layer over this temperature range, and the HF step removes only that oxidized layer. Because the amount removed per cycle is limited by how much oxide is formed in the oxidation step, the etch rate can stabilize even as temperature increases. This behavior is described as analogous to an ALD window.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, from about 250 C to 350 C.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They represent the uncertainty or variation in the measured etch rate at each temperature.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":668,"height":528}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/Thermal Atomic Layer Etching of Titanium Nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"45","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":672,"height":533,"image_format":"jpeg","image_sha256":"612e3ea4199485b2d5540a7fa710c1a315d87b70312cbd7cdf9e115bb70255f7","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_14.jpg","caption":"Figure 14. Saturation study of  $\\mathrm{WF_6 / BCl_3}$  ALE process at  $170^{\\circ}\\mathrm{C}$  using QCM: (a)  $1\\mathrm{WF_6}$  dose of  $0.2\\mathrm{s}$  per  $10\\mathrm{BCl_3}$  doses of  $0.2\\mathrm{s}$  each and (b)  $10\\mathrm{WF_6}$  doses of  $0.1\\mathrm{s}$  each per  $1\\mathrm{BCl_3}$  dose of  $1.0\\mathrm{s}$ .","id":"train/atomic-layer-etching/experimental-usecase/46/figure_14","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_14","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"QCM data demonstrating BCl₃ saturation behavior. After an initial WF₆ pulse causes fluorination and mass gain, the first few BCl₃ pulses induce a sharp mass loss. Subsequent BCl₃ pulses produce negligible change, confirming that the etching reaction is self-limiting once the fluorinated surface layer is consumed.\"},{\"panel_id\":\"b\",\"text\":\"QCM data demonstrating WF₆ saturation behavior. The first WF₆ pulse results in a large mass increase due to surface fluorination, while additional WF₆ pulses contribute minimal mass change, confirming that fluorination is also self-limiting and does not proceed indefinitely.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass loading (ng cm⁻²) | Process condition        |\\n|----------|------------------------|-------------------------|\\n| 0        | 0                      | Initial state           |\\n| 50       | 160                    | WF₆ pulse               |\\n| 100      | 60                     | 10× BCl₃                |\\n| 300      | 40                     | 10× BCl₃                |\\n| 500      | 30                     | 10× BCl₃         |\\n| 550      | 80                    | WF₆ pulse               |\\n| 600      | 20                     | 10× BCl₃                |\\n| 800      | 0                      | 10× BCl₃                |\\n| 1000     | −20                    |  10× BCl₃         |\\n| 1050     | 60                     | WF₆ pulse               |\\n| 1100     | −10                    | 10× BCl₃                |\\n| 1300     | −40                    | 10× BCl₃                |\\n| 1500     | −60                    | 10× BCl₃         |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Mass loading (ng cm⁻²) | Process condition  |\\n|----------|------------------------|------------------|\\n| 0        | 0                      | Initial state    |\\n| 100      | 150                    | 10× WF₆          |\\n| 300      | 220                    | 10× WF₆          |\\n| 500      | 260                    |  10× WF₆   |\\n| 550      | 50                     | BCl₃             |\\n| 700      | 80                     | 10× WF₆          |\\n| 900      | 70                     |  10× WF₆   |\\n| 1000     | −120                   | BCl₃             |\\n| 1150     | −80                    | 10× WF₆          |\\n| 1350     | −100                   |  10× WF₆   |\\n| 1500     | −280                   | BCl₃             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Diminishing Removal. No, the 10th pulse removes almost nothing. The first BCl₃ pulse causes a massive drop, the second is smaller, and by the 5th–10th pulse, the mass signal is essentially flat, indicating the reactant has run out of surface fluorides to react with.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The strategy in b is likely more efficient for maximizing etch rate. Multiple short WF₆ doses ensure complete fluorination of the surface without excessive precursor waste, while a single, sufficiently long BCl₃ dose guarantees complete removal of the modified layer. The method in 'a' may waste BCl₃ with many short, possibly incomplete doses.