subsets/scatter-plot/test/metadata.jsonl

{"file_name":"images/test_atomic-layer-deposition_experimental-usecase_10_fig_15.jpg","caption":"FIG. 15. Surface coverage of Ti atoms as a function of the number of ALD cycles, as measured by a quartz crystal microbalance at room temperature (Ref. 232). The precursor employed was  $\\mathrm{TiCl_4}$  and the reactant was an  $\\mathrm{H}_{2}$  plasma. For deposition on the as-received crystal, the growth showed a linear trend after  $\\sim 5$  cycles. Reprinted with permission from H. Kim et al., J. Vac. Sci. Technol. A 20, 802 (2002). Copyright 2002 American Vacuum Society.","id":"test/atomic-layer-deposition/experimental-usecase/10/fig_15","sample_id":"atomic-layer-deposition/experimental-usecase/10/fig_15","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows a linear relationship between Ti atom surface coverage and the number of ALD cycles. The direct proportionality indicates a consistent growth rate per cycle, characteristic of self-limiting reactions in atomic layer deposition (ALD).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | Ti Atom Surface Coverage (×10¹⁵ /cm²) |\\n|------------------|---------------------------------------|\\n| 0                | 0                                     |\\n| 10               | 5                                     |\\n| 20               | 10                                    |\\n| 30               | 15                                    |\\n| 40               | 20                                    |\\n| 50               | 25                                    |\\n| 60               | 30                                    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ti atom surface coverage increases linearly with the number of ALD cycles, indicating a constant growth per cycle characteristic of ideal ALD behavior.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ti atom surface coverage (×10¹⁵/cm²).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Consistent growth rate per cycle, Self-limiting surface reactions, Stable precursor adsorption behavior\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":600,"height":511}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_15.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_15.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/H. B. Profijt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":600,"height":511,"image_format":"jpeg","image_sha256":"9eeda7167bca5985b4e09df0e493cc47e696b113600a26d3c61fd50ce6818070","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/test_atomic-layer-deposition_experimental-usecase_10_fig_19.jpg","caption":"FIG. 19. (Color online) Water vapor transmission rate (WVTR) of a  $20 \\mathrm{nm}$  thick  $\\mathrm{Al}_2\\mathrm{O}_3$  films on poly(2,6-ethylenenaphthalate) (PEN) substrates as a function of the deposition temperature (Ref. 54). The  $\\mathrm{Al}_2\\mathrm{O}_3$  films were deposited by plasma-assisted ALD using  $\\mathrm{AlMe}_3$  as the precursor and an  $\\mathrm{O}_2$  plasma as the reactant. The inset shows the WVTR as a function of the film thickness for a film deposited at room temperature. The WVTR values were determined using a standard calcium test (Ref. 378) and include water permeation through pinholes which are possibly present. From E. Langereis et al., Appl. Phys. Lett. 89, 081915 (2006). Reprinted with permission. Copyright 2006, American Institute of Physics.","id":"test/atomic-layer-deposition/experimental-usecase/10/fig_19","sample_id":"atomic-layer-deposition/experimental-usecase/10/fig_19","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure compares the water vapor transmission rate (WVTR) as a function of deposition temperature for Al₂O₃ films, with an inset showing WVTR versus film thickness on PEN substrates. Increasing deposition temperature initially reduces WVTR due to improved film density but rises at higher temperatures. The inset reveals that increasing Al₂O₃ thickness significantly decreases WVTR, indicating enhanced barrier performance.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition temperature (°C) | WVTR (g m⁻² day⁻¹) |\\n|---|---|\\n| 20 | 0.008 |\\n| 30 | 0.009 |\\n| 40 | 0.01 |\\n| 50 | 0.011 |\\n| 60 | 0.012 |\\n| 70 | 0.013 |\\n| 80 | 0.014 |\\n| 90 | 0.015 |\\n| 100 | 0.016 |\"},{\"panel_id\":\"b\",\"text\":\"| Al₂O₃ thickness (nm) | WVTR (g m⁻² day⁻¹) |\\n|---|---|\\n| 0 | 0.4 |\\n| 10 | 0.01 |\\n| 20 | 0.02 |\\n| 30 | 0.03 |\\n| 40 | 0.04 |\\n| 50 | 0.05 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"WVTR (Water Vapor Transmission Rate) in g·m⁻²·day⁻¹\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Thicker Al₂O₃ improves barrier, WVTR drops to ~40 nm, Diminishing returns beyond\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"WVTR decreases with increasing film thickness due to improved barrier density, but shows a minimum at moderate deposition temperatures, rising again at higher temperatures due to potential defect formation.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":603,"height":451},{"panel_id":"b","x":122,"y":34,"width":302,"height":221}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_19.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/images/fig_19.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/10/H. B. Profijt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":608,"height":456,"image_format":"jpeg","image_sha256":"530d21b8017f2057f5a4ea6d658d9c782869704da620531dc7929468e831e84b","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/test_atomic-layer-deposition_experimental-usecase_21_fig_2.jpg","caption":"Fig. 2 Saturation curves for (a) thermal and (b) plasma-assisted ALD of  $\\mathrm{Li_2CO_3}$  obtained at  $150^{\\circ}\\mathrm{C}$  showing the self-limiting growth behavior of the ALD process. The solid line serves as a guide to the eye. Dotted lines indicate the chosen standard values and the corresponding growth per cycle. Note that the reaction time of  $\\mathrm{H}_2\\mathrm{O}$  after dosing was kept at  $4\\mathrm{s}$ .","id":"test/atomic-layer-deposition/experimental-usecase/21/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/21/fig_2","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Saturation curves for thermal ALD of Li2CO3. The line chart shows the growth per cycle (Å) as a function of LiO'Bu dose (s), H2O dose and CO2 dose.\"},{\"panel_id\":\"b\",\"text\":\"Saturation curves for  plasma-assisted ALD of Li2CO3. The line chart shows the growth per cycle (Å) as a function of LiO'Bu dose and plasma duration.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Parameter        | Growth per Cycle (Å) |\\n|-------------------|-----------------------|\\n| LiO'Bu dose (s)   | 0.6                  |\\n| H₂O dose (ms)     | 0.6                  |\\n| CO₂ dose (s)      | 0.6                  |\"},{\"panel_id\":\"b\",\"text\":\"| Parameter          | Growth per Cycle (Å) |\\n|---------------------|-----------------------|\\n| LiO'Bu dose (s)     | 0.8                  |\\n| Plasma duration (s) | 0.8                  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A LiOtBu dose of 6 s, a water dose of 50 ms and a CO2 dose of 2 s were chosen as standard conditions for thermal ALD of Li2CO3.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Without H2O dosing no film growth occurred and without CO2, LiOH was formed. Due to a slower response speed of the CO2 valve, dosing of 0.5 s or less did not deliver any CO2 in the chamber (no pressure increase was observed). Purge steps after LiOtBu and CO2 dose of 2 s are sufficient to remove the reaction products and unreacted species from the processing chamber, although 5 s purge was necessary for H2O. A growth per cycle of  0.60 °A was found for the thermal ALD process.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The saturation behavior of the surface reactions during the  two half cycles of the plasma-assisted ALD process, can be seen in b. For both half-cycles purge steps of 2 s are sufficient to remove the reaction products and unreacted species. A growth per cycle of 0.82 °A was obtained for a deposition temperature of 150 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":745,"height":403},{"panel_id":"b","x":752,"y":5,"width":541,"height":403}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/21/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/21/images/fig_2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/21/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/21/N. Hornsveld et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"21","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:image_caption"},"width":1292,"height":411,"image_format":"jpeg","image_sha256":"fdc41a460aaed2018fd2f52daa3e9d459869f1c065260ad4973d878829020736","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/test_atomic-layer-deposition_experimental-usecase_48_fig_1.jpg","caption":"FIG. 1. (Color online) GPC values as a function of (a) HyALD dose; (b) Ar purge time after precursor dose; (c)  $\\mathrm{O}_2$  plasma exposure; and (d) purge time after the plasma step. The substrate temperature during the process was  $250^{\\circ}\\mathrm{C}$ . The dashed lines serve as a guide to the eyes.","id":"test/atomic-layer-deposition/experimental-usecase/48/fig_1","sample_id":"atomic-layer-deposition/experimental-usecase/48/fig_1","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"scatter plot"},{"panel_id":"d","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between HfCp(NMe₂)₃ dose and GPC (nanometers). GPC increases sharply at short doses  and then levels off near ~0.11 nm/cycle.\"},{\"panel_id\":\"b\",\"text\":\"The scatter plot shows the relationship between HfCp(NMe₂)₃ purge time and GPC (nanometers). The data points are relatively constant around 0.11 nm with no significant changes observed.\"},{\"panel_id\":\"c\",\"text\":\"The scatter plot shows the relationship between O₂ plasma exposure time and GPC (nanometers). The data points increase slightly with increasing exposure time up to 0.11 nm.\"},{\"panel_id\":\"d\",\"text\":\"The scatter plot shows the relationship between O₂ plasma purge time and GPC (nanometers). The data points decrease slightly with increasing exposure time.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| HfCp(NMe₂)₃ dose (s) | GPC (nm) |\\n|---|---|\\n| 0.5 | 0.04 ± 0.008 |\\n| 1 | 0.09 ± 0.008 |\\n| 3 | 0.10 ± 0.008 |\\n| 5 | 0.11 ± 0.008 |\\n| 8 | 0.11 ± 0.008 |\"},{\"panel_id\":\"b\",\"text\":\"| HfCp(NMe₂)₃ purge (s) | GPC (nm) |\\n|---|---|\\n| 0 | 0.11 ± 0.008|\\n| 1 | 0.107 ± 0.008 |\\n| 3 | 0.106 ± 0.008 |\\n| 5 | 0.106 ± 0.008 |\\n| 10 | 0.108 ± 0.008 |\"},{\"panel_id\":\"c\",\"text\":\"| O₂ plasma exposure time (s) | GPC (nm) |\\n|---|---|\\n| 1 | 0.07 ± 0.008 |\\n| 3 | 0.09 ± 0.008 |\\n| 5 | 0.10 ± 0.008 |\\n| 8 | 0.11 ±0.008 |\"},{\"panel_id\":\"d\",\"text\":\"| O₂ plasma purge (s) | GPC (nm) |\\n|---|---|\\n| 0 | 0.12 ± 0.008 |\\n| 1 | 0.118 ± 0.008 |\\n| 2 | 0.111 ± 0.008 |\\n| 3 | 0.110 ± 0.008 |\\n| 4 | 0.108 ± 0.008 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After approx. 3 s  of HfCp(NMe₂)₃ dose (~0.11 nm/cycle).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, short purging after the HfCp(NMe₂)₃ step appears to adequately prevent additional (non-self-limiting) growth. The GPC at very short purge times is essentially the same as at longer purge times and remains near ≈0.11 nm/cycle within the error bars. Because increasing the purge duration does not lower the GPC relative to the short-/no-purge case, there is no evidence of extra growth from residual precursor.