{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_16_fig_2.jpg","caption":"FiG. 2. (a) Thermogravimetric analysis and (b) vapor pressures of the  $\\mathrm{Zr}$  precursors.","id":"train/atomic-layer-deposition/experimental-usecase/16/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/16/fig_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the thermogravimetric analysis curves (weight loss percentage as a function of temperature) for the Zr precursors. Zr(Cp)(ᵗBuDAD)(OⁱPr) evaporates with minimal residue, two other precursors show decomposition.\"},{\"panel_id\":\"b\",\"text\":\"The chart illustrates the vapor pressure of Zr complexes (Zr(Cp)(ᵗBuDAD)(OⁱPr),Zr(MeCp)(TMEA), Zr(Me₅Cp)(TEA))  at different temperatures. Zr(Cp)(ᵗBuDAD)(OⁱPr) achieves highest vapor pressures at low temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Weight loss (%) | Reagent + Conditions |\\n|---|---|---|\\n| 6 | 100.8 | Zr(Me_5Cp)(TEA) (vac) |\\n| 98 | 100.8 | Zr(Me_5Cp)(TEA) (vac) |\\n| 189 | 82.9 | Zr(Me_5Cp)(TEA) (vac) |\\n| 231 | 15.3 | Zr(Me_5Cp)(TEA) (vac) |\\n| 299 | 13.0 | Zr(Me_5Cp)(TEA) (vac) |\\n| 406 | 7.6 | Zr(Me_5Cp)(TEA) (vac) |\\n| 498 | 6.8 | Zr(Me_5Cp)(TEA) (vac) |\\n| 6 | 100.8 | Zr(Me_5Cp)(TEA) (atm) |\\n| 98 | 100.8 | Zr(Me_5Cp)(TEA) (atm) |\\n| 205 | 92.6 | Zr(Me_5Cp)(TEA) (atm) |\\n| 270 | 68.9 | Zr(Me_5Cp)(TEA) (atm) |\\n| 347 | 63.1 | Zr(Me_5Cp)(TEA) (atm) |\\n| 423 | 34.8 | Zr(Me_5Cp)(TEA) (atm) |\\n| 497 | 29.3 | Zr(Me_5Cp)(TEA) (atm) |\\n| 6 | 100.8 | Zr(MeCp)(TMEA) (atm) |\\n| 98 | 100.8 | Zr(MeCp)(TMEA) (atm) |\\n| 202 | 98.4 | Zr(MeCp)(TMEA) (atm) |\\n| 285 | 68.5 | Zr(MeCp)(TMEA) (atm) |\\n| 317 | 14.6 | Zr(MeCp)(TMEA) (atm) |\\n| 418 | 11.8 | Zr(MeCp)(TMEA) (atm) |\\n| 498 | 11.5 | Zr(MeCp)(TMEA) (atm) |\\n| 6 | 100.8 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\\n| 98 | 100.8 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\\n| 183 | 94.6 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\\n| 234 | 56.5 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\\n| 263 | 2.9 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\\n| 409 | 2.1 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\\n| 498 | 1.4 | Zr(Cp)(tBuDAD)(O^iPr) (atm) |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (°C) | Vapor pressure (Torr) | Reagent |\\n|---|---|---|\\n| 100 | 0.6 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 109 | 0.8 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 119 | 1.2 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 129 | 1.5 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 139 | 2.0 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 128 | 0.2 | Zr(Me_5Cp)(TEA) |\\n| 138 | 0.3 | Zr(Me_5Cp)(TEA) |\\n| 149 | 0.5 | Zr(Me_5Cp)(TEA) |\\n| 159 | 0.7 | Zr(Me_5Cp)(TEA) |\\n| 168 | 0.7 | Zr(Me_5Cp)(TEA) |\\n| 178 | 0.8 | Zr(Me_5Cp)(TEA) |\\n| 188 | 1.4 | Zr(Me_5Cp)(TEA) |\\n| 127 | 0.2 | Zr(MeCp)(TMEA) |\\n| 137 | 0.2 | Zr(MeCp)(TMEA) |\\n| 147 | 0.3 | Zr(MeCp)(TMEA) |\\n| 157 | 0.5 | Zr(MeCp)(TMEA) |\\n| 167 | 0.7 | Zr(MeCp)(TMEA) |\\n| 178 | 1.6 | Zr(MeCp)(TMEA) |\\n| 188 | 2.5 | Zr(MeCp)(TMEA) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(Cp)(ᵗBuDAD)(OⁱPr)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Applying vacuum conditions significantly reduces decomposition compared to atmospheric pressure. The thermogravimetric analysis shows a residue of ~30% at the highest temperature studied. Under vacuum, the residue is ~ 6%. The reduction in residue is a sign of reduced thermal decomposition.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(Me₅Cp)(TEA) under vacuum.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(Cp)(ᵗBuDAD)(OⁱPr) achieves highest vapor pressures at lower temperatures.\"}]}]","bbox":[{"panel_id":"a","x":40,"y":6,"width":611,"height":516},{"panel_id":"b","x":52,"y":550,"width":595,"height":489}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/Sanni Seppala et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":653,"height":1042,"image_format":"jpeg","image_sha256":"725b05417e27a28556e0e2bbb46cdd81a89ffaaf32e574799917f65d3251a011","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_16_fig_3.jpg","caption":"FiG. 3. (a) Film growth rates at different temperatures with (a) water and (b) ozone as the oxygen source. Pulse times for  $\\mathrm{Zr}$  precursors,  $\\mathrm{H}_2\\mathrm{O}$  and  $\\mathrm{O_3}$  were  $1.0\\mathrm{s}$  and purge times for all pulses  $1.5\\mathrm{s}$","id":"train/atomic-layer-deposition/experimental-usecase/16/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/16/fig_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the growth rates against temperatures for Zr precursors when using water as the oxygen source. The growth rate of films with Zr(Cp)(tBuDAD)(O^Pr), Zr(MeCp)(TMEA), and Zr(Me_5Cp)(TEA) compounds increases with temperature, with Zr(Cp)(tBuDAD)(O^Pr) showing the highest growth rate.\"},{\"panel_id\":\"b\",\"text\":\"The chart shows the growth rates against temperatures for Zr precursors when using ozone as the oxygen source. The growth rate of a film with Zr(MeCp)(TMEA) as precursor sligthtly decreases with temperatures. The growth rate with Zr(Cp)(tBuDAD)(O^Pr) and Zr(Me_5Cp)(TEA) compounds increases with temperature, with Zr(Cp)(tBuDAD)(O^Pr) it is showing the highest growth rate.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Growth rate (Å/cycle) | Reagent | \\n|---|---|---|\\n| 198 | 0,22 | Zr(MeCp)(TMEA) |\\n| 252 | 0,26 | Zr(MeCp)(TMEA) |\\n| 300 | 0,25 | Zr(MeCp)(TMEA) |\\n| 348 | 0,27 | Zr(MeCp)(TMEA) |\\n| 373 | 0,32 | Zr(MeCp)(TMEA) |\\n| 398 | 0,09 | Zr(Me_5Cp)(TEA) |\\n| 424 | 0,20 | Zr(Me_5Cp)(TEA) |\\n| 250 | 0,17 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 300 | 0,43 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 324 | 0,62 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 350 | 0,59 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 375 | 0,63 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 398 | 0,68 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 425 | 0,94 | Zr(Cp)(tBuDAD)(O^iPr) |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (°C) | Growth rate (Å/cycle) | Reagent | \\n|---|---|---|\\n| 250 | 0,09 | Zr(Me_5Cp)(TEA) |\\n| 275 | 0,19 | Zr(Me_5Cp)(TEA) |\\n| 300 | 0,29 | Zr(Me_5Cp)(TEA) |\\n| 350 | 0,38 | Zr(Me_5Cp)(TEA) |\\n| 374 | 0,36 | Zr(Me_5Cp)(TEA) |\\n| 400 | 0,38 | Zr(Me_5Cp)(TEA) |\\n| 426 | 0,39 | Zr(Me_5Cp)(TEA) |\\n| 252 | 0,39 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 300 | 0,66 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 377 | 1,28 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 198 | 0,64 | Zr(MeCp)(TMEA) |\\n| 250 | 0,57 | Zr(MeCp)(TMEA) |\\n| 300 | 0,57 | Zr(MeCp)(TMEA) |\\n| 352 | 0,52 | Zr(MeCp)(TMEA) |\\n| 375 | 0,52 | Zr(MeCp)(TMEA) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Zr(Me₅Cp)(TEA) displays the lowest growth rates among 3 precursors. It features bulky ligands that sterically shield the central zirconium atom, preventing efficient access for reactants and resulting in significantly slower growth rates. Additionally, the TEA ligand coordinates through three oxygen atoms to form chelating bonds that are  stronger than the nitrogen-based bonds in the other complexes, which reduces reactivity toward weak oxidants like water.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(Me₅Cp)(TEA) (did not react to water below 400 °C).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(MeCp)(TMEA) at approx. 250-350 °C.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(MeCp)(TMEA), Zr(Me₅Cp)(TEA).\"}]}]","bbox":[{"panel_id":"a","x":42,"y":17,"width":607,"height":553},{"panel_id":"b","x":47,"y":594,"width":602,"height":543}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/Sanni Seppala et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":653,"height":1144,"image_format":"jpeg","image_sha256":"7de551f5f77bff541302420c28602edd4f41881421dc6a50aa2d34b585565cd5","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_16_fig_4.jpg","caption":"FIG. 4. Film densities at different deposition temperatures with (a) water and (b) ozone as the oxygen source.","id":"train/atomic-layer-deposition/experimental-usecase/16/fig_4","sample_id":"atomic-layer-deposition/experimental-usecase/16/fig_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the density of films deposited with water and Zr(Cp)(ᵗBuDAD)(OⁱPr), Zr(MeCp)(TMEA), or Zr(Me₅Cp)(TEA) as a function of temperature.\"},{\"panel_id\":\"b\",\"text\":\"The chart shows the density of films deposited with ozone and Zr(Cp)(ᵗBuDAD)(OⁱPr), Zr(MeCp)(TMEA), or Zr(Me₅Cp)(TEA) as a function of temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Density (g/cm³) | Precursor |\\n|---|---|---|\\n| 398 | 4.7 | Zr (Me_5Cp) (TEA) |\\n| 423 | 4.7 | Zr (Me_5Cp) (TEA) |\\n| 200 | 3.8 | Zr (MeCp) (TMEA) |\\n| 249 | 4.4 | Zr (MeCp) (TMEA) |\\n| 300 | 5.2 | Zr (MeCp) (TMEA) |\\n| 325 | 5.6 | Zr (MeCp) (TMEA) |\\n| 347 | 5.8 | Zr (MeCp) (TMEA) |\\n| 375 | 5.8 | Zr (MeCp) (TMEA) |\\n| 248 | 4.7 | Zr (Cp) (tBuDAD) (O^iPr) |\\n| 299 | 5.1 | Zr (Cp) (tBuDAD) (O^iPr) |\\n| 324 | 5.1 | Zr (Cp) (tBuDAD) (O^iPr) |\\n| 347 | 5.2 | Zr (Cp) (tBuDAD) (O^iPr) |\\n| 374 | 5.7 | Zr (Cp) (tBuDAD) (O^iPr) |\\n| 397 | 5.7 | Zr (Cp) (tBuDAD) (O^iPr) |\\n| 424 | 5.8 | Zr (Cp) (tBuDAD) (O^iPr) |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (°C) | Density (g/cm³) | Precursor |\\n|---|---|---|\\n| 249 | 5,4 | Zr(Me_5Cp)(TEA) |\\n| 275 | 5,6 | Zr(Me_5Cp)(TEA) |\\n| 301 | 5,8 | Zr(Me_5Cp)(TEA) |\\n| 350 | 5,9 | Zr(Me_5Cp)(TEA) |\\n| 375 | 5,9 | Zr(Me_5Cp)(TEA) |\\n| 399 | 5,9 | Zr(Me_5Cp)(TEA) |\\n| 426 | 5,9 | Zr(Me_5Cp)(TEA) |\\n| 201 | 3,8 | Zr(MeCp)(TMEA) |\\n| 250 | 4,5 | Zr(MeCp)(TMEA) |\\n| 302 | 5,6 | Zr(MeCp)(TMEA) |\\n| 350 | 5,9 | Zr(MeCp)(TMEA) |\\n| 375 | 5,9 | Zr(MeCp)(TMEA) |\\n| 375 | 5,9 | Zr(MeCp)(TMEA) |\\n| 250 | 5,1 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 298 | 5,7 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 375 | 5,9 | Zr(Cp)(tBuDAD)(O^iPr) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(MeCp)(TMEA) (due to higher density).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"300°C would lead to the higher amount of deposited material. Although the growth rates are equal, the film density increased from ~4.5 g/cm3 to ~5.5 g/cm3, as the temperature rises from 250°C to 300°C. The increase in density means that more atoms are packed into the film at the higher temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After ~350 °C at ~ 5.8 g/cm3.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~350 °C and above.\"}]}]","bbox":[{"panel_id":"a","x":46,"y":17,"width":600,"height":533},{"panel_id":"b","x":54,"y":585,"width":591,"height":531}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/Sanni Seppala et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":650,"height":1114,"image_format":"jpeg","image_sha256":"07f6745093c357182f7bf461c439a466cb80f83e050f4f9f706ef0d51d34a70b","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_16_fig_9.jpg","caption":"FIG. 9. Leakage current density curves of the  $\\mathrm{ZrO_2}$  films deposited with  $\\mathrm{O_3}$  at  $300^{\\circ}\\mathrm{C}$ . The polarity of the electric field indicates the potential applied to the top electrode.","id":"train/atomic-layer-deposition/experimental-usecase/16/fig_9","sample_id":"atomic-layer-deposition/experimental-usecase/16/fig_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the leakage current density versus electric field for films deposited at 300°C using ozone and 3 different Zr precursors (Zr(Cp)(ᵗBuDAD)(OⁱPr), Zr(MeCp)(TMEA)  and Zr(Me₅Cp)(TEA))\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Electric field (MV/cm) | Current density (A/cm²) | Reagent |\\n|---|---|---|\\n| -2,5 | 0,000034413049869757 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -2,4 | 0,000043492871235381 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -2,3 | 0,000021544346900319 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -2,3 | 0,000029439237258586 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -2,3 | 0,000017957144943716 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -2,2 | 0,000022695105366947 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -2,2 | 0,000013141473626118 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,0 | 0,000000030610872343 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,1 | 0,000000033095885951 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,1 | 0,000000016394163633 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,2 | 0,000000026186598892 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,3 | 0,000000004581597669 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,4 | 0,000000012971623961 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,4 | 0,000000001153843460 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,4 | 0,000000006768750009 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,5 | 0,000000015163203393 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,6 | 0,000000005355666918 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,6 | 0,000000017269832907 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,7 | 0,000000001384340542 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,8 | 0,000000015562895291 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,8 | 0,000000001660882783 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,9 | 0,000000001916398735 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 0,9 | 0,000000010534134765 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,0 | 0,000000026876860122 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,1 | 0,000000008120906678 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,1 | 0,000000021265892581 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,3 | 0,000000029058743813 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,4 | 0,000000044062364278 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,5 | 0,000000060208944933 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 1,7 | 0,000000086666869037 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| 2,5 | 0,000097431763878509 | Zr(Cp)(tBuDAD)(O^iPr) |\\n| -1,8 | 0,000014582829828531 | Zr(Me_5Cp)(TEA) |\\n| -0,8 | 0,000000068573796153 | Zr(Me_5Cp)(TEA) |\\n| -0,6 | 0,000000134878740803 | Zr(Me_5Cp)(TEA) |\\n| -0,5 | 0,000000005355666918 | Zr(Me_5Cp)(TEA) |\\n| -0,2 | 0,000000007414066842 | Zr(Me_5Cp)(TEA) |\\n| -0,1 | 0,000000000901160733 | Zr(Me_5Cp)(TEA) |\\n| 0,0 | 0,000000048895122296 | Zr(Me_5Cp)(TEA) |\\n| 0,1 | 0,000000000501839650 | Zr(Me_5Cp)(TEA) |\\n| 0,4 | 0,000000037693909754 | Zr(Me_5Cp)(TEA) |\\n| 0,5 | 0,000000004581597669 | Zr(Me_5Cp)(TEA) |\\n| 0,6 | 0,000000031417754461 | Zr(Me_5Cp)(TEA) |\\n| 0,6 | 0,000000008554672536 | Zr(Me_5Cp)(TEA) |\\n| 0,8 | 0,000000084441059201 | Zr(Me_5Cp)(TEA) |\\n| 0,9 | 0,000000010000000000 | Zr(Me_5Cp)(TEA) |\\n| 1,2 | 0,000000040753929659 | Zr(Me_5Cp)(TEA) |\\n| 1,3 | 0,000000012638482029 | Zr(Me_5Cp)(TEA) |\\n| 1,4 | 0,000000066812659151 | Zr(Me_5Cp)(TEA) |\\n| 1,5 | 0,000000004463931423 | Zr(Me_5Cp)(TEA) |\\n| 1,9 | 0,000097431763878509 | Zr(Me_5Cp)(TEA) |\\n| -1,4 | 0,000001772505414982 | Zr(MeCp)(TMEA) |\\n| -1,0 | 0,000000248588037617 | Zr(MeCp)(TMEA) |\\n| -0,4 | 0,000000007414066842 | Zr(MeCp)(TMEA) |\\n| -0,1 | 0,000000001916398735 | Zr(MeCp)(TMEA) |\\n| 0,0 | 0,000000041828175983 | Zr(MeCp)(TMEA) |\\n| 0,1 | 0,000000000397072725 | Zr(MeCp)(TMEA) |\\n| 1,3 | 0,000002182644728397 | Zr(MeCp)(TMEA) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Zr(MeCp)(TMEA)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, it is different. The graph shows that when the electric field is negative, the leakage current is low and similar to the other compounds. When electric field is positive, the leakage current spikes compared to other precursors.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The curve shoots vertically upward at approx. 1.5 - 1.7 MV/cm.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":7,"width":645,"height":493}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/16/Sanni Seppala et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":648,"height":500,"image_format":"jpeg","image_sha256":"49bd449f26ba57082aa457905da2331b6b06c2da0a6d6ffce4143d0976ca59e9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_25_figure_10.jpg","caption":"Figure 10.  $JV$  sweeps for the best solar cells using the same RBF PDT ACIGS absorber and either TGO 0.077, ZGO 0.32, ZTO 0.2, or CdS as the ESL. Scans of solar cells using a pretreatment of either one cycle of ALD  $\\mathrm{Al}_2\\mathrm{O}_3$ $\\mathrm{(TMA + H_2O)}$  or three cycles of ALD ZnS (DEZ  $+\\mathrm{H}_2\\mathrm{S})$  prior to the deposition of TGO 0.077 are also shown.","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_10","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The PV characteristics is illustrated in the chart where the current density (in mA/cm²) is plotted as a function of bias voltage (in V) for various materials, including TGO 0.077, TGO 0.077 TMA, TGO 0.077 ZnS, ZGO 0.32, ZTO 0.20, and CdS.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Bias (V) | TGO 0.077 | TGO 0.077 TMA | TGO 0.077 ZnS | ZGO 0.32 | ZTO 0.20 | CdS |\\n|----------|-----------|---------------|---------------|----------|----------|-----|\\n| -0.5     | -35       | -35           | -35           | -35      | -35      | -35 |\\n| 0.0      | 0         | 0             | 0             | 0        | 0        | 0   |\\n| 0.3      | 0         | 0             | 0             | 0        | 0        | 0   |\\n| 0.5      | 10        | 12            | 8             | 6        | 5        | 4   |\\n| 0.6      | 20        | 22            | 18            | 15       | 12       | 10  |\\n| 0.7      | 30        | 32            | 28            | 25       | 22       | 18  |\\n| 0.8      | 38        | 40            | 36            | 32       | 30       | 25  |\\n| 0.9      | 40        | 42            | 40            | 36       | 34       | 28  |\\n| 1.0      | 40        | 42            | 40            | 36       | 34       | 28  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ZGO 0.32 with a Voc of 704 ± 8 mV. The solar cells with ZGO 0.32 ESLs had almost the same Voc as those with ZTO 0.2 but, on average, worse FF.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The TGO 0.077 ESL clearly gave lower Voc and FF compared to the ZnO or Cd-containing alternatives, just like in a previous study on TGAO ESL.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Aside from the blue losses in EQE for the devices with 2.4 eV Eg CdS, all of the solar cells showed good collection with some variation in the interference peak positions due to the difference in their ZnO:Al/ZnO/ESL stack thicknesses\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"One thing that clearly distinguishes the TGO process from the more successful ALD ESL is that it does not contain Zn or its very reactive precursor DEZ. In the first test, the initial ALD conditions of Zn-containing ESLs were therefore mimicked by using three ALD cycles of ZnS (DEZ and H2S) before\\ncontinuing on with a TGO 0.077 ESL. The resulting JV data for these solar cells show that Voc improved to values comparable to using ZGO 0.32 ESL.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":500,"height":346}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/Adam Hultqvist et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":502,"height":350,"image_format":"jpeg","image_sha256":"4e515f157adc0fd344305ab5834ce31130826a60f25caac3c8c1674bf15c53a3","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_25_figure_2.jpg","caption":"Figure 2. (a) Cation ratio measured by XRF/RBS for TGO and ZGO films in Table 1 as a function of the pulse ratio of the supercycle process. (b) GPC measured by XRR for the TGO and ZGO films (Table S2).","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between pulse ratio and cation ratio measured from RBF/XPS for TGO and ZGO.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates the growth per cycle against pulse ratio for TGO and ZGO.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse Ratio Ge/(Ge + Sn or Zn) | Cation Ratio (TGO) | Cation Ratio (ZGO) |\\n|--------------------------------|---------------------|---------------------|\\n| 0.0                            | 0.00               | 0.00               |\\n| 0.1                            | 0.08               | 0.02               |\\n| 0.2                            | 0.15               | 0.03               |\\n| 0.4                            | 0.30               | 0.05               |\\n| 0.6                            | 0.45               | 0.10               |\\n| 0.8                            | 0.70               | 0.40               |\\n| 1.0                            | 1.00               | 1.00               |\"},{\"panel_id\":\"b\",\"text\":\"| Pulse Ratio Ge/(Ge + Sn or Zn) | Growth per Cycle (TGO) [Å/cycle] | Growth per Cycle (ZGO) [Å/cycle] |\\n|--------------------------------|-----------------------------------|-----------------------------------|\\n| 0.0                            | 1.10                              | 2.00                              |\\n| 0.1                            | 1.05                              | 1.80                              |\\n| 0.2                            | 1.00                              | 1.60                              |\\n| 0.4                            | 0.95                              | 1.40                              |\\n| 0.6                            | 0.90                              | 1.20                              |\\n| 0.8                            | 0.85                              | 1.00                              |\\n| 1.0                            | 0.80                              | 0.80                              |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The GPC for these films decreased for an increasing Ge pulse ratio of the ALD ZGO process, even to such an extent that it became substantially lower than  for the binary GeOy and ZnO processes. The QCM data in the SI also showed that the MPC for each GeOy or ZnO subcycle decreased in the ternary ZGO process. All of these trends suggest that the GeOy and ZnO subcycles were not fully compatible with each other due to unfavorable surface conditions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The measured RBS/XRF [Ge]/([Ge] + [Sn]) cation ratio, x-value, for TGO films on fused silica is just a bit lower than the ALD Ge/(Ge + Sn) pulse ratio. The measured growth per cycle (GPC) for these films decreases as the pulse ratio increases (b) . This trend appears almost as a linear combination of the lower GPC GeOy subcycle and the larger GPC SnOy subcycle. Both the trend in the cation ratio and the trend in GPC are also supported by QCM data suggest that the SnOy and GeOy subcycles affect each other only to a minor extent.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Varying the Ge/(Ge + Zn) pulse ratio of the ALD process varied the XRF/RBS [Ge]/([Ge] +[Zn]) cation ratio, the x-value, measured for films on fused\\nsilica, but the film cation ratio severely lags behind the ALD pulse ratio\"}]}]","bbox":[{"panel_id":"b","x":2,"y":366,"width":504,"height":352},{"panel_id":"a","x":0,"y":0,"width":507,"height":364}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/Adam Hultqvist et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":508,"height":719,"image_format":"jpeg","image_sha256":"80089a83fb11fcacebd3bb0acf1491c25a316602751d003b45aae81f7f990c50","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_25_figure_8.jpg","caption":"Figure 8. Schematic illustration of the roughly estimated  $E_{c}$  and  $E_{v}$  trends as a function of the cation ratio,  $x_{i}$  of a TGO and b ZGO. The estimations are based on the optical absorption and the XPS  $E_{v}$  spectra from this study and from the  $\\mathrm{SnO}_{y} / \\mathrm{CIGS}$ , and  $\\mathrm{ZnO} / \\mathrm{CIGS}$  band offset and ACIGS band position data from previous studies. The cation ratio where the ACIGS solar cells (see the next section) start showing FF losses is also indicated.","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_8","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the energy difference between conduction bands and valence bands of ACIGS and TGO materials as a function of TGO cation ratio.