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Saturation proves both half-reactions are self-limiting. The WF₆ dose saturates once all available surface sites are fluorinated, forming a TiFₓ-WOₓFᵧ layer. The BCl₃ dose saturates once it completely converts/removes this modified layer to volatile products (e.g., TiCl₄). Further dosing beyond the saturation point yields no additional etch, confirming layer-by-layer control.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Clear saturation implies a robust process window. As long as dose times exceed the saturation point, minor fluctuations in precursor flow or timing will not affect the etch rate per cycle. This insensitivity is critical for high-volume manufacturing, ensuring reproducible etch depth and uniformity across the wafer and from run to run.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":475,"height":381},{"panel_id":"b","x":484,"y":5,"width":466,"height":376}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":950,"height":383,"image_format":"jpeg","image_sha256":"70bfe98a5138302a5e77ecd9c986e22679b63efce3eb3f7d4049bae6fff5d093","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_24_fig_4.jpg","caption":"Fig. 4. Influence of reactant exposure time and temperature on the sample mass changes during thermal ALE  $\\mathrm{Al}_2\\mathrm{O}_3$  films using TMA and HF. (I): effect of TMA (a) and HF (b) exposure time on mass changes ( $\\Delta \\mathrm{M}_{\\mathrm{TMA}}$  and  $\\Delta \\mathrm{M}_{\\mathrm{HF}}$ ) at  $300^{\\circ}\\mathrm{C}$ , respectively; (II): effect of temperature on  $\\Delta \\mathrm{M}_{\\mathrm{TMA}}$  and  $\\Delta \\mathrm{M}_{\\mathrm{HF}}$  of TMA, HF half-reaction (a) and mass change per ALE cycle (MCPC). Reproduced from Ref. [28].","id":"train/atomic-layer-etching/simulation-usecase/24/fig_4","sample_id":"atomic-layer-etching/simulation-usecase/24/fig_4","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"line chart"},{"panel_id":"d","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A line chart showing the change in ΔM<sub>TMA</sub> with increasing TMA exposure time at 300°C.\"},{\"panel_id\":\"b\",\"text\":\"A multiple line chart showing the change ΔMTMA, and ΔMHF as a function of the exposure time.\"},{\"panel_id\":\"c\",\"text\":\"A line chart showing the change in ΔM<sub>HF</sub> with increasing HF exposure time at 300°C.\"},{\"panel_id\":\"d\",\"text\":\"A line chart showing the change in MCPC as a function of the temperature, between 250 and 325  °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TMA Exposure Time (s) | ΔM<sub>TMA</sub> (ng/cm<sup>2</sup> cycle)) |\\n| --- | --- |\\n| 0.0 | 0 |\\n| 1.0 | -20 |\\n| 2.0 | -30 |\\n| 3.0 | - |\"},{\"panel_id\":\"b\",\"text\":\"| Exposure time | ΔM HF | ΔM TMA |\\n| --- | --- | --- |\\n| - | 5 | -10 |\\n| - | 10 | -20 |\\n| - | 18 | -30 |\\n|  | 20 | -40 |\"},{\"panel_id\":\"c\",\"text\":\"| HF Exposure Time (s) | ΔM<sub>HF</sub> (ng/cm<sup>2</sup> cycle)) |\\n| --- | --- |\\n| 0.0 | 0 |\\n| 0.5 | 12 |\\n| 1.0 | 14 |\\n| 1.5 | 15 |\\n| 2.0 | - |\"},{\"panel_id\":\"d\",\"text\":\"| Temperature (°C) | MCPC (ng/cm<sup>2</sup> cycle)) |\\n| --- | --- |\\n| 250 | 0 |\\n| 275 | -10 |\\n| 300 | -18 |\\n| 325 | -20 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The lower temperature, i.e. 250 °C\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the temperature increases, the amount of mass that changes per cycle decreases, slowing the ALE cycle down.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen in the subfigure, the mass change increases with the exposure time for HF, but decreases for TMA.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The mass changes of the samples, in this case for HF, are shown as a function of the HF exposure time (s).\"}]}]","bbox":[{"panel_id":"c","x":7,"y":285,"width":501,"height":292},{"panel_id":"b","x":532,"y":8,"width":469,"height":249},{"panel_id":"a","x":1,"y":5,"width":510,"height":274},{"panel_id":"d","x":536,"y":256,"width":467,"height":329}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/Review Paper -- Thermal atomic layer etching Mechanism, materials and prospects.