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Very short purges can leave behind reactive byproducts, that inflate GPC.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After approx. 5 s. of plasma exposure (beyond this point, the change in growth  is within the error bars).\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":501,"height":362},{"panel_id":"b","x":509,"y":5,"width":486,"height":362},{"panel_id":"c","x":1,"y":368,"width":501,"height":345},{"panel_id":"d","x":505,"y":366,"width":487,"height":345}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/48/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/48/images/fig_1.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/48/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/48/Sharma et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"48","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:image_caption"},"width":997,"height":714,"image_format":"jpeg","image_sha256":"86058ca64999979f7b706706b06ff40ef18284e0119545cbef9c8093a10c294b","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/test_atomic-layer-deposition_experimental-usecase_6_fig2.jpg","caption":"FIG.2.  $\\mathrm{SnO_x}$  growth rate determined by VASE, as a function of (a) and (c) TDMASn pulse length  $\\mathfrak{t}_1$  for timing sequences of  $\\mathrm{t}_1{:}30{:}2{:}30\\mathrm{s}$  and (b) and (d)  $\\mathrm{H}_2\\mathrm{O}$  pulse length  $\\mathfrak{t}_3$  for timing sequences of  $1:30:\\mathrm{t}_3:30\\mathrm{s}$  The substrate was Si(100) at  $150^{\\circ}\\mathrm{C}$  for (a) and (b) and at  $30^{\\circ}\\mathrm{C}$  for (c) and (d).","id":"test/atomic-layer-deposition/experimental-usecase/6/fig2","sample_id":"atomic-layer-deposition/experimental-usecase/6/fig2","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"scatter plot"},{"panel_id":"d","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents growth rate dependence on pulse length for TDMASn and H₂O across four panels (a–d), highlighting two distinct growth regimes. Panels (a) and (b) show saturation behavior around ~0.7 Å/cycle, with growth plateauing beyond 1–2 seconds of precursor exposure, indicative of self-limiting ALD reactions. In contrast, panels (c) and (d) exhibit a much higher growth rate of ~2.0 Å/cycle under shorter pulse conditions, suggesting a different growth mechanism or system configuration. These observations confirm typical ALD characteristics and may imply process differences such as thermal vs. plasma ALD, or substrate surface reactivity changes.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TDMASn Pulse Length (s) | Growth Rate (A/cycle) |\\n|---|---|\\n| 0.0 | 0.0 |\\n| 1.0 | 0.7 |\\n| 2.0 | 0.7 |\\n| 4.0 | 0.7 |\\n| 6.0 | 0.7 |\"},{\"panel_id\":\"b\",\"text\":\"| H₂O Pulse Length (s) | Growth Rate (A/cycle) |\\n|---|---|\\n| 0.0 | 0.15 |\\n| 1.0 | 0.65 |\\n| 2.0 | 0.75 |\\n| 4.0 | 0.7 |\\n| 6.0 | 0.7 |\"},{\"panel_id\":\"c\",\"text\":\"| TDMASn Pulse Length (s) | Growth Rate (A/cycle) |\\n|---|---|\\n| 0.0 | 0.0 |\\n| 1.0 | 2.0 |\\n| 2.0 | 2.0 |\\n| 4.0 | 2.0 |\"},{\"panel_id\":\"d\",\"text\":\"| H₂O Pulse Length (s) | Growth Rate (A/cycle) |\\n|---|---|\\n| 0.0 | 0.0 |\\n| 1.0 | 2.0 |\\n| 2.0 | 2.0 |\\n| 4.0 | 2.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"(a), (b), saturation at ~0.7 Å/cycle, (c), (d), saturation at ~2.0 Å/cycle, (b), (d), initial growth increases with H₂O pulse time.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.0 seconds.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The consistent saturation behavior observed in each panel confirms the self-limiting nature of ALD surface reactions. In panels (a) and (b), the growth rate increases rapidly with precursor pulse time but plateaus at around 0.7 Å/cycle, indicating sufficient surface exposure. Panels (c) and (d) follow a similar saturation trend but reach a much higher steady-state growth rate of 2.0 Å/cycle, suggesting a different process condition or chemistry, such as enhanced reactivity or modified surface structure.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":8,"width":599,"height":285},{"panel_id":"b","x":7,"y":313,"width":597,"height":288},{"panel_id":"c","x":0,"y":614,"width":605,"height":288},{"panel_id":"d","x":4,"y":916,"width":600,"height":291}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/Mullings et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":605,"height":1208,"image_format":"jpeg","image_sha256":"96299feee181c204672d670177f523f7d4dd9342f4fde29d884d56cf20daa549","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/test_atomic-layer-deposition_experimental-usecase_6_fig_3.jpg","caption":"FIG. 3.  $\\mathrm{SnO_x}$  film thickness as a function of number of ALD cycles yielding a growth rate of  $0.70\\mathrm{\\AA}$  cycle. Thicknesses were obtained using VASE for films deposited on Si(100) at  $150^{\\circ}\\mathrm{C}$  with a time sequence of  $130:2:30$","id":"test/atomic-layer-deposition/experimental-usecase/6/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/6/fig_3","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart presents a linear relationship between the number of SnO₂ ALD cycles and film thickness, highlighting consistent growth per cycle. The trend demonstrates self-limiting behavior typical of ALD, where each cycle deposits a nearly constant thickness of material. At 700 cycles, the film reaches approximately 480 Å, implying a growth rate of roughly 0.69 Å/cycle. This confirms a stable ALD process, with high conformity and control over thickness scaling.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of SnO₂ ALD cycles | Film Thickness (Å) |\\n|---|---|\\n| 0 | 0 |\\n| 100 | 50 |\\n| 200 | 100 |\\n| 500 | 350 |\\n| 700 | 480 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear relationship between ALD cycles and film thickness confirms that the SnO₂ growth is self-limiting and well-controlled. This behavior is characteristic of atomic layer deposition, where each cycle deposits a nearly uniform thickness. Such repeatability across hundreds of cycles highlights the process stability and makes it ideal for precise thickness control in thin-film applications.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"100 cycles ≈ 75 Å, 200 cycles ≈ 150 Å, 500 cycles ≈ 375 Å, 700 cycles ≈ 480 Å.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.69 Å/cycle.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":609,"height":300}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/Mullings et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":609,"height":300,"image_format":"jpeg","image_sha256":"2d001e88209262fd07b365fec00cc4faca3b6c7431ebfcb2b8fb197cb473134d","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/test_atomic-layer-deposition_experimental-usecase_6_fig_4.jpg","caption":"FIG. 4.  $\\mathrm{SnO_x}$  ALD growth rate as a function of temperature determined using VASE for films deposited on Si(100) for 200 cycles with timing sequence  $1:30:2:30\\mathrm{s}$","id":"test/atomic-layer-deposition/experimental-usecase/6/fig_4","sample_id":"atomic-layer-deposition/experimental-usecase/6/fig_4","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot illustrates the variation in growth rate as a function of deposition temperature during a material deposition process, likely ALD. The plot reveals a clear negative correlation: as the deposition temperature increases from 25 °C to 200 °C, the growth rate steadily decreases from 2.0 Å/cycle to 0.4 Å/cycle. This trend may indicate a transition from a kinetically favorable low-temperature regime to a surface-limited regime at higher temperatures, where desorption or precursor decomposition could reduce deposition efficiency\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition Temperature (°C) | Growth Rate (Å/cycle) |\\n|---|---|\\n| 25 | 2.0 |\\n| 50 | 1.8 |\\n| 75 | 1.5 |\\n| 100 | 1.2 |\\n| 125 | 1.0 |\\n| 150 | 0.8 |\\n| 175 | 0.6 |\\n| 200 | 0.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The growth rate decreases consistently as the deposition temperature increases. At lower temperatures (25 °C–75 °C), the growth rate is relatively high, indicating effective precursor adsorption and surface reaction. As the temperature increases beyond 100 °C, the growth rate gradually drops, potentially due to decreased surface sticking probability or thermal decomposition of precursors, which reduces surface saturation efficiency. This suggests that optimal ALD conditions for maximum growth occur at lower temperatures in this process.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"25 °C, 50 °C, 75 °C\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.8 Å/cycle\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":609,"height":316}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/Mullings et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":609,"height":316,"image_format":"jpeg","image_sha256":"b200b53139f6df66beeef951457b1c86ca48fc85609a9da3b192107fcab39cfb","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/test_atomic-layer-deposition_experimental-usecase_6_fig_5.jpg","caption":"FIG. 5.  $\\mathrm{SnO_x}$  growth rate at  $30^{\\circ}\\mathrm{C}$  determined by VASE as a function of nitrogen purge time following (a) TDMASn pulse length and (b)  $\\mathrm{H}_2\\mathrm{O}$  pulse length for timing sequences of  $1:1:2:1\\mathrm{t}_4\\mathrm{s}$ . The substrate was Si(100).","id":"test/atomic-layer-deposition/experimental-usecase/6/fig_5","sample_id":"atomic-layer-deposition/experimental-usecase/6/fig_5","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The two scatter plots examine how growth rate varies with increasing N₂ purge time after the TDMASn pulse during an ALD process. In both cases, a sharp decrease in growth rate is observed between 0 s and 60 s of purge time, indicating rapid removal or deactivation of reactive surface species. In plot (a), the growth rate partially recovers at intermediate purge times, while in (b), it continues to decrease monotonically. This suggests different mechanisms or sensitivities to precursor byproducts or surface reactivity in the two experimental conditions, possibly related to substrate type, reactant sequence, or deposition temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| N₂ Purge after TDMASn Pulse (s) | Growth Rate (Å/cycle) |\\n|---|---|\\n| 0 | 2.5 |\\n| 60 | 2.0 |\\n| 180 | 2.7 |\\n| 360 | 2.2 |\"},{\"panel_id\":\"b\",\"text\":\"| N₂ Purge after TDMASn Pulse (s) | Growth Rate (Å/cycle) |\\n|---|---|\\n| 0 | 3.5 |\\n| 60 | 2.0 |\\n| 180 | 1.8 |\\n| 360 | 1.7 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"0 s ≈ 3.5 Å/cycle, 60 s ≈ 2.0 Å/cycle, 180 s ≈ 1.8 Å/cycle, 360 s ≈ 1.7 Å/cycle.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure shows that increasing the N₂ purge time after the TDMASn pulse significantly affects the growth rate. In both experimental conditions, the growth rate sharply decreases between 0 and 60 seconds of purge time, suggesting a rapid loss of reactive surface species or byproduct accumulation effects. In plot (a), the growth partially recovers at 180 seconds, while in plot (b), it declines more steadily, indicating a stronger sensitivity to purge time. This highlights the importance of purge duration in maintaining process efficiency and controlling film uniformity.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.7 Å/cycle.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":6,"width":601,"height":297},{"panel_id":"b","x":1,"y":311,"width":603,"height":297}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/images/fig_5.