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates the energy difference between conduction bands and valence bands of ACIGS and ZGO materials as a function of ZGO cation ratio.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TGO cation ratio (x) | E_c (TGO) [eV] | E_v (TGO) [eV] | E_c (ACIGS) [eV] | E_v (ACIGS) [eV] |\\n|-----------------------|-----------------|-----------------|-------------------|-------------------|\\n| 0.00                   | +0.2            | -3.0            | 0.0               | -1.0              |\\n| 0.10                   | +0.3            | -3.0            | 0.0               | -1.0              |\\n| 0.20 (FF loss) | +0.4            | -3.0            | 0.0               | -1.0              |\\n| 0.30                   | +0.6            | -3.0            | 0.0               | -1.0              |\"},{\"panel_id\":\"b\",\"text\":\"| ZGO cation ratio (x) | E_c (ZGO) [eV] | E_v (ZGO) [eV] | E_c (ACIGS) [eV] | E_v (ACIGS) [eV] |\\n|-----------------------|-----------------|-----------------|-------------------|-------------------|\\n| 0.00                 | +0.2            | -3.5            | 0.0               | -1.0              |\\n| 0.10                 | +0.3            | -3.6            | 0.0               | -1.0              |\\n| 0.20                 | +0.4            | -3.7            | 0.0               | -1.0              |\\n| 0.30 (FF loss) | +0.5            | -3.8            | 0.0               | -1.0              |\\n| 0.40                 | +0.6            | -4.0            | 0.0               | -1.0              |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The estimations are based on the optical absorption and the XPS Ev spectra from this study and from the SnOy/CIGS, and ZnO/CIGS band offset and ACIGS band position data from previous studies\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"While the low-cation-ratio ZGO films likely have direct Eg, there was a transition toward X-ray amorphous films for higher cation ratios, which are likely\\nto give them indirect Eg. For simplicity, all of the Eg values of ZGO were therefore estimated using an indirect Eg model.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. For ZGO, FF losses occur at a cation ratio of 0.32\\n2. For TGO, FF losses occur at a cation ratio of 0.16\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The TGO sample set also showed a noteworthy correlation between the large tailing observed in the optical absorption and XPS Ev spectra, the high resistivity, and the amorphous nature of these films. Compared to the previous study using a high-temperature ALD process, the films in this\\nstudy are similar in terms of being amorphous, showing tails in the optical absorption and an increasing Ec position with increasing Ge content.\"}]}]","bbox":[{"panel_id":"b","x":0,"y":363,"width":506,"height":330},{"panel_id":"a","x":1,"y":1,"width":504,"height":355}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/Adam Hultqvist et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":508,"height":714,"image_format":"jpeg","image_sha256":"cf95aa1c61c3f0e226c6100b7767c5dda4f7fbfacbb9ee726b240d15dfc93eaf","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_25_figure_9.jpg","caption":"Figure 9.  $JV$  sweeps of representative solar cells using (a) TGO and (b) ZGO ESLs.","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_9","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the JV characteristics of devices using TGO as ESL with varying doping values.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the JV characteristics of devices using ZGO as ESL with varying doping values.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Bias (V) | CdS | x = 0 | x = 0.065 | x = 0.077 | x = 0.10 | x = 0.16 | x = 0.30 |\\n|----------|-----|-------|-----------|-----------|----------|----------|----------|\\n| -0.5     | -35  | -35    | -35       | -35        | -35       | -35       | -30      |\\n| 0.0      | 0   | 0     | 0         | 0         | 0        | 0        | -3       |\\n| 0.3      |-35  | -35    | -35       | -35        | -35       | -35       | 0   |\\n| 0.5      | -35  | -10    | -15        | -18        | -20       | -18       |0       |\\n| 0.7      | 0  | 20    | 25        | 28        | 30       | 20       | 0      |\"},{\"panel_id\":\"b\",\"text\":\"| Bias (V) | CdS | x = 0 | x = 0.032 | x = 0.12 | x = 0.19 | x = 0.26 | x = 0.32 | x = 0.37 |\\n|----------|-----|-------|-----------|----------|----------|----------|----------|----------|\\n| -0.4     | -35   | -35      | -40         | -35        | -35        | -35         | -35         | -35        |\\n| 0.0      |  -35   | -35      | -35         | -35        | -35        | -35         | -35         | -35      |\\n| 0.2      | -35   | -30     | -28         | -32        | -34        | -35        | -35        | -35        |\\n| 0.4      | 0   | 20    | 30        | 35       | -34       | -35       | 25       | -10       |\\n| 0.6      | 0   | -    | -        | -       | 0       | 0      | 0      | 0       |\\n| 0.8      | -   | -    | -        | -      |20       | 20       | 40       | 20       |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the Ge content was increased for the ZGO films, the solar cells showed a decrease in Voc, compared to solar cells using ZnO . When the Ge content increased further, there was an increase in Voc up to ZGO 0.37.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"From figure a it appears, the device characteristics of CdS is better than device b as the Voc seems higher in case a than case b.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. ZnO:Al/ZnO/TGO/ACIGS/Mo/SLG solar cell stacks\\n, 2. ZnO:Al/ZnO/ZGO/ACIGS/Mo/SLG solar cell stacks\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Until TGO 0.10, the Voc of the solar cells increased with increasing Ge content. At the same time, the short circuit current density (Jsc) remained unchanged. For TGO 0.16, both the Voc and the FF started to drop off due to a kink in the J−V curve around Voc. Finally, the solar cells with TGO 0.30 show low FF, low Voc and a tiny Jsc, suggesting that the transport of photoexcited electrons has been almost completely hampered.\"}]}]","bbox":[{"panel_id":"b","x":0,"y":353,"width":497,"height":353},{"panel_id":"a","x":0,"y":4,"width":499,"height":357}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/Adam Hultqvist et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":503,"height":708,"image_format":"jpeg","image_sha256":"de9a22f837f59bce27fff78c273e949785ae4e67f835ad7ad75d1fc95f50aed4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_26_figure_1.jpg","caption":"Figure 1. ALD growth rates for  $\\mathrm{In}_2\\mathrm{O}_3$ $(\\Delta)$ $\\mathrm{SnO_2}$ $(\\Omega)$  and ITO  $(\\bullet)$  versus deposition temperature. Potential ALD window for ITO growth is indicated.","id":"train/atomic-layer-deposition/experimental-usecase/26/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/26/figure_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multiple line chart compares the growth rate per cycle of Tin Oxide, Indium Oxide, and ITO films as a function of deposition temperature. Both Tin and Indium Oxide show a general increase in growth rate with temperature, while ITO exhibits a stable region between ~200 - 300 °C, identified as the “ITO Window,” where optimal co-deposition occurs.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition Temperature (°C) | Tin Oxide (Å/Cycle) | Indium Oxide (Å/Cycle) | ITO (Å/Cycle) |\\n|-----------------------------|--------------------|-------------------------|---------------|\\n| 100                         | 0.5                | 1.2                     | 1.3           |\\n| 150                         | 0.8                | 1.3                     | 1.2           |\\n| 200                         | 1.5                | 1.4                     | 1.3           |\\n| 250                         | 2.0                | 1.5                     | 1.4           |\\n| 300                         | 2.8                | 1.6                     | 1.5           |\\n| 350                         | 2.9                | 2.5                     | 1.6           |\\n| 400                         | 3.0                | 2.7                     | 1.6           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Tin Oxide shows a steep increase in growth rate with temperature, Indium Oxide increases more moderately, and ITO remains stable between 200–300 °C, forming a defined “ITO Window.”\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Growth rate (Å/Cycle).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Tin Oxide, Indium Oxide, ITO (Indium Tin Oxide)\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":597,"height":583}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/Jeffrey W. Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":597,"height":583,"image_format":"jpeg","image_sha256":"5b979595bd0059ef4b778f56f18fae8d2d3cc54d339d8565296a00a1e17c2017","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_26_figure_3.jpg","caption":"Figure 3.  $\\mathrm{SnO_2}$  content versus percentage of  $\\mathrm{SnO_2}$  cycles for ALD ITO films determined by XRF. Dashed line shows expected  $\\mathrm{SnO_2}$  content as calculated using a rule-of-mixtures formula.","id":"train/atomic-layer-deposition/experimental-usecase/26/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/26/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the increase in SnO₂ content (mol %) as a function of SnO₂ cycle percentage, as measured by XRF. A dashed line indicates the expected values from the Rule of Mixtures. The measured data rises faster than predicted, suggesting non-linear incorporation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| % SnO₂ Cycles | SnO₂ Content (mol %) |\\n|---------------|----------------------|\\n| 0             | 0                    |\\n| 5             | 2                    |\\n| 10            | 12                   |\\n| 15            | 26                   |\\n| 25            | 40                   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The measured SnO₂ content increases faster than the Rule of Mixtures prediction, indicating a non-linear trend in incorporation.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Enhanced SnO₂ uptake, Possible non-ideal mixing behavior, Deviations from theoretical assumptions\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":572,"height":570}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/Jeffrey W. Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":572,"height":570,"image_format":"jpeg","image_sha256":"62decce0f40a7ba517118b3cdd5a7bfc0108d6a4ac540d5e0780c6677e11c62b","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_30_figure_8.jpg","caption":"Figure 8. Representative magnetization versus external field curves of doped iron oxide films deposited from  $\\mathrm{CpFeC_5H_4CHN(CH_3)_2}$  at  $375^{\\circ}C$  Mg/Fe atomic ratios, film thicknesses, hysteresis parameters and measurement temperatures are given by labels. Annealing was carried out in forming gas at  $450^{\\circ}C$  for  $30\\mathrm{min}$  (Color online)","id":"train/atomic-layer-deposition/experimental-usecase/30/figure_8","sample_id":"atomic-layer-deposition/experimental-usecase/30/figure_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows hysteresis loops of magnetic material Mg/Fe = 0.017 at various temperatures (10, 50, 100, 200, 300 K) with as deposited condition , the thickness of the film ~425 nm\"},{\"panel_id\":\"b\",\"text\":\"The figure shows hysteresis loops of magnetic material Mg/Fe = 0.061 at various temperatures (10, 50, 100, 200, 300 K) with as deposited condition , the thickness of the film ~411 nm\"},{\"panel_id\":\"c\",\"text\":\"The figure shows hysteresis loops of magnetic material Mg/Fe = 0.017 at various temperatures after annealing at 450◦C\"},{\"panel_id\":\"d\",\"text\":\"The figure shows hysteresis loops of magnetic material Mg/Fe = 0.061 at various temperatures after annealing at 450◦C.  The hysteresis loops shows noticable magnetic behavior with high magnetic saturations and low coercive fields.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | Saturation Magnetization Ms (emu/cm³) | Coercivity Hc (Oe) |\\n|-----------------|---------------------------------------|-------------------|\\n| 300 | ~200 | 300 |\\n| 200 | ~200 | ~300 |\\n| 100 | ~200 | ~300 |\\n| 50 | ~200 | ~300 |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (K) | Saturation Magnetization Ms (emu/cm³) | Coercivity Hc (Oe) |\\n|-----------------|---------------------------------------|-------------------|\\n| 300 | 491 | 150 |\\n| 200 | ~500 | ~150 |\\n| 100 | ~500 | ~150 |\\n| 50 | ~500 | ~150 |\\n| 10 | ~500 | ~150 |\"},{\"panel_id\":\"c\",\"text\":\"| Temperature (K) | Saturation Magnetization Ms (emu/cm³) | Coercivity Hc (Oe) |\\n|-----------------|---------------------------------------|-------------------|\\n| 300 | 280 | 280 |\\n| 200 | ~290 | ~280 |\\n| 100 | ~290 | ~280 |\\n| 50 | ~290 | ~280 |\\n| 10 | ~290 | ~280 |\"},{\"panel_id\":\"d\",\"text\":\"| Temperature (K) | Saturation Magnetization Ms (emu/cm³) | Coercivity Hc (Oe) |\\n|-----------------|---------------------------------------|-------------------|\\n| 300 | 274 | 200 |\\n| 200 | ~280 | ~200 |\\n| 100 | ~280 | ~200 |\\n| 50 | ~280 | ~200 |\\n| 10 | ~280 | ~200 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the Mg content increases, the Hc value tends to decrease.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the as-deposited state, increasing Mg/Fe from 0.017 to 0.061 increases Ms from about 200 to about 491 emu/cm³. Over the same change, Hc decreases from about 300 to about 150 Oe. This matches the broader trend described in the paper that higher Mg in this range reduces coercivity while increasing saturation magnetization up to a limit.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The best room-temperature magnetization in this figure is the as-deposited Mg/Fe = 0.061 film (panel b). It shows Ms ≈ 491 emu/cm³ at 300 K, substantially higher than the Mg/Fe = 0.017 film and higher than either annealed case. This directly supports the process-optimization message that moderate Mg levels can strongly enhance magnetic performance (while other processing steps can change Ms and Hc in different directions).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Mg incorporation and FGA alter the cation distribution in the spinel lattice. At moderate Mg levels, magnetite formation is favored, increasing Mₛ and lowering Hc. At higher Mg levels, Mg²⁺ substitutes Fe³⁺ in octahedral sites, reducing magnetic exchange interactions and introducing lattice strain, which increases coercivity and decreases magnetization. Annealing further modifies grain boundaries and defect density, influencing domain wall motion.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. In the nondoped and as-deposited films the HC ranged from 485 to 1950 Oe between 300 and 10 K.\\n2. In the films with Mg/Fe atomic ratio of 0.017 the HC was in the range of 300–1260 Oe between 300 and 10 K. \\n3. With  increase in the Mg content to Mg/Fe ratio of 0.062 the HC decreased toward the range of 150–460 Oe between 300 and 10 K.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 10 K.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing temperature reduces both the coercive field and the saturation magnetisation. The hysteresis loops become narrower at higher temperatures, indicating weaker magnetic ordering.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The coercive field decreases as the Mg/Fe ratio increases.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Mg/Fe=0.017 and the film as-deposited\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The hysteresis parameter Hc seems to decrease for increasing measurement temperature. This is seen in all subfigures, however the trend is less big for the highly doped films.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Hc is 1500 Oe.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Saturation magnetization increases from ~200 emu cm⁻³ at low Mg content to ~491 emu cm⁻³ at higher Mg content.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"x-axis: magnetic field (Oe)\\n\\ny-axis: magnetization, M (emu/cm³)\\n\\ncurves: measurements at 300, 200, 100, 50, and 10 K for the same sample/state\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In each panel, Hc increases as temperature decreases (loops widen at low T), which is typical for ferri/ferromagnetic films. After forming-gas annealing, Hc is generally reduced at higher temperatures compared with the as-deposited state for these kinds of samples. The paper attributes the anneal-driven reduction in coercivity to crystal growth and increased average magnetic domain size.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"FGA modifies both saturation magnetization (Mₛ) and coercivity (Hc). For low Mg content (Mg/Fe = 0.017), Mₛ increases from 200 → 280 emu/cm³, and Hc slightly decreases from 300 → 280 Oe, indicating improved magnetite crystallinity and reduced pinning. For higher Mg content (Mg/Fe = 0.061), Mₛ decreases from 491 → 274 emu/cm³, while Hc increases from 150 → 200 Oe, suggesting that excess Mg and annealing lead to cation redistribution and reduced magnetic Fe content.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As-deposited: Increasing Mg/Fe from 0.017 to 0.061 significantly raises Mₛ (200 → 491 emu/cm³) and lowers Hc (300 → 150 Oe), indicating that moderate Mg promotes magnetite formation and soft magnetic behavior.\\nAfter FGA: Low Mg films gain magnetization (Mₛ = 280 emu/cm³), while high Mg films lose magnetization (Mₛ = 274 emu/cm³) and become harder (Hc rises to 200 Oe).\\nMg doping initially enhances magnetization, but excessive Mg combined with annealing reduces Fe magnetic sites and increases anisotropy.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, only up to a certain limit.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The coercive field (Hc) decreases systematically with increasing Mg doping, indicating that Mg addition softens the magnetic behaviour. For instance, Hc drops from roughly 485-1950 Oe in undoped films to 150-460 Oe at a Mg/Fe ratio of 0.062. Annealing further reduces Hc, consistent with structural relaxation and grain coarsening. Meanwhile, the saturation magnetization (μs) increases with moderate Mg incorporation, suggesting improved ferromagnetic ordering and a higher proportion of reduced iron oxide phases such as magnetite. However, excessive Mg content diminishes μs again, likely due to the formation of a nonmagnetic ternary phase like magnesioferrite.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 820 and 895 K.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Curie temperatures remain between 820 and 895 K, close to that of bulk magnetite, confirming that the dominant phase retains strong ferrimagnetic character.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Annealing of all films was carried out at 450 °C for 30 minutes.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Magnetic field (Oe).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The loop becomes wider as the temperature drops.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 300 K.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The annealing seems to have the effect to create films with roughly the same saturation magnetization, even if the as-deposited films deviate from this level.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Coercivity decreases slightly from ~300 Oe to ~280 Oe after annealing.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Annealing likely reduces structural disorder and improves cation distribution within the iron-oxide lattice. This suppresses antiferromagnetic coupling and enhances net magnetic alignment. The increase in saturation magnetization therefore reflects improved magnetic ordering rather than thickness effects.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The as-deposited film with Mg/Fe = 0.061 offers the best performance for sensors and spintronics due to its high Mₛ (491 emu/cm³) and low Hc (150 Oe), enabling strong magnetic response and easy switching. FGA-treated low-Mg films (Mg/Fe = 0.017) are suitable for applications requiring moderate coercivity and thermal stability, while high-Mg annealed films may be useful for data storage where higher Hc provides stability against demagnetization.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The coercivity and magnetic moment decreased with increase in magnesium content. Therefore, high amounts of magnesium may not be advantageous in terms of increasing both HC and μS, maybe due to the formation of an additive ternary phase, e.g. magnesioferrite, in the film bulk.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The as-deposited sample ,The Mg = 0.001 doped sample ,The Mg = 0.017 doped sample.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The main affected properties are the coercive field, Saturation magnetization, Loop shape of the hysteresis curves, Temperature dependence of magnetization\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"280 emu/cm^3.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Hc value is higher for the annealed samples than for the as-deposited one.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The annealed film\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The as-deposited films show that with increasing dopant level the saturated magnetization increases significantly, whilst the Hc parameter decreases.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Measuring hysteresis down to 10 K reduces thermal activation of domain walls, making intrinsic magnetic anisotropy more apparent. This helps separate temperature-driven loop broadening from composition-dependent effects. The comparison clarifies how Mg content and annealing influence intrinsic magnetic ordering.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The annealing procedure also decreased the HC.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Adjusting the Mg content allows fine control over the balance between coercivity and saturation magnetization, which is crucial for practical magnetic applications. Moderate Mg doping can soften the magnetic response by lowering coercivity while maintaining or slightly improving magnetization, making the films more suitable for applications such as magnetic sensors or spintronic components. However, excessive Mg incorporation can introduce secondary phases like magnesioferrite, which limits further performance gains. This figure shows that careful compositional tuning, combined with annealing and temperature control, is essential for optimizing magnetic behavior.