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"24","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":1003,"height":625,"image_format":"jpeg","image_sha256":"fe375e9ee1fe47e45d4b42ba3fedcfd6c1755a377163739fe4d22aefbb130e97","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_c7be3f3d0a3f5b7cce1c5f6182cb538c5a79b547f531c40a02988e69d6a039b3.jpg","caption":"","id":"train/atomic-layer-etching/simulation-usecase/33/c7be3f3d0a3f5b7cce1c5f6182cb538c5a79b547f531c40a02988e69d6a039b3","sample_id":"atomic-layer-etching/simulation-usecase/33/c7be3f3d0a3f5b7cce1c5f6182cb538c5a79b547f531c40a02988e69d6a039b3","subset":"line-chart","split":"train","classification":[{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"b\",\"text\":\"The chart illustrates how the etching yield of silicon nitride (Si₃N₄) during HF exposure varies with the angle of incidence (θₐₙ) of incoming particles. The yield remains nearly constant around 0.06 between 0° and 30°, then slightly increases to a peak of 0.07 at 45°. Beyond this angle, the etching yield decreases markedly, reaching approximately 0.03 at 75°. This trend indicates that normal to moderately angled incidence promotes optimal etching, while highly oblique angles reduce etching efficiency, likely due to reduced energy transfer or surface access.\"}]","data_extraction":"[{\"panel_id\":\"b\",\"text\":\"| θ<sub>in</sub> (deg) | Etching yield (Si<sub>3</sub>N<sub>4</sub>/HF) |\\n|---|---|\\n| 0 | 0.06 |\\n| 15 | 0.06 |\\n| 30 | 0.06 |\\n| 45 | 0.07 |\\n| 60 | 0.05 |\\n| 75 | 0.03 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etching yield remains relatively stable between 0° and 30°, with values around 0.06. At 45°, the yield slightly increases to 0.07, suggesting enhanced surface interaction at this moderate angle. However, beyond 45°, the etching yield decreases significantly, dropping to 0.03 at 75°. This implies that more oblique angles reduce the effectiveness of etching, likely due to lower surface impact or inefficient energy transfer from incident particles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"45 degrees.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Maximum etching efficiency occurs near 45°, Minimal change is seen from 0° to 30° , Avoid high angles (≥ 60°), reactor design should consider angular effects.\"}]}]","bbox":[{"panel_id":"b","x":0,"y":0,"width":333,"height":286}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/c7be3f3d0a3f5b7cce1c5f6182cb538c5a79b547f531c40a02988e69d6a039b3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/c7be3f3d0a3f5b7cce1c5f6182cb538c5a79b547f531c40a02988e69d6a039b3.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":"b","first_classification_label":"line chart","caption_source":"not_found"},"width":333,"height":286,"image_format":"jpeg","image_sha256":"53bebf21667e446775f46c389de786e01c5462111aa757ae5632c766787960ce","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_figure_9.jpg","caption":"Figure 9. (a) The reflection coefficient, and (b) the total etching yield per incident HF molecule, both as a function of the incident angle. The error bars indicate the standard deviations among three independent runs.","id":"train/atomic-layer-etching/simulation-usecase/33/figure_9","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_9","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents the variation of reflection coefficient as a function of incident angle (θ<sub>in</sub>) from 0° to 75°. From 0° to 45°, the reflection coefficient remains relatively constant at around 0.1, indicating minimal reflection changes over low to moderate angles. At 60°, the coefficient begins to rise, and at 75°, it increases sharply to approximately 0.45. This indicates that higher incident angles lead to significantly more surface reflection, a behavior that may influence energy input or beam penetration in angle-sensitive surface processes.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| θ<sub>in</sub> (deg) | Reflection coefficient |\\n|---|---|\\n| 0 | 0.1 |\\n| 15 | 0.1 |\\n| 30 | 0.1 |\\n| 45 | 0.1 |\\n| 60 | 0.2 |\\n| 75 | 0.45 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The reflection coefficient remains relatively stable at around 0.1 between 0° and 45°, suggesting minimal angle dependence in that range. However, beyond 60°, the coefficient increases substantially, reaching approximately 0.45 at 75°. This trend indicates that reflection becomes significantly more pronounced at high angles of incidence, which could affect energy deposition or reaction efficiency in angle-sensitive processes.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 60 degrees\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Minimise angles above 60° , Maintain angles between 0° and 45° , Adjust beam alignment in plasma\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":323,"height":286}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_9.