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/6/Mullings et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":608,"height":606,"image_format":"jpeg","image_sha256":"2ca2983ab7609432e47df2a26cdb13f195f8619500e1c8532226c183134d4324","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/test_atomic-layer-deposition_experimental-usecase_62_fig_6.jpg","caption":"FIG. 6. (Color online) (a) Time-integrated emission intensity at  $388~\\mathrm{nm}$  related to CN emission, as a function of the  $\\mathrm{Ta}[\\mathrm{N}(\\mathrm{CH}_3)_2]_5$  exposure time during plasma-assisted ALD of TaN. The emission intensity for zero  $\\mathrm{Ta}[\\mathrm{N}(\\mathrm{CH}_3)_2]_5$  exposure time is due to emission from the  $\\mathrm{H}_2$  plasma. (b) Time-integrated emission intensities of O plasma species and  $\\mathrm{H}$  and CO reaction products as a function of the number of consecutive  $\\mathrm{Al}(\\mathrm{CH}_3)_3$  doses in the  $\\mathrm{Al}_2\\mathrm{O}_3$  plasma-assisted ALD cycle. The lines serve as guides to the eye.","id":"test/atomic-layer-deposition/experimental-usecase/62/fig_6","sample_id":"atomic-layer-deposition/experimental-usecase/62/fig_6","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the integrated intensity of a CN peak after a certain Ta[N(CH₃)₂₅] exposure time.\"},{\"panel_id\":\"b\",\"text\":\"The chart displays the integrated intensity of different plasma species against the number of Al(CH₃) doses.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ta[N(CH₃)₂₅] exposure time (s) | Integrated intensity (a.u.) |\\n|---|---|\\n|0| 750 | \\n|1| 840 | \\n|2| 870 | \\n|3| 880 | \\n|4| 880 | \\n|5| 880 |\\n|6| 880 | \\n|7| 880 |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Al(CH₃) doses| Integrated intensity (a.u.) O | Integrated intensity (a.u.) H | Integrated intensity (a.u.) CO | \\n|---|---|---|---|\\n| 0 |60|0|0|\\n| 1 |52|23|38|\\n| 2 |50|42|60|\\n| 3 |42|50|68|\\n| 4 |40|58|72|\\n| 5 |40|62|80|\\n| 6 |40|62|80|\\n| 8 |40|62|80|\\n| 12 |40|62|80|\\n| 16 |40|62|80|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The required exposure time is 4 seconds.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the pure hydrogen plasma, there are also species with intensity at this wavelength. This is used as the reference for the intensity after precursor dose.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It takes 6 doses.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the data, peaks related to oxygen species are visible. This is a clear indication that the plasma is an oxygen plasma.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":617,"height":539},{"panel_id":"b","x":3,"y":540,"width":625,"height":518}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/62/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/62/images/fig_6.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/62/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/experimental-usecase/62/Mackus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"62","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:image_caption"},"width":625,"height":1066,"image_format":"jpeg","image_sha256":"f405c52eafd7accdee1e22abcd058d6134dab69ff4d8b70550b5680dc2ba1d84","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/test_atomic-layer-deposition_simulation-usecase_18_fig_1.jpg","caption":"Fig. 1. Comparison of calculated rate constant with experiment [14] for the reaction  $\\mathrm{TMG} \\rightarrow \\mathrm{DMG} + \\mathrm{CH}_3$ .","id":"test/atomic-layer-deposition/simulation-usecase/18/fig_1","sample_id":"atomic-layer-deposition/simulation-usecase/18/fig_1","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot compares calculated reaction rate constants for Trimethylgallium (TMG) decomposition with experimental data from a previous work. The Arrhenius plot of log(k) versus reciprocal temperature (10000/T) shows strong agreement between the computed values (triangles) and experimental points. This consistency confirms that the ab initio–derived activation barriers accurately represent the gas-phase decomposition kinetics.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| 10000/T (1/K) | log k (1/sec) | Work        |\\n|--------------:|--------------:|------------|\\n| 7.0           | 6.4           | This work  |\\n| 7.3           | 6.1           | This work  |\\n| 7.6           | 5.8           | This work  |\\n| 8.0           | 5.5           | This work  |\\n| 8.3           | 5.2           | This work  |\\n| 8.7           | 4.9           | This work  |\\n| 9.0           | 4.6           | This work  |\\n| 9.3           | 4.3           | This work  |\\n| 9.6           | 4.0           | This work  |\\n| 9.9           | 3.7           | This work  |\\n| 10.2          | 3.4           | Experiment |\\n| 10.5          | 3.1           | Experiment |\\n| 10.8          | 2.8           | Experiment |\\n| 11.1          | 2.5           | Experiment |\\n| 11.4          | 2.2           | Experiment |\\n| 11.7          | 1.9           | Experiment |\\n| 12.0          | 1.6           | Experiment |\\n| 12.3          | 1.3           | Experiment |\\n| 12.6          | 1.0           | Experiment |\\n| 12.9          | 0.7           | Experiment |\\n| 13.2          | 0.4           | Experiment |\\n| 13.5          | 0.1           | Experiment |\\n| 13.8          | -0.2          | Experiment |\\n| 14.1          | -0.5          | Experiment |\\n| 14.4          | -0.8          | Experiment |\\n| 14.7          | -1.1          | This work  |\\n| 15.0          | -1.4          | This work  |\\n| 15.6          | -2.5          | This work  |\\n| 16.5          | -3.8          | This work  |\\n| 17.5          | -5.5          | This work  |\\n| 19.5          | -9.5          | This work  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"I  - Region 7 (10000/T): High temperature, corresponding to a very fast reaction rate (log k ≈ 5).\\nII - Region 14 (10000/T): Moderate temperature, where the reaction rate decreases significantly (log k ≈ −2).\\nIII - Region 20 (10000/T): Low temperature, at which the reaction is effectively negligible (log k ≈ −12).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Experimental data is absent at low temperatures (10000/T > 15) because the reaction becomes too slow to measure accurately in a lab setting. However, the computational route (triangles) extends well into this region (log k-12). This makes the theoretical model superior for simulating cool-down zones or exhaust lines, ensuring that even trace decomposition at low temperatures is accounted for in safety or clogging models.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear agreement confirms that the ab initio-derived transition state geometry and its associated energy barrier accurately represent the real kinetic pathway. The fundamental connection is Transition State Theory (TST), which states that the rate constant is directly governed by the Gibbs free energy of the activated complex. The close match validates that the computationally modeled reaction coordinate and barrier height (property) correctly map to the true chemical structure of the rate-determining transition state.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies that ab initio methods are a reliable tool for predictive process design. Engineers can use them to accurately model how temperature affects TMG decomposition rates in the gas phase, which is critical for controlling precursor delivery and reactivity in the MOCVD chamber. This allows for the in silico optimization of temperature and pressure parameters to maximize growth efficiency and film quality before costly and time-consuming reactor trials.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":587,"height":425}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_1.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/Debasis Sengupta et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":589,"height":431,"image_format":"jpeg","image_sha256":"c2f0a69e7119d65b367dcbd1985233a37f0e06e7a7df893cbfa1c1aaa7a1cf13","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/test_atomic-layer-deposition_simulation-usecase_18_fig_11.jpg","caption":"Fig. 11. Comparison of numerically calculated growth rate with experiment [3] at the diffusion-limited regime using the reactor geometry shown in Fig. 10.","id":"test/atomic-layer-deposition/simulation-usecase/18/fig_11","sample_id":"atomic-layer-deposition/simulation-usecase/18/fig_11","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot compares numerical predictions with experimental growth rates from Theodoropoulos et al. specifically within the diffusion-limited regime. The data shows a nearly constant growth rate (μm/hr) across the temperature range of 1120–1300 K, which the simulation reproduces accurately.  The experiment data points are represented by diamonds, while the numerical results from the present study are shown as circles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| T (K) | Growth Rate (µm/hr) | Work                 |\\n|-------|---------------------|----------------------|\\n| 1120  | 2.50                | Experimental study   |\\n| 1160  | 2.50                | Numerical (this work)|\\n| 1180  | 2.50                | Experimental study   |\\n| 1200  | 2.55                | Numerical (this work)|\\n| 1220  | 2.50                | Experimental study   |\\n| 1240  | 2.55                | Numerical (this work)|\\n| 1280  | 2.45                | Experimental study   |\\n| 1320  | 2.50                | Numerical (this work)|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process is mass-transport (diffusion) limited.  Since the rate is independent of temperature (flat line), surface reactions are fast and not the bottleneck. Therefore, increasing the temperature further will have no effect. However, increasing the precursor flow rate (flux) would directly increase the supply of reactants reaching the surface, thereby increasing the growth rate.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The simulation results (circles) closely match the experimental data (diamonds) at 1270 K and across the entire temperature range, with overlapping deposition rates near 2.5 μm/hr. This strong agreement demonstrates that the numerical model accurately captures the reactor transport phenomena, including diffusivity, boundary layer effects, and fluid flow, confirming its predictive reliability in this regime.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The constant growth rate indicates the process has entered a diffusion-limited regime. At lower temperatures, growth is typically reaction-limited, where the rate increases exponentially with temperature. Above a certain temperature, the surface reaction becomes so fast that the rate is instead limited by the slower, weakly temperature-dependent physical transport of reactants through the gas boundary layer to the surface, resulting in the observed plateau.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The flat growth rate profile across a 180 K window demonstrates inherent robustness to temperature fluctuations. This means that minor hot or cold spots in a large reactor will not translate into significant thickness variations, leading to better wafer-scale uniformity and higher manufacturing yield without requiring prohibitively precise temperature control.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":5,"width":584,"height":428}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_11.