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":4,"width":499,"height":384},{"panel_id":"b","x":500,"y":2,"width":510,"height":393},{"panel_id":"c","x":8,"y":398,"width":502,"height":391},{"panel_id":"d","x":500,"y":394,"width":511,"height":395}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/Kaupo Kukli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"30","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1011,"height":791,"image_format":"jpeg","image_sha256":"dee718efc7dc61b42f321ddf933a34d2dddbc5d278becb3dfc59faaf204bb344","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_30_figure_9.jpg","caption":"Figure 9. Magnetization versus measurement temperature curves of selected samples after forming gas annealing measured under FGD field cooling (ZFC) and field cooling (FC) regimes. The composition, growth temperatures and film thicknesses are described by labels. The peaks associated with the Verwey transition of magnetite are denoted by  $\\mathrm{T_V}$","id":"train/atomic-layer-deposition/experimental-usecase/30/figure_9","sample_id":"atomic-layer-deposition/experimental-usecase/30/figure_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The plot shows the temperature dependence of magnetic susceptibility for Fe₂O₃ nanoparticles grown at 450 °C with an average diameter of 705 nm. Measurements were carried out using a field gradient apparatus (FGA) under an applied magnetic field of 1000 Oe, following zero-field-cooled (ZFC) and field-cooled (FC) protocols after forming gas annealing. Features associated with the Verwey transition temperature (T𝑉) of magnetite are indicated.\"},{\"panel_id\":\"b\",\"text\":\"This plot presents the temperature dependence of magnetic moment per gram for a composite consisting of 100 × Fe₂O₃ and 4 × (3 × MgO + 100 × Fe₂O₃), grown at 425 °C with an average diameter of 183 nm. Measurements were performed using an FGA under a magnetic field of 1000 Oe in both ZFC and FC modes after forming gas annealing. The Verwey transition temperature (TV) is marked.\"},{\"panel_id\":\"c\",\"text\":\"The plot shows the temperature dependence of magnetic moment per gram for a composite composed of 100 × Fe₂O₃ and 15 × (10 × MgO + 100 × Fe₂O₃), grown at 375 °C with an average diameter of 411 nm. Magnetization measurements were conducted using an FGA with an applied magnetic field of 1000 Oe under ZFC and FC conditions following forming gas annealing, with the Verwey transition temperature (T𝑉) indicated.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | FC (emu/g) | ZFC (emu/g) |\\n| --------------- | ---------- | ----------- |\\n| 10              | ~0.10      | ~0.01       |\\n| 50              | ~0.10      | ~0.03       |\\n| 100             | ~0.10      | ~0.07       |\\n| 125             | ~0.10      | ~0.10       |\\n| 200             | ~0.10      | ~0.09       |\\n| 300             | ~0.10      | ~0.08       |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (K) | FC (emu/g) | ZFC (emu/g) |\\n| --------------- | ---------- | ----------- |\\n| 10              | ~0.055     | ~0.030      |\\n| 50              | ~0.055     | ~0.040      |\\n| 100             | ~0.055     | ~0.050      |\\n| 120             | ~0.055     | ~0.055      |\\n| 200             | ~0.055     | ~0.050      |\\n| 300             | ~0.055     | ~0.045      |\"},{\"panel_id\":\"c\",\"text\":\"| Temperature (K) | FC (emu/g) | ZFC (emu/g) |\\n| --------------- | ---------- | ----------- |\\n| 10              | ~0.069     | ~0.068      |\\n| 50              | ~0.069     | ~0.069      |\\n| 100             | ~0.069     | ~0.070      |\\n| 150             | ~0.068     | ~0.069      |\\n| 200             | ~0.067     | ~0.068      |\\n| 300             | ~0.066     | ~0.067      |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Growth temperature and composition.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For all three graphs, samples with different thickness are used. So it could be that this thickness plays a factor in the different magnetizations.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It suggests increased magnetic disorder and weakened long-range ordering.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In (a) a pronounced anomaly is marked at ~125 K. In (b) a similar anomaly remains visible around ~120 K. In (c) the anomaly is much less prominent, consistent with the stated trend that the Verwey signature weakens in films that are more disordered and/or contain higher MgO amounts.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A Verwey-type transition near ~120 K is a characteristic low-temperature transformation associated with magnetite (Fe₃O₄) (cubic-to-monoclinic symmetry change tied to cation ordering). Observing a step/jump in M–T around 120–125 K is therefore used as supporting evidence for the presence of magnetite in mixed-phase samples. In this paper, that interpretation is explicitly noted for selected annealed films showing the jump.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Verwey transition depends on stoichiometric magnetite with ordered Fe²⁺/Fe³⁺ distribution. Mg substitution and oxygen vacancies disrupt this order, lowering or smearing Tᵥ. Grain size and strain also matter: larger, well-crystallized grains show sharp Tᵥ and strong ZFC/FC splitting, while smaller or defect-rich grains broaden the transition and reduce irreversibility.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The transition lost its prominence in as-deposited films, in relatively disordered films grown at the lowest temperatures examined and in films containing relatively higher amounts of magnesium oxide\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Verwey transition visible near 120–125 K for composition variations, Lower growth temperatures reduce the transition magnitude, Thicker films show minor decreases in magnetization but overall similar trends\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"120 K.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Either the growth temperature could be lowered, or the Mg/Fe ratio could be increased.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The dopant level is Mg/Fe = 0.062.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It implies well-preserved magnetite-like ordering.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Cryogenic sensors: Prefer films with a sharp Verwey transition and strong ZFC/FC separation (e.g., thick, low-Mg film), as these provide distinct low-temperature magnetic signatures. 2. Room-temperature devices: Favor films with suppressed or broadened Tᵥ (e.g., high-Mg film), ensuring stable magnetic behavior across 150–300 K without abrupt changes.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The temperature dependence of magnetic moment was measured at a magnetic moment of 1000 Oe\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The undoped Fe₂O₃ sample shows a clear ZFC peak near 125 K, indicating strong blocking behavior. The Mg-containing sample grown at 425 °C shows a similar but weaker peak near 120 K. In contrast, the sample grown at 375 °C exhibits minimal FC–ZFC separation, suggesting suppressed or absent blocking.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1000 Oe.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Fe₂O₃ sample grown at 450 °C, The Fe₂O₃–MgO composite grown at 425 °C\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 120–125 K.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"450 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The temperature is 125 K.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It seems that increasing the doping level, slightly increases the magnetic moment.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The separation reflects irreversible magnetic behavior caused by domain pinning or blocked magnetic moments. A strong divergence indicates a magnetically ordered phase with thermal history dependence. The convergence of FC and ZFC curves marks the temperature where magnetic domains become thermally unpinned.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It rises sharply and reaches a peak near 125 K.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ZFC: zero-field cooling measurement curve\\n\\nFC: field-cooling measurement curve\\n\\nApplied field during M–T: 1000 Oe\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Magnetite-like content is linked to the stronger ferrimagnetic response desired in these films, while hematite-dominant films tend to show much weaker magnetization. A Verwey-type feature provides an additional, temperature-dependent indicator that reduced/spinel-like iron oxide is present after forming-gas anneal. This helps connect processing (growth temperature, Mg dosing, anneal) to the observed magnetic behavior.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Verwey transition (Tᵥ ≈ 120–125 K) is a hallmark of magnetite (Fe₃O₄) with long-range charge ordering. Its appearance after FGA shows that annealing in a reducing atmosphere successfully converted hematite or maghemite phases into magnetite and improved crystallinity. This process also reduces oxygen content and stabilizes Fe²⁺/Fe³⁺ mixed valence states, essential for magnetite formation.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Pure Fe₂O₃ film (450 °C, thick): Shows a sharp Tᵥ at ~125 K and strong ZFC/FC separation, indicating well-ordered magnetite after FGA.\\nLow Mg doping (Mg/Fe ≈ 0.021, 425 °C): Tᵥ shifts slightly lower (~120 K) and ZFC/FC separation narrows, suggesting smaller grains and mild disorder. High Mg doping (Mg/Fe ≈ 0.062, 375 °C): Tᵥ is broadened or suppressed, and ZFC/FC curves nearly overlap, indicating disrupted charge ordering and increased structural disorder due to Mg substitution and lower growth temperature\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A jump in magnetization-versus-temperature curves is  known as the Verwey transition. At the Verwey transition temperature, TV, ca. 120 K, magnetite is supposed to transform from cubic (T > TV) to monoclinic (T < TV) symmetry. This transformation is related to changes in cation ordering and, at a slightly higher temperature (ca. 130 K) to changes in magnetocrystalline anisotropy as well as switching the easy and hard magnetization axes. The Verwey transition can be used for identification of magnetite in pure or mixed phase samples.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1000 Oe.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At the Verwey transition temperature, TV, ca. 120 K, magnetite is supposed to transform from cubic (T > T_V) to monoclinic (T < T_V) symmetry.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the Mg fraction in the film increases, T_V decreases. As the Mg fraction increases, the difference between ZFC and FC also becomes more pronounced, but the transition point becomes less clear. Finally, an increase in Mg content also causes an overall decrease of magnetization, which is consistent with earlier observations in this work.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the undoped iron oxide film, it is 125 K and a magnetization of 0.11 emu/g; The iron oxide film with Mg/Fe = 0.021, T_V = 120 K, M = 0.055; The iron oxide film with Mg/Fe = 0.062, the T_V is not stated.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 125 K, as indicated by the label Tv.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 120 K.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 125 K.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Composition strongly affects the Verwey transition, with magnetite-containing films showing a peak near 120–125 K. Growth temperature also influences magnetic behavior, as lower temperatures suppress the Verwey transition, leading to smaller differences between FC and ZFC magnetization. Film thickness has a subtler effect, slightly modifying magnetization magnitudes but not significantly altering the overall transition features. Together, these factors determine the magnetic phase evolution and the shape of the multi-line curves.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":512,"height":429},{"panel_id":"b","x":5,"y":434,"width":522,"height":415},{"panel_id":"c","x":5,"y":855,"width":522,"height":418}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/images/figure_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/30/Kaupo Kukli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"30","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":528,"height":1280,"image_format":"jpeg","image_sha256":"15c457cf92189c9aa5b6e1aa40acaa232e48cdaf1bb9d043beb39c87e84da1a9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_31_figure_1.jpg","caption":"Figure 1. XPS depth profile of thermal ALD Co on Si(001) substrate from  $\\mathrm{CoCp(CO)_2}$  and  $\\mathrm{H}_2$  showing the Co, C, O, and Si contents in the film.","id":"train/atomic-layer-deposition/experimental-usecase/31/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/31/figure_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Figure shows an XPS depth profile of a thermal-ALD Co film deposited on Si(001) using CoCp(CO)₂ and H₂. The profile reveals a carbon-rich film with ~40–50 at% C and only ~45–50 at% Co throughout most of the depth, along with a few at% oxygen contamination. The sharp rise in Si near ~25–30 min sputtering marks the Co/Si interface, confirming the film thickness. Overall, the data demonstrate that thermal ALD produces a Co–C composite rather than pure metallic Co, indicating poor ligand removal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Element|Trend across film|Interpretation|\\n|--------|-----------------|--------------|\\n|Co|~45–50 at% throughout|Indicates incomplete metal formation|\\n|C|~40–50 at% throughout|Severe carbon contamination from ligands|\\n|O|Low, few at%|Minor oxidation/impurity|\\n|Si|Rises sharply at end|Marks Co/Si interface (film thickness)|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Carbon\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Interconnect applications require low-resistivity, high-purity metallic cobalt. The XPS depth profile reveals that the thermal-ALD film contains nearly equal amounts of cobalt and carbon, forming a carbon-rich Co–C composite rather than metallic Co. Such films exhibit very high electrical resistivity and poor reliability, making them unsuitable for device applications and motivating the need for plasma-enhanced ALD.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"High carbon content (~40–50 at%), Relatively low cobalt content (~45–50 at%), Presence of oxygen impurities, Formation of a Co–C composite structure\"}]}]","bbox":[{"panel_id":"a","x":3,"y":4,"width":540,"height":409}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/31/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/31/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/31/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/31/Han-Bo-Ram Lee et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"31","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":544,"height":414,"image_format":"jpeg","image_sha256":"7866bf6abb3f8b279110bcacd69f0ef2524b6c5c33563b7f27c33e8dc514da89","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_32_figure_5.jpg","caption":"Figure 5. Leakage current densities vs electric fields of thermal and PEALD (a)  $\\mathrm{Ta}_2\\mathrm{O}_5$  and (b)  $\\mathrm{TiO_2}$  thin film. The film thickness was  $10\\mathrm{nm}$  for all films, and the growth temperatures were  $250^{\\circ}\\mathrm{C}$  for  $\\mathrm{Ta}_2\\mathrm{O}_5$  and  $200^{\\circ}\\mathrm{C}$  for  $\\mathrm{TiO_2}$  respectively.","id":"train/atomic-layer-deposition/experimental-usecase/32/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/32/figure_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Leakage current density versus electric field for 10 nm Ta₂O₅ films grown at 250°C. PE-ALD yields approximately two orders of magnitude lower leakage (~8×10⁻⁷ A/cm²) compared to thermal ALD (~6×10⁻⁵ A/cm²) at 1 MV/cm.\"},{\"panel_id\":\"b\",\"text\":\"Leakage current density versus electric field for 10 nm TiO₂ films grown at 200°C. PE-ALD again shows lower leakage (~7×10⁻⁴ A/cm²) than thermal ALD (~5×10⁻³ A/cm²) at 1 MV/cm, though the difference is smaller than for Ta₂O₅.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Electric Field, MV/cm | Thermal ALD (A/cm²) | PE-ALD (A/cm²) |\\n|---|---|---|\\n| 0 | 1×10⁻⁷ | 5×10⁻⁸ |\\n| 0.1 | 5×10⁻⁷ | 7×10⁻⁸ |\\n| 0.25 | 3×10⁻⁶ | 1×10⁻⁷ |\\n| 0.5 | 2×10⁻⁵ | 3×10⁻⁷ |\\n| 0.75 | 5×10⁻⁵ | 7×10⁻⁷ |\\n| 1.0 | 7×10⁻⁵ | 2×10⁻⁶ |\\n| 1.5 | 1×10⁻⁴ | 7×10⁻⁶ |\\n| 2.0 | 1.5×10⁻⁴ | 2×10⁻⁵ |\"},{\"panel_id\":\"b\",\"text\":\"| Electric Field, MV/cm | Thermal ALD (A/cm²) | PE-ALD (A/cm²) |\\n|---|---|---|\\n| 0 | 2×10⁻³ | 5×10⁻⁵ |\\n| 0.1 | 3×10⁻³ | 8×10⁻⁵ |\\n| 0.25 | 3.5×10⁻³ | 2×10⁻⁴ |\\n| 0.5 | 4×10⁻³ | 4×10⁻⁴ |\\n| 0.75 | 5×10⁻³ | 5×10⁻⁴ |\\n| 1.0 | 6×10⁻³ | 6×10⁻⁴ |\\n| 1.5 | 8×10⁻³ | 8×10⁻⁴ |\\n| 2.0 | 1×10⁻² | 1×10⁻³ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"PE-ALD Ta₂O₅ exhibits a leakage current density of approximately 8×10⁻⁷ A/cm² at 1 MV/cm, which is about two orders of magnitude lower than thermal ALD Ta₂O₅ at approximately 6×10⁻⁵ A/cm². This substantial improvement makes PE-ALD Ta₂O₅ from PDMAT competitive with or better than previously reported values for PE-ALD Ta₂O₅ from ethoxide precursors.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Lower oxygen vacancy concentration, reduced carbon or silicon contamination, and smoother oxide/substrate interfaces observed by TEM.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"PE-ALD Ta₂O₅, with the lowest leakage current density of approximately 8×10⁻⁷ A/cm² at 1 MV/cm.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The improvement is greater for Ta₂O₅. PE-ALD reduces the leakage current by approximately two orders of magnitude for Ta₂O₅ (from ~6×10⁻⁵ to ~8×10⁻⁷ A/cm²), whereas for TiO₂ the reduction is less than one order of magnitude (from ~5×10⁻³ to ~7×10⁻⁴ A/cm²).\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":672,"height":435},{"panel_id":"b","x":1,"y":436,"width":670,"height":511}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/W. J. Maeng et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"32","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":678,"height":953,"image_format":"jpeg","image_sha256":"a96d4db318206463e79cf7b97d58885b42b7cae1e73313bdcb06fc5c9cf3097d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_34_figure_1.jpg","caption":"Figure 1. Film thickness as function of number of cycles measured by in situ spectroscopic ellipsometry for ALD of Li-containing films using LiHMDS combined various coreactants at a table temperature of  $200^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/34/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/34/figure_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure displays a multiple -line chart showing the film thickness growth as a function of ALD cycle number for different oxidant/reactant chemistries. The growth rates vary significantly depending on the species used: H₂O exhibits the highest growth rate, followed by H₂* + O₂*, then O₂* + H₂*, and finally O₂* alone, which shows the slowest growth. The chart highlights how the choice of oxidant or co-reactant can strongly influence the cumulative thickness achieved over the same number of ALD cycles..\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle | O₂* (nm) | H₂O (nm) | H₂* + O₂* (nm) | O₂* + H₂* (nm) |\\n|-------|----------|----------|----------------|----------------|\\n| 0     | 0        | 0        | 0              | 0              |\\n| 50    | 10       | 30       | 20             | 15             |\\n| 100   | 20       | 60       | 40             | 30             |\\n| 150   | 30       | 90       | 60             | 45             |\\n| 200   | 40       | 120      | 80             | 60             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The faster and steeper thickness increase observed for H₂O compared to O₂* implies that H₂O is significantly more reactive or efficient as an oxidant in this ALD process. This could be due to more favorable surface reactions, faster ligand removal, or a more complete reaction per cycle. In contrast, O₂* likely leads to slower surface reactions, incomplete ligand removal, or passivation effects that limit the growth per cycle. This highlights the importance of selecting appropriate co-reactants for achieving desired film thickness and deposition rates.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"60 nm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"O₂*, O₂* + H₂*, H₂* + O₂*, H₂O\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":570,"height":428}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/M. J. Pieters et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":570,"height":428,"image_format":"jpeg","image_sha256":"ced65fee56d9edc144257e3ba2f39e53a3aea9ec59cb0ea477100947b978045f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_34_figure_3.jpg","caption":"Figure 3. (a) Time-resolved QMS data of selected  $m / z$  values during the LiHMDS  $+\\mathrm{H}_2\\mathrm{O}$  process. The standard ALD recipe (SR) at the start and end of the deposition, corresponding to the slow and fast growth regimes (earlier defined in Figure 1) respectively, are compared to the recipe with only  $\\mathrm{H}_2\\mathrm{O}$  and the recipe with only LiHMDS. Zooms of the HMDS-related signals of  $m / z = 15$  amu (b) and 146 amu (c) during the LiHMDS dose step.","id":"train/atomic-layer-deposition/experimental-usecase/34/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/34/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents ion current traces for various species monitored during sequential pulsing of LiHMDS and H₂O in an ALD process. Subfigure (a) shows the evolution of ion signals for multiple mass fragments (e.g., CH₃⁺, H₂O⁺, NH(SiMe₃)₂⁺) across time and marks key events such as precursor pulses and surface reaction (SR) start and end times. Subfigures (b) and (c) zoom in on the time-resolved ion currents for specific m/z values (15 and 146), further detailing the temporal behavior of key reaction byproducts during the LiHMDS pulse. Together, the data provide insight into precursor decomposition and reaction dynamics on the surface.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | LiHMDS | SR start | H₂O | SR end |\\n|----------|--------|----------|-----|--------|\\n| 0        | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 5        | 1E-11  | 1E-11    | 1E-11| 1E-11  |\\n| 10       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 15       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 20       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 25       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 30       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 75       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\\n| 80       | 1E-12  | 1E-12    | 1E-12| 1E-12  |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Ion current (A) |\\n|----------|-----------------|\\n| 0        | 1E-11           |\\n| 2        | 1E-10           |\\n| 4        | 1E-10           |\\n| 6        | 1E-10           |\\n| 8        | 1E-10           |\"},{\"panel_id\":\"c\",\"text\":\"| Time (s) | Ion current (A) |\\n|----------|-----------------|\\n| 0        | 1E-11           |\\n| 2        | 1E-10           |\\n| 4        | 1E-10           |\\n| 6        | 1E-10           |\\n| 8        | 1E-10           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ion current profiles reveal the presence and intensity of specific species over time, corresponding to precursor introduction and reaction phases. In subfigure (a), CH₃⁺ and NH(SiMe₃)₂⁺ signals spike during the LiHMDS pulse, suggesting the release of ligand fragments due to surface reactions. The H₂O⁺ signal increases during the water pulse, indicating oxidant introduction. The timing of signal changes at SR start and SR end further supports surface-limited reactions consistent with ALD. Subfigures (b) and (c) confirm this by showing that key mass fragments peak during the precursor pulse, then decay rapidly, demonstrating self-limiting behavior and supporting the controlled nature of ALD chemistry.