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":"line chart","caption_source":"content.json:image_caption"},"width":323,"height":286,"image_format":"jpeg","image_sha256":"411cc1668e980299573a3af45fa293f2c40a3055fbfdd594fa288d32a7cf45a4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_35_FIG5.jpg","caption":"Figure 5. (a-d) Snapshots representing the  $\\mathrm{H}_2\\mathrm{O}$  formation mechanism obtained in optimization of the  $\\mathrm{O}_5$  configuration, as shown in Figure 3 (the color code is the same). The left panel shows a complementary graph with relative energy along the formation pathway with respect to the initial geometry (desorbed oxygen atoms).","id":"train/atomic-layer-etching/simulation-usecase/35/FIG5","sample_id":"atomic-layer-etching/simulation-usecase/35/FIG5","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"molecular structure diagram"},{"panel_id":"c","label":"molecular structure diagram"},{"panel_id":"d","label":"molecular structure diagram"},{"panel_id":"e","label":"molecular structure diagram"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":1,"y":152,"width":615,"height":470},{"panel_id":"b","x":679,"y":8,"width":306,"height":204},{"panel_id":"c","x":680,"y":216,"width":312,"height":204},{"panel_id":"d","x":680,"y":428,"width":312,"height":201},{"panel_id":"e","x":680,"y":639,"width":285,"height":184}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/35/images/FIG5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/35/images/FIG5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/35/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/35/Density Functional Theory Study of Oxygen Adsorption on Polymer Surfaces for Atomic-Layer.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"35","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:image_caption"},"width":998,"height":828,"image_format":"jpeg","image_sha256":"de5811f88545d56a9c69af1f64770cbae23484b972b1b4d2282531eed633a3b5","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_4_figure_1.jpg","caption":"Figure 1. Etch per cycle for TiN films versus  $\\mathrm{NbF}_5$  pulse time at  $460^{\\circ}C$  The duration of  $\\mathbb{N}_2$  purges and  $\\mathrm{CCl}_4$  pulses were fixed at 6 and 1 s, respectively.","id":"train/atomic-layer-etching/simulation-usecase/4/figure_1","sample_id":"atomic-layer-etching/simulation-usecase/4/figure_1","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure illustrates how the etch per cycle of TiN films varies with NbF₅ pulse time during the NbF₅–CCl₄ etch process at 460 °C. The etch per cycle increases from about 0.3 Å without NbF₅ to approximately 0.8 Å at a 3 s NbF₅ pulse, showing a gradual increase without clear saturation under the applied conditions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| NbF₅ pulse time [s] | Etch per cycle [Å] |\\n|---|---|\\n| 0.0 | 0.3 |\\n| 1.0 | 0.6 |\\n| 3.0 | 0.8 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch per cycle increases with increasing NbF₅ pulse time. When no NbF₅ is used, the etch per cycle is about 0.3 Å, indicating that CCl₄ alone can etch TiN. As the NbF₅ pulse time increases to 1 s, the etch per cycle rises to approximately 0.6 Å, and further increases to about 0.8 Å at a 3 s pulse time. The trend shows a gradual increase without clear saturation.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Partially self-limiting surface reactions, Insufficient purge time, Low volatility or slow removal of reaction products, Presence of residual CCl₄ vapors in the etch system, Slow diffusion of fluorine species through TiN grain boundaries\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.6 Å.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":1,"width":623,"height":569}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/Combining Experimental and DFT Investigation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":633,"height":575,"image_format":"jpeg","image_sha256":"9ddaf759394920b5a45161c754403177fba77daf46ec5daa2f58239c413b30e7","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_9_1efb120580ad0809db6986794a53a1e9c383b07153d625e10037bb0738d83671.jpg","caption":"Fig. A1. Convergce tests for atomic layr nmber and cutoff energies for DFT calculations. (a) Variation of adsorption energy of chlorine with layer number, cutoff energies of  $50 / 200\\mathrm{Ry}$  were applied. (b) Variation of total system energy with cutoff energies for wave function and electron density, the cutoff energy for electron density was set to be 4 times that for wave function, as a default setting in Quantum Espresso.","id":"train/atomic-layer-etching/simulation-usecase/9/1efb120580ad0809db6986794a53a1e9c383b07153d625e10037bb0738d83671","sample_id":"atomic-layer-etching/simulation-usecase/9/1efb120580ad0809db6986794a53a1e9c383b07153d625e10037bb0738d83671","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between layer number and adsorption energy, with a decreasing trend, indicating that for a certain structure adsorption is more likely.