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/Debasis Sengupta et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":592,"height":434,"image_format":"jpeg","image_sha256":"68487a207a6a6934c89385270b76cc3237c82e94c61fdf63460e3f178bcf7fc4","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/test_atomic-layer-deposition_simulation-usecase_18_fig_8.jpg","caption":"Fig. 8. Comparison of numerically calculated growth rates with experiment [5] along the radius of the wafer for the reactor shown in Fig. 7. Case I: pressure 100torr,  $T = 1273\\mathrm{K}$ ,  $\\mathrm{NH}_3 = 2\\mathrm{slpm}$ ,  $\\mathrm{H}_2 = 10\\mathrm{slpm}$ ,  $\\mathrm{TMG} = 0.248\\mathrm{scm}$ .","id":"test/atomic-layer-deposition/simulation-usecase/18/fig_8","sample_id":"atomic-layer-deposition/simulation-usecase/18/fig_8","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the radial GaN growth rate profile in a stagnation-flow reactor under Case I conditions with high H₂ flow. The simulated curve (diamonds with line) follows the experimental data (circles), with only minor deviation near the wafer center. This demonstrates that the model can reasonably capture spatial non-uniformity in convection-dominated growth conditions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Radius (cm) | Growth Rate (mm/hr) | study type |\\n|---|---| --- |\\n| 0 | 17.9 | Experiment |\\n| 0.5 | 6.7 | Numerical (present study) |\\n| 0.5 | 17.9 | Experiment |\\n| 1.0 | 4.2 | Numerical (present study) |\\n| 1.0 | 3.1 | Experiment |\\n| 1.5 | 2.7 | Numerical (present study) |\\n| 1.5 |  4 | Experiment |\\n| 2.0 | 1.9 | Numerical (present study) |\\n| 2.0 | 2.6 | Experiment |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. Defining the reactor geometry and set the condition: high H₂ carrier gas flow.\\n\\n2. Solving the coupled equations for fluid flow, heat transfer, and chemical reactions across the 2D/3D domain.\\n\\n3. Extract the calculated growth rate at multiple points along the wafer radius.\\n\\n4. Plotting these values to create the spatial profile.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, for edge yield optimization, the model is reliable as it tracks the depletion profile accurately. However, it is totally inadequate for predicting center-wafer behavior.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The growth rate decreases from the center to the edge. This is caused by boundary layer development and precursor depletion. Under high convection, fresh precursor is delivered primarily to the center (stagnation point), then flows radially outward, becoming progressively depleted as it deposits material, leading to lower growth rates at larger radii.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, a 15-20% thickness variation would translate directly into an unacceptable variation in device electrical properties, such as threshold voltage and on-resistance, across the wafer. For power transistors, such non-uniformity would drastically reduce manufacturing yield, as devices from different radial positions would not meet the same specification.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":6,"width":584,"height":443}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_8.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/Debasis Sengupta et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":589,"height":450,"image_format":"jpeg","image_sha256":"beb86eca9b4757e9cca03de1f5f6bf4c4eb61b0d3c987c2aec0ebea9ad4ee803","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/test_atomic-layer-deposition_simulation-usecase_18_fig_9.jpg","caption":"Fig. 9. Comparison of numerically calculated growth rates with experiment [5] along the radius of the wafer for the reactor shown in Fig. 7. Case II: pressure 100torr,  $T = 1273\\mathrm{K}$ ,  $\\mathrm{NH}_3 = 2\\mathrm{slpm}$ ,  $\\mathrm{H}_2 = 0.2\\mathrm{slpm}$ ,  $\\mathrm{TMG} = 0.248\\mathrm{scm}$ .","id":"test/atomic-layer-deposition/simulation-usecase/18/fig_9","sample_id":"atomic-layer-deposition/simulation-usecase/18/fig_9","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This graph shows the growth rate profile for the same reactor as similar figure in this work but under Case II conditions with lower flow, where diffusion dominates transport. The simulation matches very closely with the experimental data across the full wafer radius. This agreement demonstrates the model’s strong accuracy in diffusion-controlled regimes.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Radius (cm) | Growth Rate (cm/hr) | Study Type |\\n|---|---|---|\\n| 0.0 | -- | Numerical (present study) |\\n| 0.0 | 2.3 | Experimental |\\n| 0.5 | 1.7 | Numerical (present study) |\\n| 0.5 | 2.2 | Experimental |\\n| 1.0 | 1.76 | Numerical (present study) |\\n| 1.0 | 1.78 | Experimental |\\n| 1.5 | 1.69 | Numerical (present study) |\\n| 1.5 | 1.8 | Experimental |\\n| 2.0 | 1.58 | Numerical (present study) |\\n| 2.0 | 1.8 | Experimental |\\n| 2.5 | 1.47 | Numerical (present study) |\\n| 2.5 | 1.9 | Experimental |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The key change is a large reduction in the carrier gas (H₂) flow rate. Lower flow decreases convective velocity, reducing the Reynolds number and shifting dominance from convection to diffusion. This allows precursors more time to diffuse radially across the wafer, smoothing out concentration gradients and producing a uniform growth rate profile, as shown in Case II.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The diffusion-dominated regime (Case II) is preferable for superior uniformity. The flatter intrinsic profile minimizes center-to-edge variation. The primary trade-off is a lower overall growth rate and potentially poorer precursor vapor-phase mixing, which could lead to compositional non-uniformities in ternary alloys and reduced reactor throughput due to the slower deposition.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The profile here is significantly flatter and more uniform than the center-high, edge-low profile of Case I. Under diffusion-dominated transport, precursor redistribution is governed by Fick's Law, which drives species from high-concentration regions (above the center) toward low-concentration regions (above the edges). This natural smoothing effect reduces radial gradients, leading to the nearly uniform growth rate observed across the wafer radius.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, the model's proven accuracy under diffusion-limited conditions means it can reliably predict how changes in reactor geometry, pressure, or flow will affect the uniformity-critical transport pathways. Engineers can use it to optimize a new reactor's design specifically for uniform precursor diffusion, confidently trading off some convective speed to achieve the target thickness uniformity.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":4,"width":582,"height":454}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/images/fig_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/18/Debasis Sengupta et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":589,"height":456,"image_format":"jpeg","image_sha256":"662deaa67d76ba55660584fb61c21abdd87738661add9174a601103b36ba1174","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/test_atomic-layer-deposition_simulation-usecase_34_figure_7.jpg","caption":"Figure 7. Two-point probe resistance measurement of Pt ALD-graphene contacts, Pt ALD growth was initiated by a  $7\\min \\mathrm{H}_{2}$  plasma treatment. The resistance is averaged over six devices and divided by a correction factor  $C$ , which takes into account the increase in the contact area with increasing ring size. Fitting Equation (3), the contact resistance is determined to be  $0.5\\pm 0.2\\mathrm{k}\\Omega$ .","id":"test/atomic-layer-deposition/simulation-usecase/34/figure_7","sample_id":"atomic-layer-deposition/simulation-usecase/34/figure_7","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows a linear relationship between the gap size (in µm) and the resistance (corrected for contact area) in Ω.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gap Size d (µm) | R<sub>tot</sub>/C (Ω) |\\n|---|---|\\n| 0 | 0 |\\n| 5 | 20 |\\n| 10 | 40 |\\n| 15 | 50 |\\n| 17.5 | 70 |\\n| 22.5 | 80 |\\n| 25 | 90 |\\n| 35 | 120 |\\n| 40 | 150 |\\n| 47.5 | 170 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"7 minutes.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.5 ± 0.2 kΩ µm, which is at the intercept of the vertical axis.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Roughly 150 Ω.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It increases linearly with a slope of about 1.01 ± 0.02 kΩ µm^-1.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":667,"height":500}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/images/figure_7.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/34/René H. J. Vervuurt et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":667,"height":500,"image_format":"jpeg","image_sha256":"bd58daa8f1017c970dbeab903eb44509d6822621aac7bda4aaacb2a7e2a6cc36","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/test_atomic-layer-deposition_simulation-usecase_40_fig_6.jpg","caption":"Fig. 6. Process efficiency as a function of the time required to perform 100 ALD cycles in the case of  $\\mathrm{ZnO}$  deposition by SALD.","id":"test/atomic-layer-deposition/simulation-usecase/40/fig_6","sample_id":"atomic-layer-deposition/simulation-usecase/40/fig_6","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents the relationship between deposition efficiency (η₍dep₎) and the time required to complete 100 SALD cycles for ZnO deposition at different DEZ flow rates (15, 30, and 50 sccm). The data reveal a trade-off between throughput and efficiency, with 30 sccm providing the highest efficiency at moderate cycle times, while lower and higher flow rates lead to reduced efficiency due to under- or over-supply of precursor.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| DEZ Flow Rate (sccm) | Time for 100 ALD cycles (s) | Deposition efficiency (η_dep) |\\n|---------------------|----------------------------|-------------------------------|\\n| 15                  | 45                         | 0.045                         |\\n| 15                  | 55                         | 0.048                         |\\n| 15                  | 65                         | 0.050                         |\\n| 15                  | 80                         | 0.043                         |\\n| 15                  | 90                         | 0.037                         |\\n| 30                  | 30                         | 0.052                         |\\n| 30                  | 35                         | 0.054                         |\\n| 30                  | 40                         | 0.055                         |\\n| 30                  | 45                         | 0.053                         |\\n| 30                  | 50                         | 0.050                         |\\n| 50                  | 30                         | 0.047                         |\\n| 50                  | 35                         | 0.045                         |\\n| 50                  | 40                         | 0.043                         |\\n| 50                  | 45                         | 0.038                         |\\n| 50                  | 50                         | 0.036                         |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This figure enables identification of an optimal operating window where deposition efficiency is maximized without sacrificing throughput. The results indicate that intermediate DEZ flow rates (e.g., 30 sccm) balance sufficient precursor supply with minimal gas-phase loss, allowing faster processing while maintaining high efficiency. Such optimization is critical for scaling SALD processes to high-throughput manufacturing of ZnO films in applications like transparent electronics and sensors.