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approx. 5 seconds.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"CH₃⁺ (m/z = 15), H₂O⁺ (m/z = 18), NH(SiMe₃)₂⁺ (m/z = 146)\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":595,"height":545},{"panel_id":"b","x":656,"y":3,"width":477,"height":260},{"panel_id":"c","x":661,"y":294,"width":472,"height":253}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/M. J. Pieters et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1136,"height":547,"image_format":"jpeg","image_sha256":"b20ed1b0bec210bf177c39e0b53b230e53c8de218d48f162b045e714118452af","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_34_figure_5.jpg","caption":"Figure 5. Time-resolved QMS data of selected  $m / z$  values during the LiHMDS  $+\\mathrm{O}_2^*$  process. The standard ALD recipe (SR, red) is compared to the recipe with only  $\\mathrm{O}_2^*$  (blue) and the recipe with LiHMDS and  $\\mathrm{O}_2$  gas (black).","id":"train/atomic-layer-deposition/experimental-usecase/34/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/34/figure_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents time-resolved mass spectrometry data for various reaction by-products during exposure to LiHMDS, O₂ plasma (O₂*), and the surface reaction (SR) step. The top four subplots show ion current intensity for species including H₂O⁺ (m/z=18), SiH₃⁺ (m/z=30), CO₂⁺ (m/z=44), and HNSi₂Me₃⁺ (m/z=146), with clear changes occurring after precursor and plasma exposure. CO₂ and H₂O peaks intensify during the SR period, consistent with ligand combustion, while HNSi₂Me₃⁺ appears following LiHMDS exposure, reflecting gas-phase fragmentation of unreacted precursor. The bottom panel tracks chamber pressure, showing correlation between dose timing and ion evolution. These measurements collectively map the dynamic release of volatile by-products during an ALD half-cycle..\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | LiHMDS (A) | O₂* (A) | SR (A) |\\n|----------|------------|---------|--------|\\n| 0        | 1.5E-12    | 1.5E-12 | 1.5E-12 |\\n| 5        | 1.6E-12    | 1.5E-12 | 1.5E-12 |\\n| 10       | 1.8E-12    | 1.5E-12 | 2.0E-12 |\\n| 15       | 2.2E-12    | 2.5E-12 | 3.0E-12 |\\n| 20       | 1.8E-12    | 3.5E-12 | 2.8E-12 |\\n| 25       | 1.5E-12    | 2.2E-12 | 1.7E-12 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 7 seconds, during the LiHMDS exposure phase.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"H₂O⁺ (m/z = 18), CO₂⁺ (m/z = 44), HNSi₂Me₃⁺ (m/z = 146)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The sharp rise in CO₂⁺ and H₂O⁺ signals during the SR step indicates that combustion of organic ligands is occurring due to interaction with reactive oxygen species. These by-products are signatures of oxidative ligand removal, suggesting that the SR step is essential for fully eliminating unreacted LiHMDS residues. Understanding the timing and completeness of these signals helps in optimising precursor dosing and plasma timing to ensure efficient surface cleaning and minimal carbon incorporation in the final film , a key control parameter in ALD processes targeting high-purity oxides or nitrides\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":672,"height":905}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/M. J. Pieters et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":672,"height":905,"image_format":"jpeg","image_sha256":"cc7428746c384840e9889026dacc4e1732ab9082892714ad15e7d8e9ac45c815","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_34_figure_6.jpg","caption":"Figure 6. Time-resolved QMS data of selected  $m / z$  values during the LiHMDS  $+\\mathrm{O}_2^*$ $+\\mathrm{H}_2^*$  process. The standard ALD recipe (SR) at the start and end of the deposition, corresponding to the slow and fast growth regimes respectively, are compared to the recipes with only LiHMDS, only  $\\mathrm{O}_2^*$ $+\\mathrm{H}_2^*$  only LiHMDS  $+\\mathrm{H}_2^*$  and only LiHMDS  $+\\mathrm{O}_2^*$ .","id":"train/atomic-layer-deposition/experimental-usecase/34/figure_6","sample_id":"atomic-layer-deposition/experimental-usecase/34/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents ion current versus time data across several subplots, each tracking a different species (e.g., H₂O⁺, SiH₃⁺, CO₂⁺, HNSi₂Me₃⁺) detected by mass spectrometry at their respective m/z values. Different reactant conditions, such as LiHMDS alone, and combinations with O₂* and H₂*,  are compared to observe how plasma and precursor combinations influence gaseous by-product formation. For example, the H₂O⁺ signal peaks during exposure to O₂*, indicating oxidative ligand removal, while HNSi₂Me₃⁺ signals provide evidence of volatile organosilicon fragments. The pressure trace at the bottom confirms timing of reagent pulses and surface reactions (SR), validating the synchrony of signal changes with process steps\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | LiHMDS (m/z=18) | O₂* (m/z=30) | H₂* (m/z=44) | H₂O (m/z=146) |\\n|----------|------------------|--------------|--------------|----------------|\\n| 0        | 1E-10            | 1E-11        | 1E-12        | 1E-12          |\\n| 10       | 1.2E-10          | 1.1E-11      | 1.3E-12      | 1.2E-12        |\\n| 20       | 1.5E-10          | 1.4E-11      | 1.6E-12      | 1.8E-12        |\\n| 30       | 1.3E-10          | 1.3E-11      | 1.4E-12      | 1.5E-12        |\\n| 40       | 1.1E-10          | 1.2E-11      | 1.2E-12      | 1.3E-12        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 20 seconds.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SiH₃⁺ (m/z = 30), CO₂⁺ (m/z = 44), and HNSi₂Me₃⁺ (m/z = 146)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ion current trends suggest that ligand combustion and surface reactions are being actively driven during the LiHMDS + O₂* + H₂* exposure. The sharp rise in H₂O⁺ indicates oxidative ligand breakdown, possibly of methyl or HMDS groups. Concurrent increases in CO₂⁺ and HNSi₂Me₃⁺ imply that both organic and silicon-containing fragments are released as volatile by-products. The sequential timing and magnitude of these signals can guide optimization of ALD cycle parameters, such as plasma duration and purge steps,  to ensure complete removal of precursor residues and maximise film purity. This is particularly important when designing ALD processes for sensitive electronics, where residual ligands can impair dielectric or conductive properties.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":666,"height":862}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/M. J. Pieters et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":666,"height":862,"image_format":"jpeg","image_sha256":"1519884a28622b3d034d2793afb139b84a0714f5d65df32101039ad5c9224ddc","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_34_figure_7.jpg","caption":"Figure 7. Time-resolved QMS data of selected  $m / z$  values during the  $\\mathrm{LiHMDS} + \\mathrm{H}_2^* +\\mathrm{O}_2^*$  process. The standard ALD recipe (SR) at the start and end of the deposition, corresponding to the slow and fast growth regimes respectively, are compared to the recipes with only LiHMDS, only  $\\mathrm{O}_2^* +\\mathrm{H}_2^*$ , only  $\\mathrm{LiHMDS} + \\mathrm{H}_2^*$  and only  $\\mathrm{LiHMDS} + \\mathrm{O}_2^*$ .","id":"train/atomic-layer-deposition/experimental-usecase/34/figure_7","sample_id":"atomic-layer-deposition/experimental-usecase/34/figure_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents time-resolved quadrupole mass spectrometry (QMS) data tracking the evolution of several ion fragments (e.g., CH₃⁺, H₂O⁺, SiH₂⁺, CO₂⁺, HNSi₂Me₅⁺) under sequential exposure to LiHMDS, hydrogen plasma (H₂*), oxygen plasma (O₂*), and combined steps in an ALD process. Notably, the CH₃⁺ and HNSi₂Me₅⁺ signals are elevated during LiHMDS dosing, consistent with precursor fragmentation. H₂O⁺ and CO₂⁺ peaks emerge during SR phases involving oxidants, suggesting combustion of ligands. The pressure trace shows dosing windows for each reactant and aligns with corresponding ion signal surges. Together, these data elucidate the chemical pathways of ligand decomposition and by-product release across plasma-activated ALD cycles\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | LiHMDS (A) | LiHMDS+H₂* (A) | H₂*+O₂* (A) | LiHMDS+O₂* (A) | SR end (A) |\\n|----------|-------------|----------------|-------------|----------------|------------|\\n| 0        | 9.5E-11     | 9.5E-11        | 9.5E-11     | 9.5E-11        | 9.5E-11    |\\n| 5        | 1.1E-10     | 1.1E-10        | 1.1E-10     | 1.1E-10        | 1.1E-10    |\\n| 10       | 1.0E-10     | 1.0E-10        | 1.0E-10     | 1.0E-10        | 1.0E-10    |\\n| 15       | 9.0E-11     | 9.0E-11        | 9.2E-11     | 9.0E-11        | 9.0E-11    |\\n| 20       | 1.2E-10     | 1.1E-10        | 1.3E-10     | 1.2E-10        | 1.2E-10    |\\n| 25       | 8.5E-11     | 9.0E-11        | 1.1E-10     | 9.5E-11        | 9.0E-11    |\\n| 30       | 6.5E-11     | 7.5E-11        | 9.0E-11     | 7.0E-11        | 7.0E-11    |\\n| 35       | 5.0E-11     | 6.0E-11        | 8.0E-11     | 6.0E-11        | 6.0E-11    |\\n| 40       | 3.5E-11     | 4.0E-11        | 6.5E-11     | 4.5E-11        | 4.5E-11    |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Pressure (mTorr) |\\n|----------|------------------|\\n| 0        | 0                |\\n| 5        | 50               |\\n| 10       | 120              |\\n| 15       | 50               |\\n| 20       | 0                |\\n| 25       | 50               |\\n| 30       | 100              |\\n| 35       | 50               |\\n| 40       | 0                |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The rise in CO₂⁺ during the SR period following O₂* exposure suggests effective combustion of organic ligands from the LiHMDS precursor, implying that oxidative plasma plays a key role in removing residual carbonaceous species. Meanwhile, HNSi₂Me₅⁺ signals peaking during the LiHMDS step indicate gas-phase fragmentation rather than surface retention. The differing behaviors underscore the sequential decomposition of ligands and highlight the importance of synchronizing plasma exposure to ensure full ligand removal.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 120 mTorr.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"H₂O⁺ (m/z=18), CO₂⁺ (m/z=44), and CH₃⁺ (m/z=15)\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":666,"height":796},{"panel_id":"b","x":4,"y":796,"width":664,"height":190}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/34/M. J. Pieters et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":670,"height":991,"image_format":"jpeg","image_sha256":"3cd0e5563920b1ba0f26f131c8f90aa436129da192d53423c3bfc246612efbb9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_36_fig_12.jpg","caption":"Fig. 12 a Normalized values of  $S_{ID} / I_{ID}^{2}$  plotted at various values of  $(V_{GS} - V_{th})$ ; b Hooge's coefficient  $(a_{\\parallel})$  plotted with respect to  $(V_{GS} - V_{th})$  for the conventional HEMT and the  $\\mathrm{Ga}_{2}\\mathrm{O}_{3}$  MOS-HEMT","id":"train/atomic-layer-deposition/experimental-usecase/36/fig_12","sample_id":"atomic-layer-deposition/experimental-usecase/36/fig_12","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Multiple line chart displaying the relationship between frequency and S<sub>id</sub>/I<sub>D</sub> for Ni/Au and MOS at V_GS=0.4.\"},{\"panel_id\":\"b\",\"text\":\"Multiple line chart displaying the relationship between frequency and S<sub>id</sub>/I<sub>D</sub> for Ni/Au and MOS at V_GS=0.6.\"},{\"panel_id\":\"c\",\"text\":\"Multiple line chart displaying the relationship between frequency and S<sub>id</sub>/I<sub>D</sub> for Ni/Au and MOS at V_GS=0.8.\"},{\"panel_id\":\"d\",\"text\":\"Multiple line chart displaying the relationship between frequency and S<sub>id</sub>/I<sub>D</sub> for Ni/Au and MOS at V_GS=01.\"},{\"panel_id\":\"e\",\"text\":\"Scatter plot showing the relationship between V<sub>GS</sub>-V<sub>th</sub> and Hooge's Parameter (α<sub>H</sub>) for Ni/Au-HEMT and Ga<sub>2</sub>O<sub>3</sub> MOS-HEMT.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Frequency (Hz) | Ni/Au | MOS |\\n| --- | --- | --- |\\n| 10^2 | 10^-10 | 9*10^-10 |\\n| 10^4 | 10^-12 | 10^-11 |\"},{\"panel_id\":\"b\",\"text\":\"| Frequency (Hz) | Ni/Au | MOS |\\n| --- | --- | --- |\\n| 10^2 | 10^-10 | 9*10^-10 |\\n| 10^4 | 10^-12 | 10^-11 |\"},{\"panel_id\":\"c\",\"text\":\"| Frequency (Hz) | Ni/Au | MOS |\\n| --- | --- | --- |\\n| 10^2 | 9*10^-10 | 9*10^-11 |\\n| 10^4 | 10^-13 | 9^-13 |\"},{\"panel_id\":\"d\",\"text\":\"| Frequency (Hz) | Ni/Au | MOS |\\n| --- | --- | --- |\\n| 10^2 | 10^-11 | 10^-11 |\\n| 10^4 | 4-10^-13 | 9*10^-12 |\"},{\"panel_id\":\"e\",\"text\":\"| V<sub>GS</sub>-V<sub>th</sub> (V) | Ni/Au-HEMT  | Ga2O3 MOS-HEMT  |\\n| --- | --- | --- |\\n| 0.2 | 8*10^-1 | 7*10^-2 |\\n| 0.4 | 10^-2 | 9*10^-2 |\\n| 0.6 | 9*10^-1 | 3*10^-3 |\\n| 0.8 | 4*10^-2 | 10^-3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the frequency increases, the S_iD/I_d ratio decreases for both the Ni/Au and the MOS curves.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Ni/Au sample.\"}]},{\"panel_id\":\"e\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"100 Hz.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The S_iD/I_d ratio decreases when the difference increases from 0.4 to 1.\"}]}]","bbox":[{"panel_id":"a","x":7,"y":31,"width":409,"height":567},{"panel_id":"b","x":410,"y":28,"width":299,"height":571},{"panel_id":"c","x":708,"y":35,"width":302,"height":558},{"panel_id":"d","x":1006,"y":37,"width":304,"height":557},{"panel_id":"e","x":312,"y":637,"width":737,"height":478}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/Huan-Yu Shih et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1317,"height":1119,"image_format":"jpeg","image_sha256":"65e3f21e9cb75c06f4fe1b3b889c72b0bbc95117f8c02c188ea641338eaefee5","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_36_fig_3.jpg","caption":"Fig. 3 a Growth rate of the  $\\mathrm{Ga}_{2}\\mathrm{O}_{3}$  thin films plotted with respect to the TEG pulse time and plasma time. b Film thickness plotted with respect to the number of applied ALD cycles","id":"train/atomic-layer-deposition/experimental-usecase/36/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/36/fig_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the growth rate vs plasma time and the growth rate vs TEG pulse time. The growth rate of a material increases with increasing TEG pulse time and plasma time up to a certain point and then plateaus.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the thickness as a function of the number of ALD cycles. The thickness of a material increases linearly with the number of ALD cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TEG pulse time | Growth rate | Plasma time | Growth rate |\\n| --- | --- | --- | --- |\\n| 0.00 | 0.015 | 0 | 0.016 |\\n| 0.05 | 0.06 | 5 | 0.06 |\\n| 0.10 | 0.06 | 10 | 0.06 |\\n| 0.20 | 0.06 | 20 | 0.06 |\"},{\"panel_id\":\"b\",\"text\":\"| ALD cycle (number) | Thickness (nm) |\\n| --- | --- |\\n| 0 | 0 |\\n| 100 | 7 |\\n| 200 | 13 |\\n| 300 | 17 |\\n| 400 | 25 |\\n| 600 | 35 |\\n| 700 | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TEG and O2 plasma.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After 5 s of plasma exposure.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The film thickness increases with the number of ALD cycles, in accordance with the ALD process.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":17,"y":22,"width":642,"height":533},{"panel_id":"b","x":669,"y":80,"width":649,"height":472}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/Huan-Yu Shih et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1319,"height":556,"image_format":"jpeg","image_sha256":"ae4675d5002450c8cc5998bc7497f26f2fe800b17762f8f80f8e6577f6ad7aa9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_38_figure_2.jpg","caption":"Figure 2. Elemental depth profiles of the samples with increasing number of  $\\mathrm{LiO}^{\\prime}\\mathrm{Bu} - \\mathrm{H}_{2}\\mathrm{O}$  cycles.","id":"train/atomic-layer-deposition/experimental-usecase/38/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/38/figure_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multiple line chart"},{"panel_id":"f","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The depth profile shows oxygen and manganese dominating the near-surface region of the MnO₂ film. Both elements remain fairly constant through most of the film thickness. Silicon rises sharply beyond ~1500 ×10¹⁵ at/cm², marking the substrate interface. No lithium is detected in this pristine sample.\"},{\"panel_id\":\"b\",\"text\":\"After 10 LiO′Bu–H₂O cycles, a small lithium signal appears near the surface. Oxygen and manganese profiles remain similar to pristine MnO₂. Lithium does not penetrate deeply and drops to zero before the Si interface. This suggests early-stage, surface-limited lithiation.\"},{\"panel_id\":\"c\",\"text\":\"With 50 cycles, lithium concentration increases and extends further into the film. The Li profile is still higher near the surface but now reaches deeper compared to 10 cycles. Mn and O remain stable within the film region. The MnO₂ framework appears preserved.\"},{\"panel_id\":\"d\",\"text\":\"At 100 cycles, lithium is clearly incorporated throughout most of the MnO₂ layer. The Li profile becomes flatter inside the film before dropping near the Si interface. Oxygen and manganese concentrations remain relatively unchanged. This indicates a transition toward bulk lithiation.\"},{\"panel_id\":\"e\",\"text\":\"The lithium signal strengthens further and becomes more uniform across the film thickness. Only a slight decrease is seen near the interface region. Mn and O profiles still show no major disruption. Lithiation is no longer confined to the surface.\"},{\"panel_id\":\"f\",\"text\":\"After 300 cycles, lithium shows a high and nearly constant concentration throughout the film. The Li profile closely tracks the MnO₂ layer thickness. Oxygen and manganese remain stable until the Si interface. This panel reflects fully developed bulk lithiation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Depth (×10^15 at/cm^2) | O (at.%) | Mn (at.%) | Si (at.%) | Li (at.%) |\\n|---|---|---|---|---|\\n| 0    | ~0  | ~0  | ~0   | ~0  |\\n| 200  | ~60 | ~38 | ~0   | ~0  |\\n| 500  | ~62 | ~36 | ~0   | ~0  |\\n| 1000 | ~62 | ~34 | ~0   | ~0  |\\n| 1400 | ~55 | ~25 | ~10  | ~0  |\\n| 1600 | ~10 | ~5  | ~80  | ~0  |\\n| 2000 | ~0  | ~0  | ~98  | ~0  |\\n| 2500 | ~0  | ~0  | ~100 | ~0  |\"},{\"panel_id\":\"b\",\"text\":\"| Depth (×10^15 at/cm^2) | O (at.%) | Mn (at.%) | Si (at.%) | Li (at.%) |\\n|---|---|---|---|---|\\n| 0    | ~0  | ~0  | ~0   | ~0  |\\n| 200  | ~60 | ~38 | ~0   | ~1  |\\n| 500  | ~60 | ~36 | ~0   | ~1  |\\n| 1000 | ~58 | ~33 | ~0   | ~1  |\\n| 1400 | ~40 | ~18 | ~20  | ~0  |\\n| 1600 | ~5  | ~2  | ~90  | ~0  |\\n| 2000 | ~0  | ~0  | ~98  | ~0  |\\n| 2500 | ~0  | ~0  | ~100 | ~0  |\"},{\"panel_id\":\"c\",\"text\":\"| Depth (×10^15 at/cm^2) | O (at.%) | Mn (at.%) | Si (at.%) | Li (at.%) |\\n|---|---|---|---|---|\\n| 0    | ~0  | ~0  | ~0   | ~0  |\\n| 200  | ~60 | ~38 | ~0   | ~3  |\\n| 500  | ~60 | ~36 | ~0   | ~3  |\\n| 1000 | ~58 | ~34 | ~0   | ~2  |\\n| 1400 | ~35 | ~15 | ~25  | ~0  |\\n| 1600 | ~5  | ~2  | ~92  | ~0  |\\n| 2000 | ~0  | ~0  | ~98  | ~0  |\\n| 2500 | ~0  | ~0  | ~100 | ~0  |\"},{\"panel_id\":\"d\",\"text\":\"| Depth (×10^15 at/cm^2) | O (at.%) | Mn (at.%) | Si (at.%) | Li (at.%) |\\n|---|---|---|---|---|\\n| 0    | ~0  | ~0  | ~0   | ~0  |\\n| 200  | ~58 | ~36 | ~0   | ~5  |\\n| 500  | ~60 | ~36 | ~0   | ~4  |\\n| 1000 | ~58 | ~33 | ~0   | ~2  |\\n| 1300 | ~45 | ~20 | ~10  | ~0  |\\n| 1500 | ~10 | ~5  | ~80  | ~0  |\\n| 2000 | ~0  | ~0  | ~98  | ~0  |\\n| 2500 | ~0  | ~0  | ~100 | ~0  |\"},{\"panel_id\":\"e\",\"text\":\"| Depth (×10^15 at/cm^2) | O (at.%) | Mn (at.%) | Si (at.%) | Li (at.%) |\\n|---|---|---|---|---|\\n| 0    | ~0  | ~0  | ~0   | ~0  |\\n| 200  | ~55 | ~34 | ~0   | ~10 |\\n| 500  | ~56 | ~32 | ~0   | ~11 |\\n| 1000 | ~56 | ~30 | ~0   | ~9  |\\n| 1400 | ~40 | ~15 | ~15  | ~2  |\\n| 1600 | ~5  | ~2  | ~92  | ~0  |\\n| 2000 | ~0  | ~0  | ~98  | ~0  |\\n| 2500 | ~0  | ~0  | ~100 | ~0  |\"},{\"panel_id\":\"f\",\"text\":\"| Depth (×10^15 at/cm^2) | O (at.%) | Mn (at.%) | Si (at.%) | Li (at.%) |\\n|---|---|---|---|---|\\n| 0    | ~0  | ~0  | ~0   | ~0  |\\n| 200  | ~52 | ~28 | ~0   | ~15 |\\n| 500  | ~52 | ~28 | ~0   | ~16 |\\n| 1000 | ~52 | ~27 | ~0   | ~13 |\\n| 1400 | ~35 | ~12 | ~20  | ~4  |\\n| 1600 | ~5  | ~2  | ~92  | ~0  |\\n| 2000 | ~0  | ~0  | ~98  | ~0  |\\n| 2500 | ~0  | ~0  | ~100 | ~0  |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It can be observed that, as the concentration of Li increases, Li can be found deeper in the sample.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxygen concentration remains approximately the same for all the samples and it can be found at depths of 1500-2000 x 10^15 at/cm^2.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"No lithiation in pristine MnO₂, Surface-limited lithiation at low cycle numbers (10c–100c), Homogeneous lithiation throughout the film at high cycle numbers (200c–300c)\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface-localized lithium signal suggests that lithium incorporation is favored near the film surface rather than uniformly throughout the bulk. This behavior is consistent with surface-limited incorporation or post-deposition modification effects, rather than bulk diffusion during growth.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Only the surface of the film is lithiated. The depth profile shows, that lithium concentration peaks near the surface and drops to neglibly low values at approx. 500x10^15 at/cm². On the other hand, the manganese and oxygene signals extend much deeper (approx. 1500 x10^15 at/cm²), indicating that the bulk of the film consists of pure manganese oxide.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lithium in the top layer rises from 0.6 at.% (10c) to 3.2 at.% (50c) to 9.6 at.% (100c). After 300 cycles, the film becomes stoichiometric LiMn₂O₄.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 50 cycles, lithium penetrates deeper into the MnO₂ film compared to 10 cycles. While the concentration remains highest near the surface, the Li signal now extends further toward the film interior. This reflects a progression from surface-limited to partial bulk lithiation.\"}]},{\"panel_id\":\"f\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It can be observed that, as the concentration of Li increases, Li can be found deeper in the sample.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxygen concentration remains approximately the same for all the samples and it can be found at depths of 1500-2000 x 10^15 at/cm^2.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The 300cLi-MnO2.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is relatively homogeneous, reaching a concentration of around15 at%. Furthermore, it can be found throughout the bulk of the material.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At low cycle numbers (10-50), lithium penetrates only the surface region. At 100 cycles, Li reaches ~50% of the film depth. Above 100 cycles (200-300), lithium distributes throughout the entire film thickness, enabling stoichiometric LiMn₂O₄ formation.