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates the decrease in total energy with increasing cut-off energies.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Layer number | Adsorption Energy (eV) |\\n|---|---|\\n| 4 | 4.2 |\\n| 6 | 4.13 |\\n| 8 | 4.12 |\\n| 10 | 4.12 |\\n| 12 | 4.12 |\"},{\"panel_id\":\"b\",\"text\":\"| Cut-off energies (Ry) | Total energy (Ry) |\\n|---|---|\\n| 15 | -945.9 |\\n| 20 | -946.1 |\\n| 25 | -946.2 |\\n| 30 | -946.3 |\\n| 35 | -946.3 |\\n| 40 | -946.3 |\\n| 45 | -946.3 |\\n| 50 | -946.3 |\\n| 55 | -946.3 |\\n| 60 | -946.35 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Energy in Ry (Rydberg) is a fundamental unit in atomic physics, representing the ionization energy of a hydrogen atom from its ground state, approximately 13.6 electron volts (eV).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"50 Ry\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It could be that from a certain thickness onwards, the substrate has no repelling influence on adsorption anymore.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Layer number 8.\"}]}]","bbox":[{"panel_id":"b","x":7,"y":455,"width":654,"height":391},{"panel_id":"a","x":9,"y":11,"width":654,"height":400}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/images/1efb120580ad0809db6986794a53a1e9c383b07153d625e10037bb0738d83671.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/images/1efb120580ad0809db6986794a53a1e9c383b07153d625e10037bb0738d83671.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/Mechanism of photoassisted atomic layer etching of chlorinated Si(111) surfaces Insight.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"9","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":667,"height":850,"image_format":"jpeg","image_sha256":"114e6d0e178f8c65967445067b4339a644ab0c8b3d44a4e188b6f0b4da1539ec","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_9_42b2341dcb88232a2b3d8fc38fe7c5cbc5345d5e59715edb80a39db016f6f34c.jpg","caption":"Fig. A2. Variation of the lowest excitation energy with the number of Si atoms of the cluster, insets are the molecular structures of the associated clusters.","id":"train/atomic-layer-etching/simulation-usecase/9/42b2341dcb88232a2b3d8fc38fe7c5cbc5345d5e59715edb80a39db016f6f34c","sample_id":"atomic-layer-etching/simulation-usecase/9/42b2341dcb88232a2b3d8fc38fe7c5cbc5345d5e59715edb80a39db016f6f34c","subset":"line-chart","split":"train","classification":[{"panel_id":"a","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The excitation energy of a system decreases as the number of silicon (Si) atoms increases, this is a stepwise relation as a function of the number of Si atoms.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Si atoms | Excitation energy (eV) |\\n|---|---|\\n| 0 | - |\\n| 3 | 4.1 |\\n| 7 | 3.7 |\\n| 11 | 3.6 |\\n| 15 | 3.2 |\\n| 24 | 3.2 |\\n| 34 | 3.1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"n=34 as actual substrates used in experimental data are composed of way more atoms than depicted in this graph.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Big substrates in simulations require way more computing power than substrates with little number of atoms.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A reason could be the effect that more connected silicon atoms shift energy levels in adsorbed species to lower energies.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No this is unlikely, except for the case where the delta time is reduced to an extremely low value\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":662,"height":433}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/images/42b2341dcb88232a2b3d8fc38fe7c5cbc5345d5e59715edb80a39db016f6f34c.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/images/42b2341dcb88232a2b3d8fc38fe7c5cbc5345d5e59715edb80a39db016f6f34c.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/Mechanism of photoassisted atomic layer etching of chlorinated Si(111) surfaces Insight.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"9","first_classification_panel_id":"a","first_classification_label":"line chart","caption_source":"content.json:img_caption"},"width":667,"height":436,"image_format":"jpeg","image_sha256":"0a3fe8f79a8e85544bb89ae30de61dd6b8698e506f7cbfb3b842802f19958580","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}