\"}]},{\"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\":\"DEZ precursor flow rate (15, 30, 50 sccm), Time required for 100 SALD cycles, Precursor utilization efficiency, Gas–surface interaction dynamics\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"30 sccm DEZ shows the highest deposition efficiency (~0.055) at intermediate cycle times (~35–50 s).\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":605,"height":547}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/images/fig_6.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-deposition/simulation-usecase/40/Viet Huong Nguyen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":605,"height":547,"image_format":"jpeg","image_sha256":"582993bb84354e3279e96a09317139dc5af6497c332ecad21958fc0a0ca13747","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/test_atomic-layer-etching_experimental-usecase_12_fig2.jpg","caption":"FIG.2. Saturation curves for (a)  $\\mathrm{SF}_6$  plasma exposure using a  $0.5s$  TMA dose, (b) the pump step following  $\\mathrm{SF}_6$  plasma exposure, (c) TMA dosing using a  $5s$ $\\mathrm{SF}_6$  plasma exposure, and (d) the pump step following  $\\mathrm{SF}_6$  TMA dose. All experiments were conducted at a substrate temperature of  $260^{\\circ}C$  with  $10s$  TMA hold steps. The dashed lines are guides to the eye.","id":"test/atomic-layer-etching/experimental-usecase/12/fig2","sample_id":"atomic-layer-etching/experimental-usecase/12/fig2","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"scatter plot"},{"panel_id":"d","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"EPC increases rapidly with SF₆ plasma exposure time and saturates at ≈3.4 Å after ~5s for a 0.5s TMA dose at 260 °C. The auto summary did not identify the saturation threshold or the self-limiting nature of the plasma step. The manual summary clarifies the time required to reach full fluorination.\"},{\"panel_id\":\"b\",\"text\":\"EPC remains constant at ≈3.1–3.3 Å across the full plasma pump-time range following SF₆ exposure. The auto summary described only a generic dependence, whereas the manual version highlights that pump time does not influence EPC once saturation is reached.\"},{\"panel_id\":\"c\",\"text\":\"EPC shows a sharp onset between 0.2 and 0.4 s TMA dosing and saturates near ≈3.3 Å for a fixed 5s SF₆ plasma exposure. The auto summary missed this threshold behavior and the implication of a self-limiting ligand-exchange step. The manual summary explicitly identifies the dosing time required for saturation.\"},{\"panel_id\":\"d\",\"text\":\"EPC is insensitive to TMA pump time and remains at ≈3.2–3.3 Å across all values shown. The auto summary implied a dependence without interpretation. The manual summary clarifies that the reaction completes during dosing and does not require extended pumping.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| SF plasma time (s) | EPC (Å) |\\n|---|---|\\n| 0 | 0 |\\n| 5 | 2 |\\n| 10 | 3 |\\n| 15 | 3 |\\n| 20 | 3 |\\n| 25 | 3 |\\n| 30 | 3 |\"},{\"panel_id\":\"b\",\"text\":\"| Plasma pump time (s) | EPC (Å) |\\n|---|---|\\n| 10 | 2.9 |\\n| 20 | 3 |\\n| 30 | 3 |\"},{\"panel_id\":\"c\",\"text\":\"| TMA dosing time (s) | EPC (Å) |\\n|---|---|\\n| 0 | 0 |\\n| 0.1 | -0.1 |\\n| 0.2 | 0 |\\n| 0.4 | 2 |\\n| 0.6 | 3 |\\n| 0.8 | 3 |\"},{\"panel_id\":\"d\",\"text\":\"| TMA pump time (s) | EPC (Å) |\\n|---|---|\\n| 5 | 3.25 |\\n| 10 | 3 |\\n| 30 | 3 |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.3 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At this dose time there is a negative EPC indicating that there is deposition of aluminum.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"≥0.4 s\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It reveals that a minimum coverage of surface fluorine species (from SF₆ plasma) must be present before TMA can effectively react and release volatile etch products.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The TMA dose saturates quickly for doses longer than 0.3 s. Doses shorter than 0.2 s were found to cause a negligible thickness change, which is likely related to the consumption of TMA by AlF3 deposition on the reactor walls (100 °C) that are at a lower temperature than the substrate holder (300 \\u0002C). Previous work has shown that AlF3 ALD using SF6 plasma and TMA readily occurs at low temperatures. It is, therefore, likely that a significant amount of dosed TMA is adsorbed on the reactor walls before it can reach the substrate\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3.1 Å.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 0.2 s (based on the EPC crossing from negative to positive).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At about ~0.3 s TMA dose. It is clearly saturated by ~0.4–0.5 s\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process would be 5-5-0.5-5.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The initial decrease suggests that some active TMA molecules are pumped away before reacting if the purge is too short. Therefore, the optimal pump time is just beyond the dip (~10-15s), where EPC stabilizes. This maximizes precursor utilization by allowing sufficient reaction time while minimizing waste and cycle time, which is critical for cost-effective high-volume production.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The measured EPC is higher than the literature results reported for thermal ALE of Al2O3 with HF/TMA at a comparable temperature.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes, the saturation of EPC in Figure 2 indicate self-limiting surface reactions characteristic of ALE.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"EPC increases with both SF₆ plasma exposure time and TMA dosing time until a saturation level is reached, indicating self-limiting surface reactions. In contrast, EPC shows little dependence on plasma pump time or TMA pump time.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The figure shows the etching per cycle as a function of plasma and precursor dose time.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SF6.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The saturated time is already around 5s, however the process soft-saturates.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"5 s\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experimentally determined Etch Per Cycle (EPC) is approximately 3.1 Å/cycle.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~2 seconds (the EPC saturates after the initial sharp rise between 0-2s).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, similar to the modification steps observed for many other ALE processes.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"260 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the film is exposed to SF6 plasma, the formed fluorine radicals react quickly with surface Al2O3 molecules to form AlF3. This AlF3 then acts as a protective layer, which inhibits the remaining fluorine radicals from reacting with Al2O3 below the AlF3 film. However, this AlF3 does not completely prevent the fluorine radicals from reaching the lower layers of Al2O3, so more AlF3 gets formed as the exposure time of SF6 increases. After about 3 seconds, this reaction does become diffusion-limited, which explains the observed change.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The initial fast increase comes from rapid fluorination of the surface. As the fluorinated layer thickens, additional fluorination becomes harder because reactants must penetrate through this layer. The fluorinated layer therefore acts as a diffusion barrier, leading to a slower EPC increase.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around ~5 s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Figure demonstrates that both SF6 plasma exposure and TMA dosing exhibit self-limiting behavior, typical of ALE processes. For SF6 plasma, the EPC rapidly increases for short exposure times and saturates around 5 s, indicating full surface fluorination. For TMA, EPC saturates after ~0.3–0.5 s, reflecting complete adsorption on the fluorinated surface. Pump steps of 10 s between half-cycles are sufficient to prevent parasitic reactions, ensuring that etching occurs only when alternating plasma and TMA doses are applied. Overall, the curves confirm the controlled, self-limiting nature of the plasma ALE process.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SF₆ plasma exposure and TMA dosing.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both subfigure b and d show that 5 s is sufficient purge time.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Neither; EPC remains constant at ≈3.3 Å.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Plasma exposure time is critical; EPC rises sharply until saturation at ~2s, indicating the time needed for surface fluorination. In contrast, plasma pump time has minimal effect after ~10s, suggesting that reactant removal is not rate-limiting. The trade-off is that precise control of the active plasma step is essential for consistent etching, while the purge step offers a wider, more forgiving time window, allowing for throughput optimization without sacrificing etch rate.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A minimum plasma pump steps of 10 s are sufficient to avoid a parasitic etching component\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 s.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SF₆ plasma exposure time, TMA dosing time.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"EPC increases rapidly at short SF6 plasma exposure times and then saturates at ~3 Å/cycle, TMA dosing exhibits quick saturation above ~0.3 s, with little additional growth for longer doses, Pump steps (both after plasma and TMA) show no significant change in EPC after 10 s, confirming removal of excess reactants, The self-limiting behavior is consistent for both plasma and TMA, ensuring precise etching per cycle\"}]}]","bbox":[{"panel_id":"a","x":8,"y":0,"width":322,"height":371},{"panel_id":"b","x":358,"y":0,"width":324,"height":369},{"panel_id":"c","x":9,"y":384,"width":321,"height":361},{"panel_id":"d","x":363,"y":384,"width":319,"height":359}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/12/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/12/images/fig2.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/12/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/12/Isotropic plasma atomic layer etching of Al2O3 using a fluorine containing plasma and Al(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"12","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":683,"height":745,"image_format":"jpeg","image_sha256":"282f63a976d7ce9cf90149fa5b7a986aa6289d45e82ebe9f1faf1aa16105f66f","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/test_atomic-layer-etching_experimental-usecase_20_figure_4.jpg","caption":"Figure 4. Temperature dependence of AlN ALE with HF and  $\\mathrm{BCl}_3$  as the reactants as measured by the change in integrated absorbance per cycle at 200, 250, 300, and  $350^{\\circ}\\mathrm{C}$ . The integrated absorbance was measured from 400 to  $1000\\mathrm{cm}^{-1}$","id":"test/atomic-layer-etching/experimental-usecase/20/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/20/figure_4","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between temperature and the change in absorption intensity per cycle at 200 °C, 250 °C, 300 °C and 350 °C. As temperature increases, there is a negative trend in the change in absorption intensity.