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"-Enables uniform electrochemical activity\\n-Reduces concentration gradients\\n-Improves cycling stability\\n-Enhances charge transport consistency\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The lithium is not evely distributed. Deeper in the film there is less lithium, this could be due to a nucleation delay.\"}]},{\"panel_id\":\"e\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It can be observed that, as the concentration of Li increases, Li can be found deeper in the sample.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxygen concentration remains approximately the same for all the samples and it can be found at depths of 1500-2000 x 10^15 at/cm^2.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is relatively homogeneous, reaching a concentration of around15 at%. Furthermore, it can be found throughout the bulk of the material.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 1500 x10^15 at/cm².\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It can be observed that, as the concentration of Li increases, Li can be found deeper in the sample.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxygen concentration remains approximately the same for all the samples and it can be found at depths of 1500-2000 x 10^15 at/cm^2.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Homogeneous lithium distribution ensures uniform conversion of MnO₂ into stoichiometric LiMn₂O₄, which is essential for stable electrochemical performance and reproducible material properties.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Oxygen.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"In the samples 200cLi−MnO2 and 300cLi−MnO2, lithium is observed to penetrate throughout the whole film thickness.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Partial lithiation creates a layered structure with only the surface enriched in Li. Full lithiation produces a uniform bulk composition. This homogeneity improves mechanical and electrochemical stability.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Depth profiles show very low impurity levels. High purity minimizes side reactions, improves cycling stability, and is essential for achieving high performance and long life in battery cathodes.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At x ≈ 0.5 in LiₓMn₂O₄ (~100 cycles), the crystal structure transforms from tetragonal β-MnO₂ to cubic spinel. The spinel structure has 3D Li⁺ diffusion pathways, enabling lithium to migrate throughout the film rather than being confined to surface channels.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The deposition is performed on a silicon wafer, based on the high silicon contant deeper in the sample. It is likely that the starting surface is thermal silicon oxide.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Lithium.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The depth profiles reveal that lithium is distributed only in the top part of the film in the samples 10cLi−MnO2, 50cLi−MnO2, and 100cLi−MnO2, that is, when the average composition of the material is LixMn2O4, x ≤ 0.5. In the lithiated top parts of the films in these samples, the lithium concentrations were 0.6, 3.2, and 9.6 at. %, respectively.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A weak lithium signal appears near the surface but does not penetrate deeply.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It can be observed that, as the concentration of Li increases, Li can be found deeper in the sample.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxygen concentration remains approximately the same for all the samples and it can be found at depths of 1500-2000 x 10^15 at/cm^2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"There is no lithium incorporation for 10 cycles of lithium deposition, this indicates that there is a delay.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It can be observed that, as the concentration of Li increases, Li can be found deeper in the sample.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxygen concentration remains approximately the same for all the samples and it can be found at depths of 1500-2000 x 10^15 at/cm^2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing the number of LiOtBu–H₂O cycles progressively enhances lithium incorporation into the MnO₂ films. At low cycle numbers, lithium is confined to the near-surface region, while higher cycle numbers enable lithium to penetrate deeper into the film. At 200 and 300 cycles, lithium is distributed throughout the entire film thickness, indicating complete and homogeneous gas-phase lithiation.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Across all samples, oxygen and manganese show relatively stable concentration profiles through most of the film thickness, while lithium is more concentrated near the surface and decreases with depth. Silicon concentration increases toward the interface region, indicating proximity to the substrate. The overall depth-dependent trends are similar between samples, with differences primarily in the relative prominence of the lithium signal.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~1:2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"33:66.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300 cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.6, 3.2 and 9.6 at.% Li, respectively.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The depth profiles show that at low cycle numbers (10-100 cycles), lithium remains confined to the upper layers of the MnO2 film. This results in compositions of LixMn2O4 with x < 0.5, where the lithium concentrations in the surface region are 0.6-9.6 at.%, depending on cycle count. At higher cycle numbers, specifically 200 and 300 cycles, lithium penetrates through the entire film, indicating complete lithiation.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The depth profiles correspond to the following samples\\na. MnO2\\nb. 10cLi-MnO2\\nc. 50cLi-MnO2\\nd. 100cLi-MnO2\\ne. 200cLi-MnO2\\nf. 300cLi-MnO2\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lithium is initially confined to the surface region after 10–100 cycles. It gradually penetrates deeper, achieving full, homogeneous distribution throughout the entire film after 200–300 cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Si signal marks the interface between the MnO₂/LiₓMn₂O₄ film and the Si(100) substrate.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Time-of-flight elastic recoil detection analysis (TOF-ERDA) using ⁷⁹Br⁷⁺ ions from a 5 MV tandem accelerator.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Oxygen and manganese dominate the near-surface region.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The data suggests that the ratio of Mn:O is not 1:2. This means that the film is a mixture of MnO2 and Mn2O3.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":327,"height":308},{"panel_id":"b","x":329,"y":5,"width":327,"height":319},{"panel_id":"c","x":667,"y":5,"width":312,"height":314},{"panel_id":"d","x":11,"y":372,"width":313,"height":300},{"panel_id":"e","x":347,"y":367,"width":311,"height":305},{"panel_id":"f","x":685,"y":366,"width":293,"height":308}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/38/Nieminen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":983,"height":675,"image_format":"jpeg","image_sha256":"b6b42c35d58106cd65fea698f87aa2f56e6be58ad6e5f7f7fa696ac4ccb8513d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_4_fig_2.jpg","caption":"Fig. 2. Thickness profile of selected  $\\mathrm{HfO_2}$  films grown from  $\\mathrm{Hf[N(CH_3) - }$ $(\\mathrm{C}_2\\mathrm{H}_5)]_4$  and  $\\mathrm{H}_2\\mathrm{O}$  on borosilicate glass substrates. Growth temperatures and numbers of growth cycles applied are indicated by labels. Thicknesses are measured at variable distance from the leading edge of the substrate, which is the edge closest to the precursor inlet.","id":"train/atomic-layer-deposition/experimental-usecase/4/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/4/fig_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the thickness of HfO₂ film as a function of distance from the leading edge at various growth temperatures and cycle counts.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| DISTANCE FROM THE LEADING EDGE [mm] | 150 °C, 1000 cycles [nm] | 250 °C, 1500 cycles [nm] | 275 °C, 1500 cycles [nm] | 300 °C, 1500 cycles [nm] | 225 °C, 2500 cycles [nm] |\\n|---|---|---|---|---|---|\\n| 5 | 110 | 140 | 160 | 180 | 210 |\\n| 10 | 110 | 140 | 160 | 180 | 210 |\\n| 15 | 110 | 140 | 160 | 180 | 210 |\\n| 20 | 110 | 140 | 150 | 180 | 200 |\\n| 25 | 100 | 140 | 150 | 170 | 200 |\\n| 30 | 100 | 130 | 150 | 170 | 200 |\\n| 35 | 100 | 130 | 150 | 160 | 200 |\\n| 40 | 100 | 130 | 150 | 160 | 200 |\\n| 45 | 100 | 130 | 150 | 160 | 200 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"150, 225, 250, 275 and 300 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1000 cycles for 150 °C, 1500 cycles for 250, 275 and 300 °C, 2500 cycles for 225 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 5 mm, the film thickness is about 110 nm and at 45 mm, the film thickness is about 100 nm. The total change in thickness is therefore around 10 nm.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The samples prepared above 250 °C have a more pronounced difference in thickness compared to the beginning and end of the leading edge. The distance from the leading edge corresponds to how close that part of the film is to the precursor inlet. It is theorised by the authors that above 250 °C, the thermal decomposition of the precursor, which happens at all temperatures, becomes more significant above that temperature. Parts of the film that are closer to the precursor inlet are therefore more likely to be thicker, leading to an additional dependency.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":633,"height":397}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/Kukli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":636,"height":400,"image_format":"jpeg","image_sha256":"1e3eb18cdf651c650285c8d9708c43da932dfb74f104ecf917c5c9ccf4ffd24e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_4_fig_7.jpg","caption":"Fig. 7. Concentrations of hydrogen, carbon, and nitrogen residues versus growth temperature. Lines are guides for the eye.","id":"train/atomic-layer-deposition/experimental-usecase/4/fig_7","sample_id":"atomic-layer-deposition/experimental-usecase/4/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the concentration of hydrogen, carbon, and nitrogen at various HfO₂ growth temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| HfO2 GROWTH TEMPERATURE, °C | at.% H | at.% C | at.% N |\\n|---|---|---|---|\\n| 150 | 5 | 0.5 | 0.1 |\\n| 175 | 3 | 0.8 | 0.2 |\\n| 200 | 3 | 0.6 | 0.3 |\\n| 225 | 2.5 | 0.7 | 0.4 |\\n| 250 | 2 | 0.9 | 0.5 |\\n| 275 | 2.5 | 1.1 | 0.6 |\\n| 300 | 2.5 | 1.2 | 0.7 |\\n| 325 | 5 | 1.5 | 0.8 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 175 and 300 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"From 275 °C on.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"250 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 2 at.%.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":556,"height":362}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/4/Kukli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":556,"height":364,"image_format":"jpeg","image_sha256":"eb750c7c36bf46b73fe6dd9abbc2afd5b1feeb03e366350cf3fe1e6365be0888","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_42_figure_1.jpg","caption":"Figure 1. Thermogravimetric analyses and evaporation rates of 1 (black) and 2 (gray) run at  $10^{\\circ}\\mathrm{C / min}$  ramp rate, with  $10^{\\circ}\\mathrm{C}$  increments for the stepped isotherm used to determine evaporation rate.","id":"train/atomic-layer-deposition/experimental-usecase/42/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/42/figure_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Thermogravimetric analyses and evaporation rates of 1 (black) and 2 (gray) run at 10 °C/min ramp rate, with 10 °C increments for the stepped isotherm used to determine evaporation rate. TGA graph shows how much mass is left as Compound 1 is heated. The black trace indicates that all of the compound vaporized cleanly starting at 91 °C, leaving 0% residue by about 150 °C.\"},{\"panel_id\":\"b\",\"text\":\"This label shows the evaporation rate (in nmoles/min) of two compounds (1 in black, 2 in gray) as temperature increases (plotted as 1/T in K⁻¹ × 10³). Both lines are straight and sloping upward,the steeper slope for Compound 1 suggests it has a higher activation energy for evaporation than Compound 2, meaning it is more temperature-sensitive.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| temperature (°C) | mass loss (%) |material  |\\n|---|---|---|\\n| 50 | 100 | 1 |\\n| 50 | 100 | 2 |\\n| 100 | 98 | 1 |\\n| 100 | 98 | 2 |\\n| 150 | 92 | 1 |\\n| 150 | 92 | 2 |\\n| 200 | 0  | 1 |\\n| 200 | 20 | 2 |\\n| 250 | 0  | 1 |\\n| 250 | 18 | 2 |\\n| 300 | 0  | 1 |\\n| 300 | 17 | 2 |\\n| 350 | 0  | 1 |\\n| 350 | 15 | 2 |\"},{\"panel_id\":\"b\",\"text\":\"| temperature (1/K x10^3) | evaporation rate (nmol/min) |material  |\\n|---|---|---|\\n| 2.25 | 1200 | 1  |\\n| 2.25 | 1100 | 2  |\\n| 2.3  | 1050 | 1  |\\n| 2.3  | 1030 | 2  |\\n| 2.35 | 1000 | 1  |\\n| 2.35 | 800 | 2  |\\n| 2.41 | 800 | 1  |\\n| 2.41 | 600 | 2  |\\n| 2.47 | 600 | 1  |\\n| 2.47 | 500 | 2  |\\n| 2.5  | 500 | 1  |\\n| 2.5  | 300 | 2  |\\n| 2.52 | 300 | 1  |\\n| 2.52 | 150 | 2  |\\n| 2.6 | 150 | 1  |\\n| 2.6 | 100 | 2  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TGA measures how much mass is lost as a sample is subject to increased temperature. Thermogravimetric analysis (TGA) showed the onset of volatility at 91 °C and a residual mass of 0%.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Compound 1 vaporized completely with 0% residue, while Compound 2 left 7.4% solid residue, meaning it was less cleanly volatile and likely partially decomposed during heating. Since the mass of copper in 2 comprises 15.72% of the total molecular mass, this compound was possibly undergoing volatilization and decomposition over the thermal range of the TGA.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Because Compound 1’s evaporation rate changes sharply with small temperature changes (steep slope). To keep the vapor supply stable and repeatable for uniform film growth, the precursor temperature must be controlled very accurately (±1 °C or better).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Plotting against 1/T (inverse Kelvin) gives a straight line when evaporation follows the Arrhenius equation. The slope of the line tells us the activation energy for evaporation – a steeper slope (like Compound 1) means evaporation is more sensitive to temperature changes.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":501,"height":396},{"panel_id":"b","x":491,"y":3,"width":436,"height":395}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/42/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/42/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/42/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/42/Coyle et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"42","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":930,"height":400,"image_format":"jpeg","image_sha256":"9b0cb07188191a4ae89c68a43ea404abc713d6e6ffa937be89db6879ebe82323","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_45_figure_1.jpg","caption":"Figure 1. (Color online) Thickness measured by in situ SE as a function of the number of cycles for the  $\\mathrm{Pt}$  and  $\\mathrm{PtO_2}$  ALD processes. The process conditions are listed in Table I. The starting substrate at 0 cycles was  $\\mathrm{Si(100)}$  with  $400 \\mathrm{nm} \\mathrm{SiO_2}$ . After 150 cycles of remote plasma ALD, the  $\\mathrm{Pt}$  film growth is continued by thermal ALD.","id":"train/atomic-layer-deposition/experimental-usecase/45/figure_1","sample_id":"atomic-layer-deposition/experimental-usecase/45/figure_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the thickness of materials deposited using different ALD methods (Remote plasma ALD Pt, Remote plasma ALD PtO₂, and Thermal ALD Pt) as a function of the number of cycles. Thermal Pt ALD shows no growth unless  applied after a Pt seed layer has been formed by remote plasma ALD. The remote plasma processes show near-linear thickness increase after a short initial nucleation period.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sample ID | Deposition mode and Material |  Number of cycles | Thickness (nm) | \\n|---|---|---|---|\\n| A | Thermal ALD Pt | 0 | 0 |\\n| A | Thermal ALD Pt | 50 | 0.05 |\\n| A | Thermal ALD Pt | 100 | 0.05 |\\n| A | Thermal ALD Pt | 150 | 0.05 |\\n| A | Thermal ALD Pt | 200 | 0.05 |\\n| A | Thermal ALD Pt | 250 | 0.06 |\\n| B | Remote plasma ALD PtO2| 0 | 0 |\\n| B | Remote plasma ALD PtO2| 20 | 0.08 |\\n| B | Remote plasma ALD PtO2| 30 | 0.37 |\\n| B | Remote plasma ALD PtO2| 40 | 0.77 |\\n| B | Remote plasma ALD PtO2| 50 | 1.47 |\\n| B | Remote plasma ALD PtO2| 60 | 2.47 |\\n| B | Remote plasma ALD PtO2| 130 | 4.92 |\\n| B | Remote plasma ALD PtO2| 200 | 7.76 |\\n| B | Remote plasma ALD PtO2| 250 | 9.98 |\\n| C | Remote plasma ALD Pt| 0 | 0 |\\n| C | Remote plasma ALD Pt| 20 | 0.52 |\\n| C | Remote plasma ALD Pt| 40 | 1.43 |\\n| C | Remote plasma ALD Pt| 60 | 2.99 |\\n| C | Remote plasma ALD Pt| 100 | 5.07 |\\n| C | Remote plasma ALD Pt| 150 | 7.23 |\\n| C | Thermal ALD Pt| 160 | 7.45 |\\n| C | Thermal ALD Pt| 200 | 8.97 |\\n| C | Thermal ALD Pt| 250 | 10.89 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After approximately 50 cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the plot, thermal ALD Pt on the initial surface stays essentially at zero thickness over the full cycle range, indicating little to no growth. After 150 cycles of remote plasma ALD Pt, the process is switched to thermal ALD on the same sample and the thickness then increases steadily. This contrast shows that thermal ALD Pt growth proceeds when a Pt surface (seed layer) is present, but not on the starting  surface.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Remote plasma ALD Pt.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After roughly 50 cycles.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":607,"height":464}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/45/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/45/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/45/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/45/Knoops et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"45","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":611,"height":467,"image_format":"jpeg","image_sha256":"e136f2c9af1cd0f9431b4fee3598307cd7cfa973f1ac24eef80baf5a090478c7","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_5_fig_11.jpg","caption":"Fig. 11. Chemical compositions of TiN ALD films measured using XPS techniques. (a) Composition vs. deposition temperature using the short reactant exposure pulse sequence. (b) Composition vs. TDMAT exposure at  $180^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/5/fig_11","sample_id":"atomic-layer-deposition/experimental-usecase/5/fig_11","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"As deposition temperature increases, oxygen content decreases markedly while nitrogen content increases, whereas titanium remains nearly constant, indicating improved Ti–N formation at higher temperatures. Carbon remains low and only weakly temperature dependent.\"},{\"panel_id\":\"b\",\"text\":\"Increasing TDMAT exposure at 180 °C raises oxygen content and reduces nitrogen content, while titanium stays stable, suggesting that excess precursor exposure promotes oxygen incorporation rather than additional nitridation. Overall, deposition temperature appears to be the dominant parameter governing TiN film stoichiometry.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition Temperature (°C) | O (at.%) | Ti (at.%) | N (at.%) | C (at.%) |\\n|---|---|---|---|---|\\n| ~60 | ~56 | ~31 | ~9 | ~3 |\\n| ~120 | ~44 | ~32 | ~19 | ~5 |\\n| ~180 | ~39 | ~32 | ~23 | ~5 |\\n| ~240 | ~37 | ~33 | ~28 | ~6 |\"},{\"panel_id\":\"b\",\"text\":\"| TDMAT Exposure (s) | O (at.%) | Ti (at.%) | N (at.%) | C (at.%) |\\n|---|---|---|---|---|\\n| 0 | ~40 | ~32 | ~23 | ~5 |\\n| ~20 | ~45 | ~31 | ~18 | ~6 |\\n| ~100  | ~46 | ~31 | ~17 | ~7 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Extended precursor exposure increases surface residence time of reactive species, which can promote unintended oxygen uptake. Rather than enhancing nitridation, excess exposure appears to favor impurity incorporation. This suggests saturation of Ti sites occurs early in the exposure window.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Oxygen content saturates after short TDMAT exposures and remains high at longer exposure times.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The opposing oxygen and nitrogen trends indicate competitive incorporation during growth. As temperature increases, nitrogen more effectively replaces oxygen within the film, promoting Ti–N bonding. This reflects a shift toward a more stoichiometric TiN phase.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Higher deposition temperatures are most favorable due to lower oxygen and higher nitrogen content.\"}]}]","bbox":[{"panel_id":"b","x":1,"y":581,"width":541,"height":564},{"panel_id":"a","x":1,"y":0,"width":541,"height":558}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":547,"height":1152,"image_format":"jpeg","image_sha256":"5402d7ed615a9506cceb32baab8e48bafce4b58a23182d8a5ef0314d519c98ca","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_5_fig_12.jpg","caption":"Fig. 12. QCM mass ratios predicted for TiN ALD vs. the number of dimethylamine molecules produced during the TDMAT reaction,  $x$ , and the number of dimethylamino ligands incorporated per Ti atom,  $y$ .","id":"train/atomic-layer-deposition/experimental-usecase/5/fig_12","sample_id":"atomic-layer-deposition/experimental-usecase/5/fig_12","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents predicted QCM mass ratios for TiN ALD as a function of the number of dimethylamine molecules produced during the TDMAT reaction (x). Each curve corresponds to a different number of dimethylamino ligands retained per Ti atom (y). For all values of y, the QCM mass ratio increases monotonically with x. At a fixed x, higher ligand retention shifts the curves upward, indicating larger net mass uptake. The model therefore highlights the coupled influence of byproduct formation and incomplete ligand removal on QCM signals.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| y (ligand per Ti) | x (approx. range) | QCM mass ratio R (approx.) |\\n|----------------------|-----------------------|--------------------------------------|\\n| 0.0 | ~2.7–3.0 | ~0.60–0.70 |\\n| 0.5 | ~2.1–2.9 | ~0.70–0.90 |\\n| 1.0 | ~1.6–2.7 | ~0.65–1.00 |\\n| 1.5 | ~1.3–2.4 | ~0.75–1.10 |\\n| 2.0 | ~1.0–2.0 | ~0.85–1.15 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Increased production of dimethylamine byproducts during the TDMAT reaction.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It increases the QCM mass ratio.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Retaining more dimethylamino ligands increases the mass remaining on the surface after reaction. This added mass contribution raises the predicted QCM mass ratio for any given amount of dimethylamine production.