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Δ Int. Abs./ Cycle (cm⁻¹) |\\n|-----------------|--------------------------|\\n| 200             | 0                        |\\n| 250             | -1.5                     |\\n| 300             | -2.2                     |\\n| 350             | -3.0                     |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"With this figure it is possible to see the temperature dependence of the AlN cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the temperature increases, higher absorbance changes and higher etching rates are observed.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 350 °C\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"This means that there is negligible AlN etching.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":655,"height":493}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/Thermal Atomic Layer Etching of Aluminum Nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":658,"height":500,"image_format":"jpeg","image_sha256":"18339f82d20902ddebfb02373f97e45687ff17fe256c834ba330d3aa418e9a3c","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/test_atomic-layer-etching_experimental-usecase_20_figure_9.jpg","caption":"Figure 9. Etch rate changes during AlN ALE vs  $\\mathrm{BCl}_3$  exposure at 255  $^\\circ \\mathrm{C}$  using an  $\\mathrm{XeF}_2$  exposure of 0.060 Torr s.","id":"test/atomic-layer-etching/experimental-usecase/20/figure_9","sample_id":"atomic-layer-etching/experimental-usecase/20/figure_9","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between BCl₃ exposure time (in Torr·s) and etch rate (in Å/cycle) at a constant temperature of 255°C and with a constant pressure of 0.060 Torr s of XeF₂, and varying BCl₃ pressure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| BCl₃ Exposure (Torr·s) | Etch Rate (Å/cycle) |\\n| --- | --- |\\n| 0 | 0.40 |\\n| 50 | 0.65 |\\n| 100 | 0.80 |\\n| 150 | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experiment was conducted at a temperature of 255 °C and a pressure of 0.060 Torr s of XeF2.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As BCl3 exposure increases, so does the etch rate, rising from 0.4 to 0.8 Å/cycle.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At approximately 50 Torr s.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":0,"y":1,"width":656,"height":513}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/images/figure_9.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/20/Thermal Atomic Layer Etching of Aluminum Nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":661,"height":517,"image_format":"jpeg","image_sha256":"33ed721febee3e0aa7437fd16f291cf341930b9ea9d6b0f78704cae6a07a5c3f","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/test_atomic-layer-etching_experimental-usecase_25_figure_10.jpg","caption":"Figure 10. Arrhenius plot of temperature-dependent etch rates for  $\\mathrm{WO}_3$  ALE. Slope of the Arrhenius plot yields an activation barrier of  $8.6\\mathrm{kcal / mol}$ .","id":"test/atomic-layer-etching/experimental-usecase/25/figure_10","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_10","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the logarithmic relationship between the inverse temperature (1000/T) and the etch rate of WO₃, which is called an Arrhenius plot.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| 1000/T (K⁻¹) | ln [WO₃ Etch Rate (Å/cycle)] |\\n|---|---|\\n| 2.1 | 1.4 |\\n| 2.2 | 0.9 |\\n| 2.3 | 0.5 |\\n| 2.4 | 0.1 |\\n| 2.5 | -0.3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The slope gives information regarding the activation energy. For the reaction to occur.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there is an uncertainty. This is mainly caused by the fact that not all datapoints are precisely located on the line.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A negative value on this y-axis represents an etch rate that is smaller than 1 angstrom per cycle.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The temperature is a probe of the kinetic energy reactants help. Usually a reaction requires an activation energy and by providing a certain temperature, the kinetic energy will increase leading to the reaction to occur.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":8,"width":655,"height":558}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_10.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":670,"height":570,"image_format":"jpeg","image_sha256":"148214790a55dca57c3e1346c32a480176df572b039c7a0a8fba1b7c4d3a2515","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/test_atomic-layer-etching_experimental-usecase_25_figure_11.jpg","caption":"Figure 11. W thickness versus number of cycles during W ALE at 207  $^\\circ \\mathrm{C}$  using  $\\mathrm{O}_2 / \\mathrm{O}_3$ ,  $\\mathrm{BCl}_3$ , and HF as reactants. Etch rate during W ALE is  $2.56\\mathrm{\\AA} / \\mathrm{cycle}$ .","id":"test/atomic-layer-etching/experimental-usecase/25/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_11","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A line chart showing the decrease in W thickness (Å) as the number of cycles increases, with a slope of 2.56.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | W Thickness (Å) |\\n|------------------|-----------------|\\n| 20              | 263             |\\n| 30              | 239             |\\n| 40              | 210             |\\n| 50              | 183             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"207 degrees Celcius\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After the 50 cycles, around 185 angstrom is still present, the EPC is 2.56 angstrom per cycle, so still around 70 cycles are required.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"225 angstrom.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The EPC is 0.256 nm/cycle and this does relate to ALE as the etch per cycle is not too high.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":669,"height":551}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_11.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":558,"image_format":"jpeg","image_sha256":"1f356e7364dd1230a347871ec3d444980dbbf9fe58552a03a2e0517a255efce0","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/test_atomic-layer-etching_experimental-usecase_25_figure_14.jpg","caption":"Figure 14. W etch rate versus  $\\mathrm{O}_2 / \\mathrm{O}_3$  exposure during W ALE at  $207^{\\circ}\\mathrm{C}$ .  $\\mathrm{BCl}_3$  and HF exposures were held at 500 and 2800 mTorr s, respectively.","id":"test/atomic-layer-etching/experimental-usecase/25/figure_14","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_14","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between oxygen/ozone exposure (in mTorr s) and the etch rate of tungsten (W) in angstrom per cycle. The etch rate increases with increasing exposure up to 3000 mTorr s, after which it plateaus.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| O₂/O₃ Exposure (mTorr s) | W Etch Rate (Å/cycle) |\\n|---|---|\\n| 0 | 0.0 |\\n| 1000 | 0.7 |\\n| 2000 | 1.9 |\\n| 3000 | 2.4 |\\n| 4000 | 2.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The feature that is visible is self-limiting. This is visible by the formation of a plateau after 3000 mTorr s.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2.4 angstrom per cycle\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As O3 is usually more reactive than O2, a higher fraction of O3 is likely to result in a decrease in required exposure time.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":3,"y":1,"width":663,"height":554}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_14.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":670,"height":556,"image_format":"jpeg","image_sha256":"cd32800c523792d16f82286d5522ac5c84a8187c8104294cc4e5f595d2c9d941","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/test_atomic-layer-etching_experimental-usecase_25_figure_15.jpg","caption":"Figure 15. W etch rate versus reactant exposure during W ALE. (a)  $\\mathrm{BCl}_3$  exposure was varied with HF and  $\\mathrm{O}_2 / \\mathrm{O}_3$  exposures held at 2800 and  $3150~\\mathrm{mTorr}$  s, respectively. (b) HF exposure was varied with  $\\mathrm{BCl}_3$  and  $\\mathrm{O}_2 / \\mathrm{O}_3$  exposures held at 500 and  $3150~\\mathrm{mTorr}$  s, respectively.","id":"test/atomic-layer-etching/experimental-usecase/25/figure_15","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_15","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch rate of W increases with increasing BCl₃ exposure at a constant temperature of 207°C, and constant HF and O2/O3 exposures.\"},{\"panel_id\":\"b\",\"text\":\"The etch rate of W increases with increasing HF exposure at a constant temperature of 207°C, and constant BCl3 and O2/O3 exposures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| BCl₃ Exposure (mTorr s) | W Etch Rate (Å/cycle) |\\n|---|---|\\n| 0 | 0.0 |\\n| 200 | 2.2 |\\n| 300 | 2.4 |\\n| 500 | 2.4 |\"},{\"panel_id\":\"b\",\"text\":\"| HF Exposure (mTorr s) | W Etch Rate (Å/cycle) |\\n|---|---|\\n| 0 | 0.0 |\\n| 1000 | 1.8 |\\n| 2000 | 2.4 |\\n| 3000 | 2.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"If this is not the case the, the outcome of the BCl3 exposure sweep will not represent the highest achievable etch rate. In addition, it will not represent ALE, as the reactions have not been performed till saturation is achieved.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.5 angstrom per cycle\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"2.5 angstrom per cycle, and this does correspond to the other achieved EPCs.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It could be that temperature has an influence on the etch per cycle which might interfere with the results achieved by changing the HF exposure.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":636,"height":532},{"panel_id":"b","x":1,"y":533,"width":643,"height":536}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_15.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_15.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":647,"height":1073,"image_format":"jpeg","image_sha256":"e42cbf1cada2777832cf7f632214b6404e0b7c64120e5f0aa626860f423234e3","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/test_atomic-layer-etching_experimental-usecase_25_figure_16.jpg","caption":"Figure 16.  $\\mathrm{WO}_3$  thickness versus number of  $\\mathrm{O}_2 / \\mathrm{O}_3$  exposures for initial W ALD film. Each  $\\mathrm{O}_2 / \\mathrm{O}_3$  exposure was 70 mTorr for  $45\\mathrm{~s~}$","id":"test/atomic-layer-etching/experimental-usecase/25/figure_16","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_16","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The graph shows the thickness of WO₃ as a function of the number of O₂/O₃ exposures at 207°C, where the thickness increases for increasing number of exposures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of O₂/O₃ Exposures | WO₃ Thickness (Å) |\\n|---|---|\\n| 0 | 33 |\\n| 1 | 33 |\\n| 2 | 33 |\\n| 3 | 35 |\\n| 4 | 37 |\\n| 5 | 42 |\\n| 6 | 45 |\\n| 8 | 49 |\\n| 10 | 52 |\\n| 12 | 55 |\\n| 16 | 59 |\\n| 20 | 61 |\\n| 24 | 62 |\\n| 28 | 63 |\\n| 32 | 64 |\\n| 34 | 65 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1 up until 3 exposures\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The is no perfect saturation achieved for ever increasing number of exposures, but a linear line is observed. This is usually referred to as soft-saturation.