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":665,"height":681}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":667,"height":686,"image_format":"jpeg","image_sha256":"bc9b46e688b773052837b6135c081f2b0faa240bd2da912aa8196f53d5eac890","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_5_fig_3.jpg","caption":"Fig. 3. Integrated absorbance of  $\\mathrm{C - H}$  and  $\\mathrm{N - H}$  stretching vibrations vs. TDMAT and  $\\mathrm{NH}_3$  exposures at  $152^{\\circ}\\mathrm{C}$ . (a) Third TDMAT exposure in ABAB... sequence. (b) Fourth  $\\mathrm{NH}_3$  exposure in ABAB... sequence.","id":"train/atomic-layer-deposition/experimental-usecase/5/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/5/fig_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure tracks surface functional group evolution during an ABAB… ALD sequence using integrated IR absorbances of C–H and N–H stretching modes at 152 °C. In panel (a), increasing TDMAT exposure causes the N–H signal to rapidly decrease toward zero while the C–H signal rises and approaches saturation, indicating that TDMAT consumption of NHx*-terminated sites is accompanied by accumulation of methyl-containing ligands on the surface.\"},{\"panel_id\":\"b\",\"text\":\"In this figure, increasing NH₃ exposure produces the opposite behavior: N–H absorbance increases and saturates while C–H absorbance decreases and levels off, consistent with NH₃-driven ligand removal and surface re-termination by NHx* species. Together, the two panels demonstrate complementary half-reactions and self-limiting surface chemistry characteristic of ALD cycling.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TDMAT Exposure (Torr·min) | C–H (integrated absorbance) | N–H (integrated absorbance) |\\n|---|---|---|\\n| 0   | ~0.23 | ~1.00 |\\n| 3   | ~0.78 | ~0.25 |\\n| 6   | ~0.88 | ~0.15 |\\n| 10  | ~0.90 | ~0.10 |\\n| 15  | ~0.94 | ~0.06 |\\n| 20  | ~0.95 | ~0.04 |\\n| 25  | ~0.98 | ~0.02 |\\n| 35  | ~1.00 | ~0.01 |\"},{\"panel_id\":\"b\",\"text\":\"| NH₃ Exposure (Torr·min) | C–H (integrated absorbance) | N–H (integrated absorbance) |\\n|---|---|---|\\n| 0     | ~1.00 | ~0.00 |\\n| 500   | ~0.45 | ~0.78 |\\n| 1000  | ~0.38 | ~0.95 |\\n| 1500  | ~0.36 | ~0.95 |\\n| 2000  | ~0.34 | ~0.85 |\\n| 3000  | ~0.34 | ~0.95 |\\n| 3500  | ~0.33 | ~0.98 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It implies that NH₃ effectively removes methyl-derived ligands and re-terminates the surface with NHx* species. This regeneration of reactive sites is essential for sustaining cyclic ALD growth.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Residual C–H indicates carbon-containing fragments remain on the surface, which can incorporate into the growing film as impurities. Carbon contamination can increase resistivity and degrade barrier performance, so effective NH₃ exposure is critical for high-quality TiN films.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The N–H absorbance decreases most rapidly during the initial TDMAT exposure.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Saturation of the C–H signal implies that available reactive surface sites have been fully occupied by methyl-containing ligands. Additional precursor exposure does not increase adsorption, indicating complete surface coverage. This confirms self-limiting precursor chemisorption.\"}]}]","bbox":[{"panel_id":"b","x":3,"y":535,"width":505,"height":503},{"panel_id":"a","x":3,"y":6,"width":505,"height":508}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":511,"height":1042,"image_format":"jpeg","image_sha256":"56c206d3ac55af7eff76ac66f732192ac0be3aca21d16ed8d3b85c68076db0df","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_5_fig_8.jpg","caption":"Fig. 8. QCM mass ratio,  $R = (m_0 + m_1) / m_0$ , measured during TiN ALD using long, medium and short reactant pulse sequences at deposition temperatures of  $60 - 240^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/5/fig_8","sample_id":"atomic-layer-deposition/experimental-usecase/5/fig_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This plot shows the QCM mass ratio as a function of deposition temperature for three reactant pulse sequences. The short-pulse sequence remains lowest across the full temperature range (≈0.60 → 0.93), indicating a smaller relative contribution from the NH₃ half-cycle. The medium-pulse sequence starts high at 50 °C, dips near 90 °C, and then increases steadily to ~1.16 at 240 °C. The long-pulse sequence stays near ~0.85–0.95 up to 150 °C and then rises above unity at higher temperatures, reflecting a growing contribution from the second half-reaction.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse sequence | Deposition Temperature (°C) | QCM Mass Ratio R |\\n|---|---|---|\\n| Long (100-10-20-30) | 50  | ~0.88 |\\n| Long (100-10-20-30) | 90  | ~0.84 |\\n| Long (100-10-20-30) | 120 | ~0.84 |\\n| Long (100-10-20-30) | 150 | ~0.92 |\\n| Long (100-10-20-30) | 180 | ~1.02 |\\n| Long (100-10-20-30) | 210 | ~1.07 |\\n| Long (100-10-20-30) | 240 | ~1.11 |\\n| Medium (20-10-6-10) | 50  | ~1.06 |\\n| Medium (20-10-6-10) | 90  | ~0.86 |\\n| Medium (20-10-6-10) | 120 | ~0.88 |\\n| Medium (20-10-6-10) | 150 | ~0.98 |\\n| Medium (20-10-6-10) | 180 | ~1.08 |\\n| Medium (20-10-6-10) | 210 | ~1.14 |\\n| Medium (20-10-6-10) | 240 | ~1.16 |\\n| Short (1-0.5-1-5)   | 50  | ~0.60 |\\n| Short (1-0.5-1-5)   | 90  | ~0.66 |\\n| Short (1-0.5-1-5)   | 120 | ~0.67 |\\n| Short (1-0.5-1-5)   | 150 | ~0.70 |\\n| Short (1-0.5-1-5)   | 180 | ~0.78 |\\n| Short (1-0.5-1-5)   | 210 | ~0.80 |\\n| Short (1-0.5-1-5)   | 240 | ~0.93 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Medium (20-10-6-10), R ≈ 1.16.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"180 °C (R ≈ 1.20)\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"An increasing R indicates that the net mass change after a full cycle grows relative to the precursor-only mass gain. This suggests that the NH₃ half-cycle contributes more strongly at higher temperatures, likely due to more complete ligand removal or enhanced reaction progress.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":663,"height":643}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":670,"height":644,"image_format":"jpeg","image_sha256":"7b1f3fe5f9a1ad9d366a84108def0591f22f5c461839c24ce72ea77b4c313574","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_5_fig_9.jpg","caption":"Fig. 9. Resistivities of TiN ALD films measured using the four-point probe for long, medium and short reactant pulse sequences at deposition temperatures of  $60 - 240^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/5/fig_9","sample_id":"atomic-layer-deposition/experimental-usecase/5/fig_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure shows the electrical resistivity of TiN ALD films as a function of deposition temperature. For all pulse lengths, resistivity decreases sharply with increasing temperature. Short pulse sequences consistently yield the lowest resistivity at high temperature. Medium pulses show intermediate behavior, while long pulses retain substantially higher resistivity, especially at lower temperatures. The results indicate that both deposition temperature and pulse timing strongly influence TiN film electrical quality.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse sequence | Deposition temperature (°C) | Resistivity (µΩ·cm) |\\n|---|---|---|\\n| Short  | 60  | ~2.5×10⁷ |\\n| Short  | 120 | ~5×10⁶ |\\n| Short  | 180 | ~5×10⁴ |\\n| Short  | 240 | ~1.5×10⁴ |\\n| Medium | 120 | ~2×10⁷ |\\n| Medium | 180 | ~3×10⁵ |\\n| Medium | 240 | ~2×10⁴ |\\n| Long   | 180 | ~1×10⁷ |\\n| Long   | 240 | ~3×10⁵ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The short pulse sequence.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It decreases by several orders of magnitude.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Shorter pulses limit excessive precursor exposure and reduce residual ligand incorporation in the growing film. At higher temperatures, this favors cleaner surface reactions, improved film densification, and better grain connectivity. These structural improvements reduce electron scattering and lead to lower resistivity compared with longer pulse sequences.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":656,"height":667}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/5/Elam et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":661,"height":673,"image_format":"jpeg","image_sha256":"1900f10398e03594987ffddf0793abb20dcec6410b3927286280b01aba4d05a2","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_50_fig_10.jpg","caption":"Fig. 10 Absorption coefficient  $(\\alpha)$  as a function of the photon energy  $(E)$  of the ALD ZnO:B films deposited at  $150^{\\circ}C$  on glass substrates.","id":"train/atomic-layer-deposition/experimental-usecase/50/fig_10","sample_id":"atomic-layer-deposition/experimental-usecase/50/fig_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"table"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the absorption coefficient  of ZnO:B as a function of photon energy E, with various doping levels (DF) indicated by different colors, it also highlights the shift in the absorption edge with increasing doping.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| DF | \\\\(\\\\alpha\\\\) \\\\((10^{3} \\\\text{ cm}^{-1})\\\\) @ 1.2 eV | n \\\\((10^{20} \\\\text{ cm}^{-3})\\\\) |\\n| --- | --- | --- |\\n| intrinsic ZnO | 0.41 | 0.13 |\\n| 0.014 | 0.79 | 1.22 |\\n| 0.016 | 0.96 | 1.79 |\\n| 0.018 | 1.50 | 2.25 |\\n| 0.027 | 1.90 | 2.25 |\\n| 0.034 | 2.60 | 3.10 |\\n| 0.053 | 2.66 | 2.70 |\\n| 0.063 | 2.18 | 1.85 |\\n| 0.075 | 1.59 | 1.34 |\\n| 0.094 | 1.39 | 1.08 |\"},{\"panel_id\":\"b\",\"text\":\"\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No. From SE data.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Clearly in the NIR region, the absorption decreases with increasing DF up to 0.034 whereas it increases again for\\nhigher DF values.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The absorption increase in NIR region with increase in doping can be attributed to the free-carrier absorption in the films as described by the Drude model\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The carrier concentration is highest at the following DF,\\n1. At DF 0.034 = 3.1\\n2. At DF 0.053 = 2.7\\nfurther increase of DF does not increase the concentration and hence this is attributed to FCA\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":531,"height":360},{"panel_id":"b","x":531,"y":34,"width":432,"height":327}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/Garcia-Alonso et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"50","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":961,"height":367,"image_format":"jpeg","image_sha256":"93c18505542a2e0043ca341569e72583ba73b9acc8aece269e02c9ac15a8c949","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_50_fig_9.jpg","caption":"Fig. 9 Transmittance of selected ZnO:B samples  $(\\mathrm{DF} = 0.000 - 0.063)$  deposited at  $150^{\\circ}C$  and with thicknesses of  $45\\pm 5$  nm.","id":"train/atomic-layer-deposition/experimental-usecase/50/fig_9","sample_id":"atomic-layer-deposition/experimental-usecase/50/fig_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the transmittance of ZnO:B at different wavelengths for various doping fractions. The transmittance increases with increasing wavelength, reaching nearly 100% above 1000 nm.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Wavelength (nm) | Intrinsic ZnO (%) | DF = 0.016 (%) | DF = 0.018 (%) | DF = 0.034 (%) | DF = 0.063 (%) |\\n|-----------------|--------------------|-----------------|-----------------|-----------------|-----------------|\\n| 300             | 40                | 38              | 36              | 34              | 32              |\\n| 400             | 85                | 82              | 80              | 78              | 75              |\\n| 600             | 92                | 90              | 89              | 87              | 85              |\\n| 1000            | 94                | 93              | 92              | 91              | 90              |\\n| 1500            | 95                | 94              | 93              | 92              | 91              |\\n| 2000            | 95                | 94              | 93              | 92              | 91              |\\n| 2400            | 95                | 94              | 93              | 92              | 91              |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The transmittance values were over 90% in the major part of the visible region\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A decrease in transmittance was observed in the NIR region with increasing doping fraction until DF = 0.034. For higher DF values, the transmittance in the NIR increased again. This behaviour can be attributed to free-carrier absorption.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A blue shift was observed in the ultraviolet absorption edge with increasing doping concentration from intrinsic ZnO to doped ZnO:B films with a DF of 0.063. This is due to the increase of the optical band gap.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":458,"height":338}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/Garcia-Alonso et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"50","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":461,"height":344,"image_format":"jpeg","image_sha256":"30ae994cd64be7d5e94a99c6fc8202055aa4e4d77bf9e7f2fe50f5bd451f54ea","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_52_figure_5.jpg","caption":"Figure 5. The imaginary part of the dielectric function  $(\\epsilon_{2})$  of the  $\\mathrm{Co}_3\\mathrm{O}_4$ ,  $\\mathrm{Li}_2\\mathrm{CO}_3$  and  $\\mathrm{LiCoO_2}$  films as determined by in situ spectroscopic ellipsometry.","id":"train/atomic-layer-deposition/experimental-usecase/52/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/52/figure_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The imaginary part of the dielectric function, epsilon2, is plotted versus photon energy for Co3O4, LiCoO2, and Li2CO3, as determined by in situ spectroscopic ellipsometry. Co3O4 shows the largest epsilon2 values, LiCoO2 is much lower, and Li2CO3 remains near zero across most of the range, indicating very weak contribution compared with the other films.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Photon energy (eV) | Co3O4 (epsilon_2) | LiCoO2 (epsilon_2) | Li2CO3 (epsilon_2) |\\n|---|---|---|---|\\n| 1.25 | 1.2 | 0.15 | 0.00 |\\n| 1.50 | 2.5 | 0.18 | 0.00 |\\n| 1.65 | 5.5 | 0.20 | 0.00 |\\n| 2.00 | 4.1 | 0.25 | 0.00 |\\n| 2.40 | 4.8 | 0.35 | 0.00 |\\n| 2.75 | 6.0 | 0.50 | 0.00 |\\n| 3.10 | 5.2 | 0.70 | 0.00 |\\n| 3.40 | 4.6 | 0.60 | 0.01 |\\n| 3.70 | 4.5 | 0.60 | 0.01 |\\n| 4.00 | 4.7 | 0.75 | 0.02 |\\n| 4.25 | 4.9 | 0.90 | 0.03 |\\n| 4.75 | 5.6 | 1.25 | 0.05 |\\n| 5.00 | 5.8 | 1.20 | 0.06 |\\n| 5.40 | 6.0 | 1.15 | 0.08 |\\n| 5.80 | 5.8 | 1.05 | 0.10 |\\n| 6.00 | 5.4 | 1.00 | 0.12 |\\n| 6.50 | 4.6 | 0.80 | 0.05 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"These reference spectra help distinguish which phases contribute to the optical response measured during growth. The paper notes a clear difference between the three films and uses this comparison to interpret the LiCoO2 signal. In particular, the discussion indicates that Li2CO3 contributes very little to the LiCoO2 epsilon2 spectrum under these conditions.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Co3O4\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The x axis is photon energy (eV) and the y axis is the dielectric function epsilon2. Epsilon2 is the imaginary part of the dielectric function and is directly related to optical absorption in the material. Larger epsilon2 values generally indicate stronger absorption at those photon energies.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Co3O4\\n\\nLiCoO2\\n\\nLi2CO3\\nSmallest epsilon2 curve: Li2CO3\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":541,"height":370}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/Donders et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"52","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":544,"height":375,"image_format":"jpeg","image_sha256":"22b61f796669c9d3188569d8f062a76ac2cae3b54b1de0134a67793ab997a19c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_52_figure_9.jpg","caption":"Figure 9. Constant current (CC) (dis)charge cycling between  $3.0\\mathrm{V}$  and  $4.1\\mathrm{V}$  (0.35 C-rate) for an ALD-deposited  $\\mathrm{LiCoO_2}$ $(x = 4)$  film on  $\\mathrm{Si / TiO_2 / Pt}$ , using  $\\mathrm{LiClO_4}$  in ethylene carbonate/diethyl carbonate (EC/DEC 1/1) as liquid electrolyte. (a) (dis)charge of  $\\mathrm{LiCoO_2}$ , (b) the derivative of the storage capacity with respect to the electrode potential and (c) the electrochemical storage capacity upon cycling showing data for  $x = 2$  and  $x = 4$ .","id":"train/atomic-layer-deposition/experimental-usecase/52/figure_9","sample_id":"atomic-layer-deposition/experimental-usecase/52/figure_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multi-axis chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Constant current charge and discharge curves of an ALD LiCoO2 film (x = 4) between 3.0 V and 4.1 V versus Li/Li+, plotted as E versus Q. The paper notes an about 12% capacity loss between charge and discharge in this panel.\"},{\"panel_id\":\"b\",\"text\":\"Differential capacity dQ/dE versus E for the same cycling window, showing a clear charge transfer feature near 3.9 V that the paper associates with LiCoO2.\"},{\"panel_id\":\"c\",\"text\":\"Capacity versus cycle number comparing x = 2 and x = 4 films, showing higher capacity for x = 4 and fairly stable cycling behavior over the shown cycles, with areal capacity on the left axis and gravimetric capacity on the right axis.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Q, mAh/g | E (V) vs Li/Li⁺ (Charge) | E (V) vs Li/Li⁺ (Discharge) |\\n|---|---|---|\\n| 5 | 3.20 | 3.00 |\\n| 10 | 3.55 | 3.20 |\\n| 20 | 3.70 | 3.50 |\\n| 40 | 3.85 | 3.70 |\\n| 60 | 3.92 | 3.85 |\\n| 80 | 4.00 | 3.92 |\\n| 95 | 4.15 | 4.00 |\"},{\"panel_id\":\"b\",\"text\":\"| E (V) vs Li/Li⁺ | dQ/dE (mAh/Vg) Charge | dQ/dE (mAh/Vg) Discharge |\\n|---|---|---|\\n| 3.2 | 0 | 0 |\\n| 3.4 | 0 | 0 |\\n| 3.6 | 0.05 | -0.05 |\\n| 3.8 | 0.10 | -0.10 |\\n| 3.9 | 0.35 | -0.40 |\\n| 3.95 | 0.80 | -0.75 |\\n| 4.0 | 0.40 | -0.30 |\\n| 4.05 | 0.20 | -0.15 |\"},{\"panel_id\":\"c\",\"text\":\"| (Dis)charge cycles | Capacity Co/Li (x)=4 (μAh μm⁻¹cm⁻²) | Capacity Co/Li (x)=4 (mAh/g) | Capacity Co/Li (x)=2 (μAh μm⁻¹cm⁻²) | Capacity Co/Li (x)=2 (mAh/g) |\\n|---|---|---|---|---|\\n| 1 | 26 | 95 | 11.5 | 42 |\\n| 2 | 25.5 | 93 | 11.5 | 42 |\\n| 3 | 25 | 92 | 11.5 | 42 |\\n| 5 | 25 | 91 | 11 | 40 |\\n| 7 | 24.5 | 90 | 11 | 40 |\\n| 10 | 24.5 | 89 | 10.5 | 38 |\\n| 12 | 24 | 88 | 10 | 37 |\\n| 14 | 24 | 88 | 10 | 37 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Voltage window: 3.0 V to 4.1 V versus Li/Li+\\n\\nConstant current cycling at 0.35 C rate\\n\\nLiquid electrolyte: LiClO4 in EC/DEC 1/1\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The y axis is dQ/dE (differential capacity) and the x axis is E versus Li/Li+.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A charge transfer reaction at about 3.9 V as expected for LiCoO2.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":18,"width":624,"height":466},{"panel_id":"b","x":8,"y":487,"width":601,"height":427},{"panel_id":"c","x":3,"y":956,"width":652,"height":464}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/images/figure_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/52/Donders et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"52","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":664,"height":1472,"image_format":"jpeg","image_sha256":"8d72921325c02f472d475bc813b27d5f639b96157d2fd6114f0ee7dca8703ba7","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_54_figure_10.jpg","caption":"Figure 10. (Color online) Cyclic voltammetric scans at  $1\\mathrm{mV / s}$  of  $60~\\mathrm{nm}$  thick TaN and TiN films showing the reaction of the films toward Li ions.","id":"train/atomic-layer-deposition/experimental-usecase/54/figure_10","sample_id":"atomic-layer-deposition/experimental-usecase/54/figure_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows a cyclic voltammetric scan for two different materials, TaN (H₂ plasma) and TiN (H₂-N₂ 30:4 plasma), performed at 1 mV/s.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Potential (V) | Current (µA/cm²) TaN | Current (µA/cm²) TiN |\\n|---|---|---|\\n| 0.0 | -9 | -1 |\\n| 0.5 | -6 | -0.8 |\\n| 1.0 |  0 | -0.5 |\\n| 1.5 | 9 | -0.3 |\\n| 2.0 | 4 | -0.2 |\\n| 2.5 |  0 | -0.1 |\\n| 3.0 | 0 | 0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1 mV/s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Both were 60 nm thick.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0 to 3 V.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The TaN films have a much higher current density than TiN, indicating a higher reactivity toward the Li-containing electrolyte.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":0,"width":660,"height":499}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/Knoops et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"54","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":666,"height":503,"image_format":"jpeg","image_sha256":"243c247a4e8cfc531551375547b598e4e318c8e61728a446dcb7e787e57fc199","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_54_figure_2.jpg","caption":"Figure 2. (Color online) The thickness of the TaN and TiN films as a function of number of ALD cycles. The thicknesses have been determined by in situ spectroscopic ellipsometry.","id":"train/atomic-layer-deposition/experimental-usecase/54/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/54/figure_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between the number of cycles and film thickness for TaN (H₂ plasma) and TiN (H₂-N₂ 30:4 plasma). Both materials exhibit a linear increase in film thickness with increasing cycle count.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of cycles | Film thickness (nm) TaN | Film thickness (nm) TiN |\\n|---|---|---|\\n| 0 | 0 | 0 |\\n| 50 | 2.5 | 2 |\\n| 100 | 5 | 4 |\\n| 150 | 7 | 6 |\\n| 200 | 10 | 8 |\\n| 250 | 12 | 10 |\\n| 300 | 14.5 | 12 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TaN and TiN.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 7.5 nm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Roughly 14.5 nm.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":0,"width":653,"height":516}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/Knoops et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"54","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":661,"height":520,"image_format":"jpeg","image_sha256":"ce2b6a506e061c8460b45cb7aeae3e6a89222ec66e860b2f4e1f637d4053d516","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_54_figure_3.jpg","caption":"Figure 3. (Color online) (a) The growth per cycle and (b) the FPP resistivity of TaN and TiN films as a function of the plasma exposure time in the ALD cycle. The lines serve as guides to the eyes.","id":"train/atomic-layer-deposition/experimental-usecase/54/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/54/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure depicts the growth rate of TiN and TaN films as a function of plasma exposure time, where the growth rate of TiN plateaus after 5 seconds, but the growth rate of TaN keeps increasing after 5 seconds.\"},{\"panel_id\":\"b\",\"text\":\"The figure depicts the FPP resistivity of the TiN and TaN films as a function of plasma exposure time. The resistivity of TiN is practically constant, but the resistivity of TaN decreases exponentially as a function of plasma exposure time.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Plasma exposure time (s) | TaN (H₂ plasma) | TiN (H₂-N₂ 30:4 plasma) |\\n|---|---|---|\\n| 0 | 0.02 | 0.03 |\\n| 5 | 0.03 | 0.04 |\\n| 10 | 0.04 | 0.04 |\\n| 20 | 0.05 | 0.04 |\\n| 30 | 0.06 | 0.04 |\"},{\"panel_id\":\"b\",\"text\":\"| Plasma exposure time (s) | TaN (H₂ plasma) | TiN (H₂-N₂ 30:4 plasma) |\\n|---|---|---|\\n| 0 | 3*10^4 | 2*10^2 |\\n| 5 | 4*10^3 | 2*10^2 |\\n| 10 | 10^3 | 2*10^2 |\\n| 30 | 4*10^2 | 10^2 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The growth rates for both species as a function of time increase rapidly during the first few seconds of plasma exposure. The growth rate of TiN quickly plateaus after 5 seconds of exposure, whilst the growth rate of TaN keeps increasing steadily, albeit significantly slower than the initial time exposure.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.04 nm/cycle.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3*10^4 μΩ cm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10^2 μΩ cm.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":-1,"width":675,"height":526},{"panel_id":"b","x":0,"y":535,"width":674,"height":516}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/Knoops et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"54","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":677,"height":1050,"image_format":"jpeg","image_sha256":"3a0e0bc595cf525066846799183053b1f093cf327441f8a31876d8672e901478","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_54_figure_5.jpg","caption":"Figure 5. (Color online) The sheet resistance at  $150^{\\circ}\\mathrm{C}$  of the TaN-Si, TiN-Si, Cu-TaN-Si, and Cu-TiN-Si stacks after heating to the anneal temperature indicated at the horizontal axis. The annealing temperature was increased in steps of  $50^{\\circ}\\mathrm{C}$  up to a maximum temperature of  $700^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/54/figure_5","sample_id":"atomic-layer-deposition/experimental-usecase/54/figure_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the sheet resistance at 150°C for different stack configurations (TaN-Si, TiN-Si, Cu-TaN-Si, Cu-TiN-Si) as a function of temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Sheet resistance (Ω/□) TaN-Si stack | Sheet resistance (Ω/□) TiN-Si stack | Sheet resistance (Ω/□) Cu-TaN-Si stack | Sheet resistance (Ω/□) Cu-TiN-Si stack |\\n|---|---|---|---|---|\\n| 150 | 60 | 10 | 0.2 | 0.1 |\\n| 200 | 60 | 10 | 0.2 | 0.1 |\\n| 250 | 60 | 10 | 0.2 | 0.1 |\\n| 300 | 60 | 10 | 0.2 | 0.1 |\\n| 350 | 60 | 10 | 0.2 | 0.1 |\\n| 400 | 60 | 10 | 0.2 | 0.1 |\\n| 450 | 50 | 10 | 0.2 | 0.1 |\\n| 500 | 30 | 9 | 0.2 | 0.1 |\\n| 550 | 20 | 9 | 0.2 | 0.1 |\\n| 600 | 30 | 9 | 0.2 | 0.1 |\\n| 650 | 30 | 8 | 1 | 3 |\\n| 700 | 20 | 7 | 4 |  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 10 Ω/\\u0001.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 60 Ω/\\u0001.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 0.2 Ω/\\u0001.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure shows that the TaN–Si stack exhibits a noticeable reduction in sheet resistance once the annealing temperature exceeds roughly 400 °C, indicating an annealing-induced improvement in film quality and demonstrating that TaN maintains thermal stability well above its 225 °C deposition temperature. The TiN–Si stack shows a similar but weaker decrease in resistance, consistent with its higher deposition temperature of 400 °C and thus less room for annealing-driven improvements. In contrast, the Cu–TaN–Si and Cu–TiN–Si stacks initially display very low sheet resistances (approximately 0.2 and 0.1 in the units indicated on the plot), dominated by the highly conductive Cu layer. However, once annealed above about 600 °C, both Cu-containing stacks undergo a dramatic increase in sheet resistance. This sudden rise signals failure of the diffusion barrier; in other words, Cu diffuses into the underlying layers at high temperature, degrading the stack and destroying the low-resistance pathway.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":1,"width":671,"height":516}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/54/Knoops et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"54","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":675,"height":519,"image_format":"jpeg","image_sha256":"29a9f3a2f7536e26eb5de2f3b3f0cd2b26102f8d9204d573d085930ac18925d3","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_57_figure_3.jpg","caption":"Figure 3. (a) Transmittance spectra of a  $100\\mathrm{nm}$  CsI film on sapphire and bare sapphire substrate. (b) Tauc plot of the CsI film in panel (a). The CsI film was deposited at  $150^{\\circ}\\mathrm{C}$ , pulse durations were 1.5 and  $0.5\\mathrm{s}$  for Cs and I precursors, respectively, and  $1.0\\mathrm{s}$  purge durations.","id":"train/atomic-layer-deposition/experimental-usecase/57/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/57/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the optical characterization of a 100 nm ALD-grown CsI film. The transmittance spectrum indicates that the CsI film (blue) is highly transparent (>80%) across the visible and near-infrared range (300–1100 nm), closely matching the bare sapphire substrate (black), confirming high optical quality with minimal scattering or defect-related absorption. A sharp decrease in transmittance near 225 nm marks the onset of strong fundamental absorption.\"},{\"panel_id\":\"b\",\"text\":\"The Tauc plot (𝛼ℎ𝜈)^2 vs. photon energy shows a linear region, whose extrapolation intersects the energy axis at ~5.5 eV, confirming that CsI is a wide-bandgap insulator with a direct optical bandgap of ~5.5 eV.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Wavelength (nm) | Transmittance (%) | Material |\\n|----------------|-------------------|------|\\n| 200            | 79                 | CsI |\\n| 200            | 0                 | Sapphire |\\n| 300            | 84                | CsI |\\n| 300            | 81                | Sapphire |\\n| 400            | 84               | CsI |\\n| 400            | 84                | Sapphire |\\n| 500            | 85                | CsI |\\n| 500            | 85                | Sapphire |\\n| 600            | 85                | CsI |\\n| 600            | 85                | Sapphire |\\n| 700            | 85                | CsI |\\n| 700            | 85                | Sapphire |\\n| 800            | 85                | CsI |\\n| 800            | 85                | Sapphire |\\n| 900            | 85               | CsI |\\n| 900            | 85               | Sapphire |\"},{\"panel_id\":\"b\",\"text\":\"| hv (eV) | (αhv)^2 (a.u.) |\\n|----------|-----------------|\\n| 5.0      | 0               |\\n| 5.25     | 0               |\\n| 5.5      | 1E+16           |\\n| 5.62      | 8E+16            |\\n| 5.75     | 3.7E+16           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Blue curve would drop significantly below the Black curve (lower transmittance). Impurities typically act as absorbing centers (darkening the film) or scattering sites. If the film were contaminated, it would appear brownish or cloudy rather than clear.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The critical absorption feature for CsI is at ~225 nm (5.5 eV).  Borosilicate glass is opaque in this UV region (cutoff ~300 nm). If used, the substrate itself would absorb all the UV light before the CsI band edge was reached. The spectrometer would measure the glass cutoff (4.0 eV) instead of the CsI bandgap (5.5 eV), yielding false data.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The x-intercept indicates a direct band gap of approximately 5.7 eV. This signifies that the CsI film is a wide-band-gap insulator, where a photon with at least 5.7 eV of energy is required to excite an electron from the valence band to the conduction band.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The high transmittance and sharp cutoff imply it is an excellent solar-blind VUV coating. It would be transparent to wavelengths above its band edge (~220 nm), blocking visible light, while potentially converting shorter VUV photons into detectable electrons via the photoelectric effect, enabling selective VUV sensing.\"}]}]","bbox":[{"panel_id":"a","x":16,"y":10,"width":474,"height":346},{"panel_id":"b","x":514,"y":14,"width":496,"height":342}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/Weia et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"57","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1011,"height":356,"image_format":"jpeg","image_sha256":"3e445cf0b31c6f7c70fe545d16b4b38481f4190e81189b03dbb6e38c2b563c22","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_57_figure_6.jpg","caption":"Figure 6. (a) Transmittance spectra of the samples made by depositing CsI on a sapphire wafer and then depositing  $\\mathrm{PbI}_2$  on CsI. Transmittance spectra of saphire and the as-deposited CsI film are shown for reference. Measurement were made from the center of the sample. (b) Digital photographs of selected samples from panel (a). CsI films were deposited at  $150^{\\circ}\\mathrm{C}$  with 300 cycles. The pulse durations were  $1.5~\\mathrm{s}$  for  $\\mathrm{Cs(btsa)}$  and  $1.0~\\mathrm{s}$  for  $\\mathrm{SnI_4}$  and purge durations were  $1.0~\\mathrm{s}$ . In the  $\\mathrm{PbI}_2$  deposition, the durations of the precursor pulses were 0.8 and  $2.0~\\mathrm{s}$  for  $\\mathrm{Pb(btsa)}_2$  and  $\\mathrm{SnI_4}$ , respectively, whereas purge durations were  $1.0~\\mathrm{s}$ . The diameter of the sapphire wafers was  $5.08~\\mathrm{cm}$ .","id":"train/atomic-layer-deposition/experimental-usecase/57/figure_6","sample_id":"atomic-layer-deposition/experimental-usecase/57/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"apparatus diagram"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure illustrates the solid-state reaction between CsI and PbI₂ to form perovskite CsPbI₃ as the number of PbI₂ ALD cycles increases. The CsI film is transparent in the visible region with an absorption edge near 225 nm, while PbI₂ shows an absorption edge at approximately 520 nm. With 2000–3000 PbI₂ cycles (the absorption edge shifts to around 700 nm (1.7–1.8 eV), indicating successful formation of the perovskite phase γ-CsPbI₃.At higher doses (3500–4000 cycles), the perovskite absorption feature weakens and the spectrum increasingly resembles pure PbI₂, indicating complete consumption of CsI and accumulation of excess unreacted PbI₂ on the surface.\"},{\"panel_id\":\"b\",\"text\":\"The image panel shows microscopic images of different materials at various  PbI₂ cycle counts.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Wavelength (nm) | Transmittance (%) | Sample  |\\n|----------------:|------------------:|----------------------|\\n| 250 | 80 | Sapphire |\\n| 250 | 5  | CsI film |\\n| 250 | 0  | CsI + 2000 PbI₂ cycles |\\n| 250 | 0  | CsI + 2500 PbI₂ cycles |\\n| 250 | 0  | CsI + 3000 PbI₂ cycles |\\n| 250 | 0  | CsI + 3500 PbI₂ cycles |\\n| 250 | 0  | CsI + 4000 PbI₂ cycles |\\n| 250 | 10 | PbI₂ film |\\n| 300 | 83 | Sapphire |\\n| 300 | 75 | CsI film |\\n| 300 | 0  | CsI + 2000 PbI₂ cycles |\\n| 300 | 0  | CsI + 2500 PbI₂ cycles |\\n| 300 | 0  | CsI + 3000 PbI₂ cycles |\\n| 300 | 0  | CsI + 3500 PbI₂ cycles |\\n| 300 | 0  | CsI + 4000 PbI₂ cycles |\\n| 300 | 8  | PbI₂ film |\\n| 350 | 85 | Sapphire |\\n| 350 | 82 | CsI film |\\n| 350 | 0  | CsI + 2000 PbI₂ cycles |\\n| 350 | 0  | CsI + 2500 PbI₂ cycles |\\n| 350 | 0  | CsI + 3000 PbI₂ cycles |\\n| 350 | 0  | CsI + 3500 PbI₂ cycles |\\n| 350 | 5  | CsI + 4000 PbI₂ cycles |\\n| 350 | 6  | PbI₂ film |\\n| 400 | 86 | Sapphire |\\n| 400 | 85 | CsI film |\\n| 400 | 5  | CsI + 2000 PbI₂ cycles |\\n| 400 | 5  | CsI + 2500 PbI₂ cycles |\\n| 400 | 3  | CsI + 3000 PbI₂ cycles |\\n| 400 | 10 | CsI + 3500 PbI₂ cycles |\\n| 400 | 5  | CsI + 4000 PbI₂ cycles |\\n| 400 | 25 | PbI₂ film |\\n| 450 | 86 | Sapphire |\\n| 450 | 86 | CsI film |\\n| 450 | 20 | CsI + 2000 PbI₂ cycles |\\n| 450 | 10 | CsI + 2500 PbI₂ cycles |\\n| 450 | 8  | CsI + 3000 PbI₂ cycles |\\n| 450 | 30 | CsI + 3500 PbI₂ cycles |\\n| 450 | 40 | CsI + 4000 PbI₂ cycles |\\n| 450 | 50 | PbI₂ film |\\n| 500 | 87 | Sapphire |\\n| 500 | 87 | CsI film |\\n| 500 | 35 | CsI + 2000 PbI₂ cycles |\\n| 500 | 25 | CsI + 2500 PbI₂ cycles |\\n| 500 | 20 | CsI + 3000 PbI₂ cycles |\\n| 500 | 45 | CsI + 3500 PbI₂ cycles |\\n| 500 | 60 | CsI + 4000 PbI₂ cycles |\\n| 500 | 65 | PbI₂ film |\\n| 600 | 87 | Sapphire |\\n| 600 | 87 | CsI film |\\n| 600 | 60 | CsI + 2000 PbI₂ cycles |\\n| 600 | 55 | CsI + 2500 PbI₂ cycles |\\n| 600 | 50 | CsI + 3000 PbI₂ cycles |\\n| 600 | 65 | CsI + 3500 PbI₂ cycles |\\n| 600 | 80 | CsI + 4000 PbI₂ cycles |\\n| 600 | 78 | PbI₂ film |\\n| 700 | 88 | Sapphire |\\n| 700 | 88 | CsI film |\\n| 700 | 80 | CsI + 2000 PbI₂ cycles |\\n| 700 | 75 | CsI + 2500 PbI₂ cycles |\\n| 700 | 72 | CsI + 3000 PbI₂ cycles |\\n| 700 | 75 | CsI + 3500 PbI₂ cycles |\\n| 700 | 84 | CsI + 4000 PbI₂ cycles |\\n| 700 | 82 | PbI₂ film |\\n| 900 | 88 | Sapphire |\\n| 900 | 88 | CsI film |\\n| 900 | 87 | CsI + 2000 PbI₂ cycles |\\n| 900 | 85 | CsI + 2500 PbI₂ cycles |\\n| 900 | 83 | CsI + 3000 PbI₂ cycles |\\n| 900 | 85 | CsI + 3500 PbI₂ cycles |\\n| 900 | 88 | CsI + 4000 PbI₂ cycles |\\n| 900 | 86 | PbI₂ film |\"},{\"panel_id\":\"b\",\"text\":\"| | Sapphire | CsI & 3000 PbI2 cyc. | CsI & 3500 PbI2 cyc. | CsI & 4000 PbI2 cyc. |\\n|---|---|---|---|---|\\n| 1 | Sapphire | CsI & 3000 PbI2 cyc. | CsI & 3500 PbI2 cyc. | CsI & 4000 PbI2 cyc. |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The appearance of a strong absorption edge at ~700 nm (the film turns black). Pure CsI is transparent, so this darkening confirms the reaction to form CsPbI₃ .\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"3000 cycles (Black) is optimal. It has the strongest absorption. Going to 4000 cycles is worse because the excess PbI₂ layer blocks light (increasing transmittance to 80% at 600 nm), reducing the cell's efficiency.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Excess Lead Iodide ( PbI₂). The curve shape reverting to match the  PbI₂ reference indicates that all CsI was consumed and unreacted  PbI₂ is now accumulating on top.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Because it would be transparent to visible light. Without the Perovskite phase (which absorbs < 700 nm), the unreacted bilayer lets sunlight pass right through without generating any electricity.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":1,"width":512,"height":417},{"panel_id":"b","x":520,"y":3,"width":487,"height":413}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/57/Weia et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"57","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1008,"height":417,"image_format":"jpeg","image_sha256":"c88b1bfa1cd8c364f1c104bce40aea60149ef4c28fcb93f8b68344f6d1322840","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_59_fig_2.jpg","caption":"FIG. 2. (Color online) The imaginary part of the dielectric function  $\\epsilon_{2}$  of (a) a conductive, cubic  $\\mathrm{TaN}_{x,x\\leqslant 1}$  film and (b) a semiconductive  $\\mathrm{Ta}_3\\mathrm{N}_5$  film as deduced from spectroscopic ellipsometry measurements. The dielectric functions are parametrized using (a) a Drude and two Lorentz oscillators and (b) using two Tauc-Lorentz oscillators. The photon energy ranges from the near infrared  $(0.75 \\mathrm{eV})$  to the ultraviolet  $(6.5 \\mathrm{eV})$ . The conductive  $\\mathrm{TaN}_{x,x\\leqslant 1}$  film was deposited using  $10 \\mathrm{s} \\mathrm{H}_2$  plasma in the ALD cycle, while the  $\\mathrm{Ta}_3\\mathrm{N}_5$  film was deposited using  $10 \\mathrm{s} \\mathrm{NH}_3$  plasma.","id":"train/atomic-layer-deposition/experimental-usecase/59/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/59/fig_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the dielectric function ε₂ as a function of photon energy for different models: Drude Lorentz, Drude, Lorentz 1, and Lorentz 2 for the conductive cubic TaN film\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the dielectric function ε₂ as a function of photon energy for different models: Tauc-Lorentz, Tauc-Lorentz 1, and Tauc-Lorentz 2 for the semiconductive Ta3N5 film.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Photon energy (eV) | Drude Lorentz  | Drude | Lorentz 1 | Lorentz 2 |  \\n|---|---|---|---|---|\\n| 1 | 14 | 8 | 0 | 3 |\\n| 2 | 9 | 3 | 1 | 5 |\\n| 3 | 6 | 2 | 2 | 2 |\\n| 4 | 4 | 1 | 3 | 1 |\\n| 5 | 5 | 0.5 | 4 | 0.5 |\\n| 6 | 6 | 0 | 5.5 | 0 |\"},{\"panel_id\":\"b\",\"text\":\"| Photon energy (eV) | Tauc-Lorentz  | Tauc-Lorentz 1 | Tauc-Lorentz 2 |\\n|---|---|---|---|\\n| 1 | 0 | 0 | 0 | \\n| 2 | 0 | 0 | 0 | \\n| 3 | 1.5 | 1.5 | 0 |\\n| 4 | 6 | 5.5 | 0 |\\n| 5 | 5.8 | 4.5 | 2 |\\n| 6 | 5 | 2 | 3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The imaginary part of the dielectric function, ε₂.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A Drude and two Lorentz oscillators.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Photon energy varies from the near infrared to the ultraviolet.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Similar to the Tauc-Lorentz 1 model, the dielectric constant of the Tauc-Lorentz model increases until around 4 eV and then decreases. However, the Tauc-Lorentz 2 model shows an increase after 4 eV.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":599,"height":460},{"panel_id":"b","x":0,"y":474,"width":602,"height":461}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/Langereis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"59","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":602,"height":939,"image_format":"jpeg","image_sha256":"25aff3c027bc9a759355bad7af95b6327e06377b8100441ce1022d52532c3cff","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_59_fig_3.jpg","caption":"FIG. 3. (Color online) Film thickness as a function of number of remote plasma ALD cycles shown for the growth of cubic  $\\mathrm{TaN}_{x,x\\leqslant 1}$  (closed symbols) and  $\\mathrm{Ta}_3\\mathrm{N}_5$  (open symbols) as deduced from in situ spectroscopic ellipsometry measurements. The solid lines are linear fits to the data. The cubic  $\\mathrm{TaN}_{x,x\\leqslant 1}$  film was deposited using  $10\\mathrm{~s~}$  of  $\\mathrm{H}_{2}$  plasma in the ALD cycle, while the  $\\mathrm{Ta}_3\\mathrm{N}_5$  film was deposited using  $10\\mathrm{~s~}$  of  $\\mathrm{NH}_3$  plasma.","id":"train/atomic-layer-deposition/experimental-usecase/59/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/59/fig_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between the number of cycles and film thickness, indicating a linear increase for both TaN x,x⩽1 and Ta3N5.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of cycles | TaN | Ta3N5 |\\n|------------------|--------------------|--------------------|\\n| 0                | 0                  | 0                  |\\n| 100              | 5                  | 6                  |\\n| 200              | 10                 | 12                  |\\n| 300              | 15                 | 17                  |\\n| 400              | 20                 | 15                  |\\n| 500              | 25                 | 30                  |\\n| 600              | 30                 | 35                  |\\n| 700              | 35                 | 40                  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Measurements were taken for two films: cubic taN and Ta3N5.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"In accordance with an ALD process, the film thickness increases with the number of cycles for both samples.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ta3N5 film is thicker than TaN, because it shows higher film thickness values.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":593,"height":444}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/Langereis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"59","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":597,"height":450,"image_format":"jpeg","image_sha256":"91b6ac60b42ea5ca092f8c3b33b31c9e902f6b94e47af7c04b63130aa07be937","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_59_fig_7.jpg","caption":"FIG. 7. (Color online) The imaginary part of the dielectric function  $\\epsilon_{2}$  of  $\\mathrm{TaN}_x$  films deposited under various plasma conditions in the ALD cycle and as determined by in situ spectroscopic ellipsometry: (1)  $30\\mathrm{~s~H}_{2}$  plasma  $(x = 0.44)$ , (2)  $10\\mathrm{~s~H}_{2}$  plasma  $(x = 0.49)$ , (3)  $5\\mathrm{~s~H}_{2} - \\mathrm{N}_{2}$  (98:2) plasma  $(x = 1.0)$ , (4)  $10\\mathrm{~s~H}_{2} - \\mathrm{N}_{2}$  (1:1) plasma  $(x = 1.55)$ , and (5)  $10\\mathrm{~s~NH}_{3}$  plasma  $(x = 1.67)$ . The ratio  $x = [\\mathrm{N}] / [\\mathrm{Ta}]$  of the  $\\mathrm{TaN}_x$  films was determined by RBS analysis.","id":"train/atomic-layer-deposition/experimental-usecase/59/fig_7","sample_id":"atomic-layer-deposition/experimental-usecase/59/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart illustrates the dielectric function ε₂ as a function of photon energy (eV). Multiple curves represent different materials and conditions for the TiN ALD cycle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Photon energy (eV) | 1 | 2 | 3 | 4 | 5 |\\n| --- | --- | --- | --- | --- | --- |\\n| 1 | 30 | 15 | 9 | 1 | 0 |\\n| 2 | 13 | 10 | 9 | 1 | 0 |\\n| 3 | 8 | 5 | 5 | 8 | 1 |\\n| 4 | 7 | 5 | 5 | 10 | 7 |\\n| 5 | 6 | 5 | 5 | 9 | 7 |\\n| 6 | 6 | 6 | 5 | 7 | 6 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"In-situ spectroscopic ellipsometry.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The dielectric function of sample \\\"1\\\" decreases significantly as the photon energy increases.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Each film was deposited under different plasma conditions. These conditions could include different exposurre times or plasma gases.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Photon energies varies from the near infrared to the ultraviolet.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":603,"height":459}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/59/Langereis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"59","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":611,"height":461,"image_format":"jpeg","image_sha256":"84c644ae1f92412411fec023e8f875d9b33248b2ed2efcdaca594d40bee9c518","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_60_85ead34ddeac10d73af92922cd512497aa52e7e6f3abc2a8b4c0729255402d13.jpg","caption":"","id":"train/atomic-layer-deposition/experimental-usecase/60/85ead34ddeac10d73af92922cd512497aa52e7e6f3abc2a8b4c0729255402d13","sample_id":"atomic-layer-deposition/experimental-usecase/60/85ead34ddeac10d73af92922cd512497aa52e7e6f3abc2a8b4c0729255402d13","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":6,"y":1,"width":416,"height":379},{"panel_id":"b","x":419,"y":1,"width":483,"height":374}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/images/85ead34ddeac10d73af92922cd512497aa52e7e6f3abc2a8b4c0729255402d13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/images/85ead34ddeac10d73af92922cd512497aa52e7e6f3abc2a8b4c0729255402d13.