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Underlayin tungsten (W).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"28 angstrom\"}]}]","bbox":[{"panel_id":"a","x":5,"y":1,"width":662,"height":547}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_16.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_16.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":670,"height":550,"image_format":"jpeg","image_sha256":"9025016d377ffeed455d2379654c06f855485fe2a53fef80e3c82f009e00b519","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/test_atomic-layer-etching_experimental-usecase_25_figure_17.jpg","caption":"Figure 17.  $\\mathrm{WO}_3$  removal after W ALE using  $\\mathrm{BCl}_3$  and HF as reactants. (a)  $\\mathrm{WO}_3$  thickness versus number of cycles showing reduction of  $\\mathrm{WO}_3$  thickness to limiting value of  $\\sim 3\\mathrm{\\AA}$ . (b) W thickness versus number of cycles showing that W thickness remains nearly constant during  $\\mathrm{WO}_3$  removal.","id":"test/atomic-layer-etching/experimental-usecase/25/figure_17","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_17","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The thickness of WO₃ decreases rapidly at first and then plateaus for increasing number of cycles.\"},{\"panel_id\":\"b\",\"text\":\"The thickness of W remains relatively constant for increasing number of cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | WO₃ Thickness (Å) |\\n|---|---|\\n| 0 | 20 |\\n| 3 | 14 |\\n| 6 | 7 |\\n| 12 | 4 |\\n| 18 | 4 |\\n| 24 | 3 |\\n| 30 | 3 |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Cycles | W Thickness (Å) |\\n|---|---|\\n| 0 | 219 |\\n| 3 | 217 |\\n| 6 | 217 |\\n| 12 | 216 |\\n| 18 | 216 |\\n| 24 | 215 |\\n| 30 | 215 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3 angstrom\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the etching of W, it must first be modified into WO3. As this process does not contain a O2/O3 step, this does not happen.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process is selective. Only WO3 is etched whereas W is unharmed. This means the process is selective towards WO3 w.r.t. W.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"None\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":661,"height":414},{"panel_id":"b","x":6,"y":432,"width":659,"height":454}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_17.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_17.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":669,"height":891,"image_format":"jpeg","image_sha256":"ca193ef1d48f1405b159d1896e5ea230bac41fde70aee21bb3314872129e6721","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/test_atomic-layer-etching_experimental-usecase_25_figure_3.jpg","caption":"Figure 3.  $\\mathrm{WO}_3$  thickness versus number of cycles showing  $\\mathrm{WO}_3$  ALE at  $207^{\\circ}\\mathrm{C}$  using  $\\mathrm{BCl}_3$  and HF as reactants. W film under  $\\mathrm{WO}_3$  layer acts as an etch stop.","id":"test/atomic-layer-etching/experimental-usecase/25/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_3","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the linear decrease in thickness of WO₃ with increasing number of cycles at 207°C. The text annotations indicate that W acts as an etch stop.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | WO₃ Thickness (Å) |\\n|---|---|\\n| 0 | 150 |\\n| 5 | 120 |\\n| 10 | 100 |\\n| 15 | 75 |\\n| 20 | 50 |\\n| 25 | 30 |\\n| 30 | 10 |\\n| 35 | 10 |\\n| 40 | 10 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate can be determined by calculating the slope, which gives a value around 5 angstrom per cycle.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process without O2/O3 exposure is selective for WO3 w.r.t. W, so there is no etching of W.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"15 nm.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there will be, tungsten will oxidize after exposure to ambient so there will be a thin layer of WO3 at the surface.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":665,"height":545}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":550,"image_format":"jpeg","image_sha256":"2d6241cacf354d8bc02d2e9e2651ca695e1ea372aad27ffd5b48490afcdc1591","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/test_atomic-layer-etching_experimental-usecase_25_figure_4.jpg","caption":"Figure 4.  $\\mathrm{WO}_3$  thickness versus number of cycles showing  $\\mathrm{WO}_3$  ALE at  $207^{\\circ}\\mathrm{C}$  using  $\\mathrm{BCl}_3$  and HF as reactants. Etch rate during  $\\mathrm{WO}_3$  ALE is  $4.18\\mathrm{\\AA}$  /cycle.","id":"test/atomic-layer-etching/experimental-usecase/25/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/25/figure_4","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the decrease in thickness of WO₃ with increasing number of cycles at 207°C. The etch rate is calculated as 4.18 Å/cycle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | WO₃ Thickness (Å) |\\n|---|---|\\n| 10 | 135 |\\n| 15 | 110 |\\n| 20 | 85 |\\n| 25 | 60 |\\n| 30 | 50 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The slope of the line is equal to the etch rate, which is 4.18 angstrom/cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"85 angstrom.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch rate entails that is a thickness over a time unit that is removed, usually a cycle is seen as this timeunit. However, this does not always hold.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Still between 10 and 15 cycles\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":667,"height":541}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/images/figure_4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/25/WO3 and W Thermal Atomic Layer Etching Using Conversion-Fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":673,"height":547,"image_format":"jpeg","image_sha256":"ebe7bf428c9af2a26875be99a56d5105a444373512ee0e1684c6f2443b6b5cea","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/test_atomic-layer-etching_experimental-usecase_37_fig_3.jpg","caption":"FIG. 3. Growth per cycle (GPC) of  $\\mathrm{TiO_2}$  thin films deposited on Si-OH surface measured by in situ ellipsometry. One of the precursor dosing times is held constant at  $300\\mathrm{ms}$  while the dosing time of (a)  $\\mathrm{TiCl_4}$  and (b)  $\\mathrm{H}_2\\mathrm{O}$  is varied.","id":"test/atomic-layer-etching/experimental-usecase/37/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/37/fig_3","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The growth per cycle remains constant at approximately 0.05 nm/cycle across various TiCl4 doses ranging from 0 ms to 350 ms.\"},{\"panel_id\":\"b\",\"text\":\"The growth per cycle remains constant at approximately 0.05 nm/cycle across various H2O doses ranging from 0 ms to 350 ms.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TiCl4 dose (ms) | Growth per cycle (nm/cycle) |\\n|---|---|\\n| 0 | 0.00 |\\n| 50 | 0.05 |\\n| 100 | 0.05 |\\n| 150 | 0.05 |\\n| 200 | 0.05 |\\n| 250 | 0.05 |\\n| 300 | 0.05 |\\n| 350 | 0.05 |\"},{\"panel_id\":\"b\",\"text\":\"| H2O Dose (ms) | Growth per cycle (nm/cycle) |\\n|---|---|\\n| 0 | 0.00 |\\n| 50 | 0.04 |\\n| 100 | 0.05 |\\n| 150 | 0.05 |\\n| 200 | 0.05 |\\n| 250 | 0.05 |\\n| 300 | 0.05 |\\n| 350 | 0.05 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"170 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.05 nm/cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The GPC as a function of TiO2 exposure plateaus at 0.05 nm/cycle after 50 ms of exposure time. The same is observed for the water exposure after about 150 ms. After these points, the growth per cycle is no longer dependent on exposure time, indicating self-limiting growth. This means that this process meets the requirement of ALD.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.05 nm/cycle.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":667,"height":550},{"panel_id":"b","x":1,"y":551,"width":667,"height":535}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/37/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/37/images/fig_3.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/37/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/37/Effect of reactant dosing on selectivity during area-selective.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"37","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":1091,"image_format":"jpeg","image_sha256":"b11943e7dba153b820183e93fe05c73702087a4cf25f6e961c46acc0b39acbe0","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/test_atomic-layer-etching_experimental-usecase_44_fig_4.jpg","caption":"FIG. 4. (Color online) Film thickness vs number of AlN ALE reaction cycles at  $275^{\\circ}\\mathrm{C}$  in pure AlN region of AlN film. Change of thickness vs number of AlN ALE reaction cycles gives an etch rate of  $0.38\\mathrm{\\AA}$  /cycle.","id":"test/atomic-layer-etching/experimental-usecase/44/fig_4","sample_id":"atomic-layer-etching/experimental-usecase/44/fig_4","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the thickness of an aluminum nitride (AlN) film as a function of the number of Atomic Layer Etching (ALE) reaction cycles at 275 °C, specifically in the pure AlN region of the film. From the linear change in thickness with cycle count, an average etch rate of 0.38 Å/cycle is determined.  The number of cycles were taken from 472 to 492.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle Number | Thickness (Å) |\\n|---|---|\\n| 472 | 412 |\\n| 474 | 411 |\\n| 476 | 410 |\\n| 478 | 409 |\\n| 480 | 408 |\\n| 482 | 407 |\\n| 484 | 407 |\\n| 486 | 406 |\\n| 488 | 405 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"HF exposure: Fluorinates the surface to form an AlF₃ layer.\\n\\nSn(acac)₂ exposure: Performs a ligand-exchange reaction to volatilize the fluorinated layer.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The trade-off is between process efficiency and surface preparation. The pure AlN etches efficiently, but the unavoidable native oxide acts as a \\\"slow-etch barrier.\\\" This forces a choice: accept a long, non-productive overetch to clear the oxide, or add a separate pre-cleaning step to remove it, adding complexity to the process flow.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Its high thermodynamic stability (sublimation point 1291°C), which allows it to remain as a solid, well-defined surface layer for the Sn(acac)₂ to react with.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The four-center transition state ensures the reaction is highly specific and occurs only at the immediate interface between the Sn(acac)₂ precursor and the solid AlF₃ surface layer. This localized, surface-limited reaction prevents spontaneous gas-phase decomposition or bulk etching. Consequently, the etch process is intrinsically self-limiting, removing exactly one reacted layer per cycle and enabling atomic-scale depth control critical for nanofabrication.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":598,"height":475}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/44/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/44/images/fig_4.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/44/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/44/Thermal atomic layer etching of crystalline aluminum nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"44","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:image_caption"},"width":606,"height":480,"image_format":"jpeg","image_sha256":"9637ec1086968b26886134527ef7ce44a099607a119c2b79e426ac34dc02b085","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/test_atomic-layer-etching_experimental-usecase_47_FIG8_a.jpg","caption":"FIG. 8. (a)  $R_{\\mathrm{q}}$  of molybdenum films as a function of the number of ALE cycles at  $300^{\\circ}\\mathrm{C}$  as measured by AFM. AFM images of (b) an unetched film with a thickness of around  $55\\mathrm{nm}$  and (c) a film etched for 200 cycles, with a remaining thickness of  $41.6\\mathrm{nm}$ .","id":"test/atomic-layer-etching/experimental-usecase/47/FIG8_a","sample_id":"atomic-layer-etching/experimental-usecase/47/FIG8_a","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The image plots number of ALE cycles and the surface roughness.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| no. cycles | R<sub>q</sub> (nm) |\\n|---|---|\\n| 0 | 3.2 |\\n| 50 | 3.5 |\\n| 100 | 3.6 |\\n| 200 | 4.2 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"AFM was used to study the film morphology before and after partial etching at 300 °C\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A slight increase of the roughness was seen. The average Rq of the unetched films was 3.3 nm, and for the film etched for 200 cycles, i.e., 21 nm in thickness, the average Rq was 4.3 nm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, we need the AFM images with 2D/3D contour plot to evaluate the grain boundaries and morphology more precisely\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":664,"height":570}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG8_a.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/images/FIG8_a.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/experimental-usecase/47/Atomic layer etching of molybdenum with O2 and NbCl5.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"47","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:image_caption"},"width":678,"height":509,"image_format":"jpeg","image_sha256":"db11c50f028e2070658cc258c7c5482a4b26fa17a268d17763cd513b5e692a2d","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/test_atomic-layer-etching_simulation-usecase_18_fig7.jpg","caption":"FIG.7. (a) Scatter plot for Hf-F coverage vs total HF coverage for the surface coverage values in Table III. Plots (b) and (c) show the change in binding energy per square nanometer with an increase in HF and Hf-F coverage, respectively.","id":"test/atomic-layer-etching/simulation-usecase/18/fig7","sample_id":"atomic-layer-etching/simulation-usecase/18/fig7","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows a scatter plot with points representing the relationship between coverage of HF per nm² and Hf-F per nm². The coverage of Hf-f increases as well as the coverage of Hf.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays a scatter plot with points indicating the binding energy (E_bind) in eV/nm² against the coverage of HF per nm². Energy decreases as the coverage increases.\"},{\"panel_id\":\"c\",\"text\":\"The figure shows a scatter plot with points representing the binding energy (E_bind) in eV/nm² against the coverage of Hf-F per nm². Energy decreases as the coverage increases.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Coverage [HF/nm²] | Coverage [Hf-F/nm²] |\\n|---|---|\\n| 0 | 0 |\\n| 5 | 4 |\\n| 10 | 6.5 |\\n| 15 | 7 |\\n| 20 | 9 |\"},{\"panel_id\":\"b\",\"text\":\"| Coverage [HF/nm²] | E_bind [eV/nm²] |\\n|---|---|\\n| 0 | -1 |\\n| 5 | -10 |\\n| 10 | -15 |\\n| 15 | -18 |\\n| 20 | -21 |\"},{\"panel_id\":\"c\",\"text\":\"| Coverage [HF/nm²] | E_bind [eV/nm²] |\\n|---|---|\\n| 0 | -2 |\\n| 2 | -4 |\\n| 4 | -10 |\\n| 6 | -18 |\\n| 8 | -20 |\\n| 10 | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen in the subfigure a, the HF coverage resulted in a complete dissociation. This lay along the correlation line.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is possible to observe a plateau close to 9.0 Hf-F/nm.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the HF coverage increases from 0 to 20 HF/nm2, the binding energy decreases to more negative values.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":11,"width":326,"height":361},{"panel_id":"b","x":361,"y":14,"width":314,"height":357},{"panel_id":"c","x":694,"y":14,"width":313,"height":357}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/images/fig7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/images/fig7.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/18/Origin of enhanced thermal atomic layer etching of amorphous HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"18","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":1009,"height":372,"image_format":"jpeg","image_sha256":"8a351b2c4fba36f8a8263e32a7bb144a62b34095bc932f3fde332426303b17b4","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/test_atomic-layer-etching_simulation-usecase_25_figure_15.jpg","caption":"Figure 15. Experimental etch rates for the thermal ALE of  $\\mathrm{HfO_2}$  and  $\\mathrm{ZrO_2}$  using precursors HF and  $\\mathrm{Sn(acac)}_2,^{4,9}\\mathrm{Al(CH_3)_3,^{4}AlCl(CH_3)_2,^{4,12}}$ $\\mathrm{SiCl_4}^4$  and  $\\mathrm{TiCl_4}$  111 (a) Etch rates for the thermal ALE of  $\\mathrm{HfO_2}$  using HF and  $\\mathrm{Sn(acac)}_2$  for the temperature range of  $150 - 250^{\\circ}C$  b) Etch rates for the thermal ALE of  $\\mathrm{HfO_2}$  and  $\\mathrm{ZrO_2}$  using HF and  $\\mathrm{Sn(acac)}_2 / \\mathrm{AlCl(CH_3)_2 / Al(CH_3)_3 / SiCl_4}$  at the temperatures shown. (c) Etch rates for the thermal ALE using HF and  $\\mathrm{TiCl_4}$  of  $\\mathrm{HfO_2}$  for the temperature range  $200 - 300^{\\circ}C$  and  $\\mathrm{ZrO_2}$  at  $250^{\\circ}C$  d) Etch rates for the thermal ALE of  $\\mathrm{HfO_2}$  and  $\\mathrm{ZrO_2}$  using HF and  $\\mathrm{AlCl(CH_3)_2}$  for the temperature range  $200 - 300^{\\circ}C$","id":"test/atomic-layer-etching/simulation-usecase/25/figure_15","sample_id":"atomic-layer-etching/simulation-usecase/25/figure_15","subset":"scatter-plot","split":"test","classification":[{"panel_id":"a","label":"scatter plot"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"scatter plot"},{"panel_id":"d","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Etch rate for thermal ALE of HfO₂ using HF + Sn(acac)₂ increases from ~0.07 Å/cycle at 150 °C to a maximum of ~0.125 Å/cycle at 225 °C, then drops slightly at 250 °C (~0.115 Å/cycle). The temperature response is weakly non-monotonic with a clear peak near 225 °C.\"},{\"panel_id\":\"b\",\"text\":\"Single-temperature comparisons across precursor pairs show a wide spread in etch rate. HF + AlCl(CH₃)₂ at 250 °C is highest in this panel (~0.75 Å/cycle for HfO₂; ~0.95 Å/cycle for ZrO₂), while HF + Sn(acac)₂ at 200 °C is much lower (~0.05 for HfO₂; ~0.15 for ZrO₂). HF + Al(CH₃)₃ at 300 °C (HfO₂ point shown) is ~0.10 Å/cycle, and HF + SiCl₄ at 350 °C ranges from ~0.05 (HfO₂) to ~0.15 Å/cycle (ZrO₂).\"},{\"panel_id\":\"c\",\"text\":\"With HF + TiCl₄, etch rate increases with temperature from ~0.10 Å/cycle at 200 °C to ~0.58 Å/cycle at 300 °C for HfO₂. At 250 °C, ZrO₂ (~0.38 Å/cycle) is higher than HfO₂ (~0.29 Å/cycle), indicating a material-dependent etch response under the same precursor condition.\"},{\"panel_id\":\"d\",\"text\":\"With HF + AlCl(CH₃)₂, etch rate increases from 200 to 300 °C for both oxides, and ZrO₂ is consistently higher than HfO₂ at each temperature. Approximate values rise from ~0.35→~1.25 Å/cycle for HfO₂ and ~0.90→~1.60 Å/cycle for ZrO₂ over 200→300 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Precursor | Material | Temperature (°C) | Etch Rate (Å/cycle) |\\n|-----------|----------|------------------|---------------------|\\n| Hf / Sn(acac)2 (Red) | HfO₂ | 150 | ~0.07 |\\n| Hf / Sn(acac)2 (Red) | HfO₂ | 175 | ~0.085 |\\n| Hf / Sn(acac)2 (Red) | HfO₂ | 200 | ~0.115 |\\n| Hf / Sn(acac)2 (Red) | HfO₂ | 225 | ~0.125 (max) |\\n| Hf / Sn(acac)2 (Red) | HfO₂ | 250 | ~0.115 |\"},{\"panel_id\":\"b\",\"text\":\"| Precursor | Material | Temperature (°C) | Etch Rate (Å/cycle) |\\n|-----------|----------|------------------|---------------------|\\n| Hf / Sn(acac)2 (Red) | HfO₂ | 200 | ~0.05 |\\n| Hf / Sn(acac)2 (Red) | ZrO₂ | 200 | ~0.15 |\\n| Hf / AlCl(CH₃)₂ (Green) | HfO₂ | 250 | ~0.75 | \\n| Hf / AlCl(CH₃)₂ (Green) | ZrO₂ | 250 | ~0.95 |  \\n| Hf / Al(CH₃)₃ (Blue) | HfO₂ | 300 | ~0.10 | \\n| Hf/ SiCl₄ (Purple) | HfO₂ | 350 | ~0.05 |\\n| Hf / SiCl₄ (Purple) | ZrO₂ | 350 | ~0.15 |\"},{\"panel_id\":\"c\",\"text\":\"| Precursor | Material | Temperature (°C) | Etch Rate (Å/cycle) |\\n|-----------|----------|------------------|---------------------|\\n| Hf / TiCl₄ (Orange) | HfO₂ | 200 | ~0.10 |\\n| Hf / TiCl₄ (Orange) | HfO₂ | 225 | ~0.18 |\\n| Hf / TiCl₄ (Orange) | HfO₂ | 250 | ~0.29 |\\n| Hf / TiCl₄ (Orange) | ZrO₂ | 250 | ~0.38 |\\n| Hf / TiCl₄ (Orange) | HfO₂ | 275 | ~0.46 |\\n| Hf / TiCl₄ (Orange) | HfO₂ | 300 | ~0.58 |\"},{\"panel_id\":\"d\",\"text\":\"| Precursor | Material | Temperature (°C) | Etch Rate (Å/cycle) |\\n|-----------|----------|------------------|---------------------|\\n| Hf / AlCl(CH₃)₂ (Green) | HfO₂ | 200 | ~0.35 |\\n| Hf / AlCl(CH₃)₂ (Green) | ZrO₂ | 200 | ~0.90 | \\n| Hf / AlCl(CH₃)₂ (Green) | HfO₂ | 225 | ~0.80 | \\n| Hf / AlCl(CH₃)₂ (Green) | ZrO₂ | 225 | ~1.18 | \\n| Hf / AlCl(CH₃)₂ (Green) | HfO₂ | 250 | ~0.98 |\\n| Hf / AlCl(CH₃)₂ (Green) | ZrO₂ | 250 | ~1.35 | \\n| Hf / AlCl(CH₃)₂ (Green) | HfO₂ | 275 | ~1.10 |\\n| Hf / AlCl(CH₃)₂ (Green) | ZrO₂ | 275 | ~1.50 |\\n| Hf / AlCl(CH₃)₂ (Green) | HfO₂ | 300 | ~1.25 | \\n| Hf / AlCl(CH₃)₂ (Green) | ZrO₂ | 300 | ~1.60 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ZrO₂.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate of HfO₂ using HF + TiCl₄ increases steadily and monotonically as the temperature is raised from 200 °C to 300 °C. The rate roughly doubles from ~0.29 Å/cycle at 250 °C to ~0.58 Å/cycle at 300 °C, indicating a strong thermal dependence for this etch chemistry within the studied range.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The data show an etch-rate maximum near 225 °C rather than a steady increase with temperature. That suggests there’s an operating window where the surface chemistry yields the largest net removal per cycle under HF + Sn(acac)₂. Running below the peak likely underdrives one of the steps, while running above it appears to reduce net etching, consistent with a narrower optimum rather than a broad Arrhenius-like rise.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes; by roughly 0.20 Å/cycle (~0.95 vs ~0.75 Å/cycle).\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":359,"height":306},{"panel_id":"b","x":363,"y":3,"width":496,"height":304},{"panel_id":"c","x":0,"y":311,"width":356,"height":309},{"panel_id":"d","x":352,"y":305,"width":512,"height":315}],"source":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/25/images/figure_15.jpg","provenance":{"source_annotation":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/25/images/figure_15.json","source_content":"icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/25/content.json","source_pdf":["icdar2026-competition-data/test/gold_standard_test_set/atomic-layer-etching/simulation-usecase/25/Self-Limiting Temperature Window for Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"scatter plot","caption_source":"content.json:img_caption"},"width":867,"height":622,"image_format":"jpeg","image_sha256":"590a621dd1a1ccf79290c7e61618aec2532440dbec80dd8a9f20e212c6907ce5","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"}