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/Zhao et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"60","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"not_found"},"width":902,"height":380,"image_format":"jpeg","image_sha256":"84dfca5a33d60875551ab7efdd394295db503f9d54e5b1d9b429a9b9036d2bb8","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_60_figure_3.jpg","caption":"Figure 3. GPC as a function of substrate temperature.","id":"train/atomic-layer-deposition/experimental-usecase/60/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/60/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the temperature dependence of the atomic layer deposition growth per cycle (GPC) for Y₂O₃ films using Y(EtCp)₂(iPr-amd) with H₂O, O₂ plasma, and H₂O plasma reactants.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Substrate temperature (°C) | GPC (Å/cycle) | Precursor |\\n|---|---|---|\\n| 119 | 0.0 | Y(EtCp)2(iPr-amd) without reactant |\\n| 151 | 0.0 | Y(EtCp)2(iPr-amd) without reactant |\\n| 180 | 0.0 | Y(EtCp)2(iPr-amd) without reactant |\\n| 200 | 0.1 | Y(EtCp)2(iPr-amd) without reactant |\\n| 220 | 0.1 | Y(EtCp)2(iPr-amd) without reactant |\\n| 250 | 0.1 | Y(EtCp)2(iPr-amd) without reactant |\\n| 280 | 0.2 | Y(EtCp)2(iPr-amd) without reactant |\\n| 300 | 0.3 | Y(EtCp)2(iPr-amd) without reactant |\\n| 320 | 0.5 | Y(EtCp)2(iPr-amd) without reactant |\\n| 350 | 0.6 | Y(EtCp)2(iPr-amd) without reactant |\\n| 380 | 0.8 | Y(EtCp)2(iPr-amd) without reactant |\\n| 120 | 0.6 | H2O |\\n| 149 | 0.7 | H2O |\\n| 179 | 0.8 | H2O |\\n| 199 | 0.8 | H2O |\\n| 220 | 0.8 | H2O |\\n| 250 | 0.8 | H2O |\\n| 279 | 0.9 | H2O |\\n| 300 | 1.0 | H2O |\\n| 320 | 1.1 | H2O |\\n| 349 | 1.3 | H2O |\\n| 380 | 1.4 | H2O |\\n| 120 | 0.7 | O2 plasma |\\n| 149 | 0.8 | O2 plasma |\\n| 180 | 0.8 | O2 plasma |\\n| 200 | 0.9 | O2 plasma |\\n| 220 | 0.9 | O2 plasma |\\n| 251 | 0.9 | O2 plasma |\\n| 279 | 1.0 | O2 plasma |\\n| 298 | 1.2 | O2 plasma |\\n| 319 | 1.2 | O2 plasma |\\n| 349 | 1.4 | O2 plasma |\\n| 380 | 1.5 | O2 plasma |\\n| 120 | 0.8 | H2O plasma |\\n| 149 | 1.0 | H2O plasma |\\n| 180 | 1.0 | H2O plasma |\\n| 201 | 1.1 | H2O plasma |\\n| 220 | 1.1 | H2O plasma |\\n| 250 | 1.1 | H2O plasma |\\n| 280 | 1.2 | H2O plasma |\\n| 300 | 1.3 | H2O plasma |\\n| 320 | 1.4 | H2O plasma |\\n| 350 | 1.6 | H2O plasma |\\n| 381 | 1.7 | H2O plasma |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"150 - 300 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"150 - 300 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, there is clear evidence of thermal decomposition of the yttrium precursor above 300°C. The growth rate for the precursor without any reactant (gray line) begins to rise significantly above this temperature. this indicates self-deposition.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Above 300°C, GPC rises for all tested processes. The systematic increase is most likely caused by the  thermal decomposition of the yttrium precursor molecules. This  disrupts the self-limiting surface reactions required for ALD, leading to uncontrolled material deposition. This is confirmed by the significant growth observed when pulsing the yttrium precursor without any co-reactant.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":7,"width":463,"height":336}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/60/Zhao et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"60","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":470,"height":345,"image_format":"jpeg","image_sha256":"6fafaf3930cdc3a642b74f6fc1ef68fab4bec352adf08d63c71380cc1c851bc5","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_66_figure_4.jpg","caption":"Figure 4. Selected  $\\theta_{\\mathrm{Kerr}}$  vs field hysteresis loops for (a)  $\\mathrm{ZnO}$ ; (b) oxygen-deficient  $\\mathrm{ZnO}$ ; and (c)  $\\mathrm{Ni - }$ , (d)  $\\mathrm{Co - }$ , and (e) Fe-doped  $\\mathrm{ZnO}$  layers. The insets show example curves with large hysteresis and Kerr angle values.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_4","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A line chart showing the relationship between magnetic field strength and Kerr rotation angle for undoped ZnO.\"},{\"panel_id\":\"b\",\"text\":\"A line chart showing the relationship between magnetic field strength and Kerr rotation angle for oxygen-deficient ZnO.\"},{\"panel_id\":\"c\",\"text\":\"A line chart showing the relationship between magnetic field strength and Kerr rotation angle for  Ni doped ZnO\"},{\"panel_id\":\"d\",\"text\":\"A line chart showing the relationship between magnetic field strength and Kerr rotation angle for Co doiped ZnO.\"},{\"panel_id\":\"e\",\"text\":\"A line chart showing the relationship between magnetic field strength and Kerr rotation angle for Fe doped ZnO.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Magnetic field (Oe) | θ_Kerr (mdeg) |\\n|---|---|\\n|-4000 |-15 |\\n|-3000 |-10 |\\n|-2000 |-5 |\\n|-1000 |0 |\\n|0 |0 |\\n|1000 |5 |\\n|2000 |10 |\\n|3000 |15 |\\n|4000 |20 |\"},{\"panel_id\":\"b\",\"text\":\"| Magnetic field (Oe) | θ_Kerr (mdeg) |\\n|---|---|\\n|-4000 |-15 |\\n|-3000 |-10 |\\n|-2000 |-5 |\\n|-1000 |0 |\\n|0 |0 |\\n|1000 |5 |\\n|2000 |10 |\\n|3000 |15 |\\n|4000 |20 |\"},{\"panel_id\":\"c\",\"text\":\"| Magnetic field (Oe) | θ_Kerr (mdeg) |\\n|---|---|\\n|-4000 |-100 |\\n|-3000 |-80 |\\n|-2000 |-60 |\\n|-1000 |-40 |\\n|0 |0 |\\n|1000 |40 |\\n|2000 |60 |\\n|3000 |80 |\\n|4000 |100 |\"},{\"panel_id\":\"d\",\"text\":\"| Magnetic field (Oe) | θ_Kerr (mdeg) |\\n|---|---|\\n|-4000 |-60 |\\n|-3000 |-50 |\\n|-2000 |-40 |\\n|-1000 |-30 |\\n|0 |0 |\\n|1000 |30 |\\n|2000 |40 |\\n|3000 |50 |\\n|4000 |60 |\"},{\"panel_id\":\"e\",\"text\":\"| Magnetic field (Oe) | θ_Kerr (mdeg) |\\n|---|---|\\n|-4000 |-90 |\\n|-3000 |-70 |\\n|-2000 |-50 |\\n|-1000 |-30 |\\n|0 |0 |\\n|1000 |30 |\\n|2000 |50 |\\n|3000 |70 |\\n|4000 |90 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. For undoped ZnO, θ_Kerr varies from ~-15 mdeg to + 15 mdeg\\n2. For oxygen-deficient ZnO,  θ_Kerr varies from ~-15 mdeg to + 15 mdeg\\n3. For Ni-doped ZnO,  θ_Kerr varies from ~-100 mdeg to + 100 mdeg\\n4. For Co-doped ZnO,   θ_Kerr varies from ~-60 mdeg to + 60 mdeg\\n5. For Fe-doped ZnO,  θ_Kerr varies from ~-90 mdeg to + 90 mdeg\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Inhomogeneity is observed for both undoped and TM-doped ZnO layers. The observed complex micro MOKE (diffracted and asymmetrical) hysteresis loops, particularly for the TM-doped samples, reveal the variations and non-uniformity of the MO response. As is seen, these variations are significant in the case of TM-doped layers and for a certain fraction of the measurements shown in the insets with broad hysteresis contours, and θKerr can reach significant values at the maximum applied field.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"MOKE stands for magneto optic kerr effect,\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":518,"height":482},{"panel_id":"b","x":523,"y":1,"width":476,"height":504},{"panel_id":"c","x":6,"y":551,"width":500,"height":472},{"panel_id":"d","x":505,"y":555,"width":451,"height":468},{"panel_id":"e","x":958,"y":557,"width":445,"height":468}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/Paskaleva et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"66","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1406,"height":1027,"image_format":"jpeg","image_sha256":"6769e0edc553a67eaf73161f4a17bbce44bafd5f2b7d4c504401e3b9b5b87880","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_7_fig_1.jpg","caption":"Fig. 1. Mass sensor signal as a function of time recorded at temperatures of  $100$  and  $300^{\\circ}\\mathrm{C}$  during a single ALD cycle. Mass sensor was covered with a buffer layer deposited at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/7/fig_1","sample_id":"atomic-layer-deposition/experimental-usecase/7/fig_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Real-time mass sensor signal vs. time during a single ALD cycle at growth temperatures of 100C and 300C. The plot defines the specific mass changes: $\\\\Delta m_1$ (gain from TiCl_4 pulse), $\\\\Delta m_2$ (loss during purge/reaction), $\\\\Delta m_2'$ (initial desorption), and $\\\\Delta m_0$ (net mass increment per cycle).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Event | Signal 300°C (Top) | Signal 100°C (Bottom) |\\n|---|---|---|---|\\n| 10 | Start TiCl4 Pulse | ~0 | ~0 |\\n| 12 | End TiCl4 Rise (Δm1) | ~70 | ~60 |\\n| 20 | End TiCl4 Pulse | ~70 | ~60 |\\n| 25 | Start H2O Pulse | ~65 | ~55 |\\n| 26 | Drop (Δm2') | ~35 | ~35 |\\n| 30 | End H2O Pulse | ~35 | ~35 |\\n| 40 | Final Level (Δm0) | ~35 | ~35 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1.300°C\\n, 2. 100°C\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, the vertical step height labeled Δm1 is clearly larger in the top trace (300°C).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"During the TiCl₄ pulse, the top trace shows a sharp step increase labeled Δm₁, followed by a plateau while the pulse continues. Around the H₂O pulse, the signal decreases, with the drop region annotated by Δm₂ and Δm₂′. The trace then settles to a lower plateau and finishes the cycle at a baseline shifted upward from the start, labeled Δm₀. Overall, it’s a step-up (TiCl₄) then step-down (H₂O-related) response with a nonzero net offset.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiCl4\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":441,"height":541}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/Aarik et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"7","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":442,"height":547,"image_format":"jpeg","image_sha256":"3fa369bab6ab2fc3709cc00052557b6a571c305836e02eb7ba6bcfc5139e099a","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_11_fig4.jpg","caption":"FIG.4. Growth profiles for a  $200ms$  20 mTorr precursor dose for different values of the reaction probability  $\\beta_{10}$  Higher reaction probabilities lead to steeper growth profiles.","id":"train/atomic-layer-deposition/simulation-usecase/11/fig4","sample_id":"atomic-layer-deposition/simulation-usecase/11/fig4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the surface coverage of a material along the axial position. Different lines represent different reaction probabilities (β10).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Axial position (m) | Coverage (β10 = 0.0001) | Coverage (β10 = 0.0002) | Coverage (β10 = 0.0005 | Coverage (β10 = 0.001) | Coverage (β10 = 0.002) | Coverage (β10 = 0.005) | Coverage (β10 = 0.01) | Coverage (β10 = 0.02) |\\n|---|---|---|---|---|---|---|---|---|\\n| 0.0 | 0.1 | 0.15 | 0.3 | 0.5 | 0.75 | 0.95 | 1 | 1 | \\n| 0.1 | 0.1 | 0.15 | 0.3 | 0.45 | 0.6 | 0.85 | 1 | 1 |\\n| 0.2 | 0.1 | 0.15 | 0.3 | 0.4 | 0.5 | 0.6 | 0.75 | 0.9 |\\n| 0.3 | 0.1 | 0.15 | 0.25 | 0.35 | 0.4 | 0.4 | 0.2 | 0.175 |\\n| 0.4 | 0.1 | 0.15 | 0.25 | 0.3 | 0.3 | 0.2 | 0.1 | 0.05 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"200 ms.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"20 mTorr.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The reaction probability (β10).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Figure 4 shows that the reaction probability β₁₀ strongly controls how rapidly surface coverage develops along the reactor, but its effect depends on whether the process is transport-limited or reaction-limited. At relatively high reaction probabilities, decreasing β₁₀ reduces the steepness of the unsaturated growth profile, leading to a more gradual increase in surface coverage without a proportional decrease in the total mass uptake; in this regime, the deposited mass is primarily redistributed spatially within the reactor rather than reduced overall. As β₁₀ is lowered further, the system transitions into a reaction-limited regime, where surface reactions become the bottleneck and the surface coverage decreases directly in proportion to β₁₀.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":663,"height":491}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/Angel Yanguas-Gil et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":672,"height":492,"image_format":"jpeg","image_sha256":"7b61aa4aae946d9a835719b6bdad45f88ae0ad81d8ca3373df170ac93adb2a30","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_11_fig_10.jpg","caption":"FIG. 10. Reactor growth profiles for a soft-saturating self-limited model with two components (a)  $\\beta_{1b} = 10^{-2}$  and  $\\beta_{1b} = 10^{-3}$  and (b)  $\\beta_{1b} = 10^{-2}$  and  $\\beta_{1b} = 10^{-4}$ . All profiles correspond to the same unsaturated dose time of  $0.2 \\text{s}$  for an average precursor pressure at the inlet of 50 mTorr. (c) Surface coverage for increasing dose times using (b) conditions.","id":"train/atomic-layer-deposition/simulation-usecase/11/fig_10","sample_id":"atomic-layer-deposition/simulation-usecase/11/fig_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the reactor growth profiles for a soft-saturating self-limited model with two components, with β_1a = 10^-2 and β_1b = 10^-3.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the reactor growth profiles for a soft-saturating self-limited model with two components, with β_1a = 10^-2 and β_1b = 10^-4.\"},{\"panel_id\":\"c\",\"text\":\"The figure shows the reactor growth profiles with the same conditions as (b), but with four different dose times (t_d).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Axial position (m) | f₂ = 0 | f₂ = 0.1 | f₂ = 0.2 | f₂ = 0.4 | f₂ = 0.6 | f₂ = 0.8 | f₂ = 1 |\\n|---|---|---|---|---|---|---|---|\\n| 0.00 | 1.00 | 0.95 | 0.90 | 0.80 | 0.65 | 0.55 | 0.45 |\\n| 0.10 | 0.98 | 0.92 | 0.88 | 0.75 | 0.60 | 0.50 | 0.40 |\\n| 0.20 | 0.85 | 0.80 | 0.75 | 0.65 | 0.55 | 0.45 | 0.35 |\\n| 0.30 | 0.30 | 0.35 | 0.45 | 0.55 | 0.50 | 0.40 | 0.30 |\\n| 0.40 | 0.05 | 0.10 | 0.15 | 0.30 | 0.35 | 0.30 | 0.25 |\"},{\"panel_id\":\"b\",\"text\":\"| Axial position (m) | f₂ = 0 | f₂ = 0.1 | f₂ = 0.2 | f₂ = 0.4 | f₂ = 0.6 | f₂ = 0.8 | f₂ = 1 |\\n|---|---|---|---|---|---|---|---|\\n| 0.00 | 1.00 | 0.95 | 0.85 | 0.65 | 0.45 | 0.25 | 0.08 |\\n| 0.10 | 0.98 | 0.92 | 0.82 | 0.62 | 0.44 | 0.25 | 0.08 |\\n| 0.20 | 0.90 | 0.85 | 0.75 | 0.60 | 0.43 | 0.25 | 0.08 |\\n| 0.30 | 0.35 | 0.45 | 0.55 | 0.50 | 0.42 | 0.25 | 0.08 |\\n| 0.40 | 0.08 | 0.12 | 0.20 | 0.40 | 0.40 | 0.25 | 0.08 |\"},{\"panel_id\":\"c\",\"text\":\"| Axial position (m) | t_d = 0.2 s | t_d = 0.3 s | t_d = 0.5 s | t_d = 1.0 s |\\n|---|---|---|---|---|\\n| 0.00 | 0.90 | 0.92 | 0.93 | 0.94 |\\n| 0.10 | 0.88 | 0.91 | 0.92 | 0.94 |\\n| 0.20 | 0.75 | 0.90 | 0.92 | 0.94 |\\n| 0.30 | 0.40 | 0.88 | 0.92 | 0.94 |\\n| 0.40 | 0.12 | 0.70 | 0.92 | 0.94 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Roughly 0.9.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"50 mTorr.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.2 seconds.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Figure 10 shows that the presence and relative contribution of a slow self-limited reaction component can substantially alter growth profiles and lead to apparent saturation behaviour that is not fully representative of true surface saturation. When both fast and slow components have relatively high reaction probabilities, increasing the fraction of the slow component produces a gradual transition from a steep, step-like saturation profile to a more gently varying, nearly linear axial coverage profile. When the reaction probability of the slow component is reduced by two orders of magnitude, profiles near the reactor inlet can exhibit a plateau characteristic of saturated growth even though the absolute surface coverage remains well below full saturation. Increasing the dose time in this regime further accentuates this effect: beyond a certain dose duration, the process appears saturated, yet the slow component continues to evolve and yields a small but measurable increase in surface coverage.\"}]}]","bbox":[{"panel_id":"a","x":14,"y":28,"width":618,"height":395},{"panel_id":"b","x":14,"y":457,"width":618,"height":398},{"panel_id":"c","x":14,"y":886,"width":618,"height":436}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/Angel Yanguas-Gil et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":661,"height":1322,"image_format":"jpeg","image_sha256":"0fbfdb368d36306bb58667c6136ff196a83da887ed77018b4fa98ba574713ac1","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_11_fig_13.jpg","caption":"FIG. 13. Coverage profiles in the presence of competitive adsorption by reaction by-products. (a) Precursor coverage profile for a process characterized by  $\\beta_{10} = 10^{-2}$  and  $p_0 = 50$  mTorr. (b) By-products coverage profiles for the same conditions used in (a). Note how  $\\Theta +\\Theta_{bp} = 1$  indicating that the process is fully saturated. (c) Impact of precursor partial pressure for a process characterized by  $\\beta_{10} = 10^{-2}$  and  $\\beta_{bp0} = 10^{-3}$","id":"train/atomic-layer-deposition/simulation-usecase/11/fig_13","sample_id":"atomic-layer-deposition/simulation-usecase/11/fig_13","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the precursor coverage (Precursor cov.) against axial position (m) for different values of $\\\\beta_{bp}$.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the byproduct coverage (Byproduct cov.) against axial position (m) for different values of $\\\\beta_{bp}$.\"},{\"panel_id\":\"c\",\"text\":\"The figure shows the precursor coverage against the axial position for different values of precursor partial pressure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Axial position (m) | Precursor cov. $\\\\beta_{bp}$ = 0 | Precursor cov. $\\\\beta_{bp}$ = 0.0001 | Precursor cov. $\\\\beta_{bp}$ = 0.001 | Precursor cov. $\\\\beta_{bp}$ = 0.01 |\\n|---|---|---|---|---|\\n| 0.0 | 1.00 | 1.00 | 1.00 | 0.85 |\\n| 0.1 | 1.00 | 1.00 | 0.95 | 0.75 |\\n| 0.2 | 1.00 | 1.00 | 0.90 | 0.65 |\\n| 0.3 | 1.00 | 1.00 | 0.85 | 0.60 |\\n| 0.4 | 1.00 | 1.00 | 0.80 | 0.50 |\"},{\"panel_id\":\"b\",\"text\":\"| Axial position (m) | Byproduct cov. $\\\\beta_{bp}$ = 0 | Byproduct cov. $\\\\beta_{bp}$ = 0.0001 | Byproduct cov. $\\\\beta_{bp}$ = 0.001 | Byproduct cov. $\\\\beta_{bp}$ = 0.01 |\\n|---|---|---|---|---|\\n| 0.0 | 0.00 | 0.00 | 0.00 | 0.15 |\\n| 0.1 | 0.00 | 0.00 | 0.05 | 0.25 |\\n| 0.2 | 0.00 | 0.00 | 0.10 | 0.35 |\\n| 0.3 | 0.00 | 0.00 | 0.15 | 0.40 |\\n| 0.4 | 0.00 | 0.00 | 0.20 | 0.50 |\"},{\"panel_id\":\"c\",\"text\":\"| Axial position (m) | Precursor cov. 100 mTorr | Precursor cov. 50 mTorr | Precursor cov. 25 mTorr |\\n|---|---|---|---|\\n| 0.0 | 0.98 | 0.98 | 0.98 |\\n| 0.1 | 0.95 | 0.95 | 0.95 |\\n| 0.2 | 0.93 | 0.92 | 0.90 |\\n| 0.3 | 0.93 | 0.89 | 0.85 |\\n| 0.4 | 0.90 | 0.85 | 0.80 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Increasing the precursor pressure reduces the impact of by-product adsorption, leading to weaker axial coverage gradients.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"50 mTorr.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"β₁₀ = 10⁻².\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing the by-product reaction probability introduces progressively stronger axial gradients in the precursor surface coverage, even though the process is saturated, because adsorbed by-products occupy surface sites and reduce the available precursor coverage according to Θ = 1 − Θ_bp.\"}]}]","bbox":[{"panel_id":"a","x":10,"y":8,"width":662,"height":308},{"panel_id":"b","x":0,"y":352,"width":672,"height":297},{"panel_id":"c","x":18,"y":683,"width":653,"height":333}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/Angel Yanguas-Gil et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":672,"height":1017,"image_format":"jpeg","image_sha256":"1ee0347ca6012187741c91cb287ec8f7475469165b0538a3b52bbd279d23ead9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_15_fig_6.jpg","caption":"Fig. 6. Predicted energies for the second chlorine loss pathways via hydrolysis (R3a and R3b). The zero is the sum energy of the products of R1a/R1b and water.","id":"train/atomic-layer-deposition/simulation-usecase/15/fig_6","sample_id":"atomic-layer-deposition/simulation-usecase/15/fig_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure illustrates a potential energy diagram plotting the energy changes  (ΔE)  for two reaction pathways (R3a and R3b). The chart compares the pathways through their chemisorption states (CS),  transition states (TS) and intermediate physisorption states (PS), revealing that pathway R3a is energetically more favorable than R3b.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"|  Reaction coordinate, a.u.| Species Label | ΔE, kJ/mol | \\n| --- | --- | --- |\\n| 1 | CS(R3a) | -64.5 | \\n| 2 | TS(R3a) | 2.2 |\\n| 3 | PS(R3a) | -20.4 |\\n| 4  | Product(R3a) | -11.1 |\\n| 1 | CS(R3b) | -59.7 |\\n| 2 | TS(R3b) | 10.1 |\\n| 3 | PS(R3b) | -12.4 |\\n| 4 | Product(R3b) | 11.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"R3a.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Pathway R3a proceeds faster than pathway R3b. This conclusion is supported by the lower potential energy barrier found at the transition state (TS). The diagram shows that R3a has a barrier at only 2.2 kJ/mol, whereas R3b faces a higher barrier of 10.1 kJ/mol.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Product(R3a) is the more stable product.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Pathway R3b is endothermic.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":0,"width":643,"height":412}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/15/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/15/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/15/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/15/Chengxing Cui et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"15","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":650,"height":414,"image_format":"jpeg","image_sha256":"66ed04ec34fcf636f6d50bbae7366f20e88a3b9be05bb0e11e36878fe3825223","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_17_figure_14.jpg","caption":"Figure 14: (a) Blue line shows the total density profiles at 0 ALD cycles, red for four ALD cycles, and green for eight ALD cycles with respect to the Z-direction distance of the  $\\mathsf{Al}_2\\mathsf{O}_3$  film at  $300^{\\circ}C$  [186]. (b) The number density profiles for oxygen (blue), aluminium (green), and hydrogen (red) as a function of the Z-direction distance of  $\\mathsf{Al}_2\\mathsf{O}_3$  film at  $300^{\\circ}C$  [186].","id":"train/atomic-layer-deposition/simulation-usecase/17/figure_14","sample_id":"atomic-layer-deposition/simulation-usecase/17/figure_14","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line graph shows the density profiles of an Al₂O₃ film along its thickness (Z-direction) at three stages of ALD: 0 cycles (blue), 4 cycles (red), and 8 cycles (green), all at 300°C. The plot reveals how film density evolves with ALD cycles\"},{\"panel_id\":\"b\",\"text\":\"The graph shows the number density profiles (in atoms/nm³) for Oxygen (O - blue), Aluminum (Al - green), and Hydrogen (H - red) atoms across the thickness (Z-direction) of an Al₂O₃ film deposited at 300°C. The profiles reveal the layered atomic composition of the film: Oxygen and Aluminum densities peak in the main film region (~8–15 Å) where the Al₂O₃ matrix forms, while Hydrogen density is prominent at the film's interfaces, likely indicating surface hydroxyl groups (-OH) or precursor ligands.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z (Å) | L0 density (g/cm³) | L4 density (g/cm³) | L8 density (g/cm³) |\\n| --- | --- | --- | --- |\\n| 3.0 | 1.10 | 1.10 | 1.10 |\\n| 4.0 | 0.55 | 0.56 | 0.56 |\\n| 5.0 | 0.56 | 0.56 | 0.56 |\\n| 6.0 | 0.80 | 0.85 | 0.85 |\\n| 7.0 | 0.55 | 0.70 | 0.75 |\\n| 8.0 | 0.25 | 0.50 | 0.65 |\\n| 9.0 | 0.05 | 0.30 | 0.55 |\\n| 10.0 | 0.00 | 0.15 | 0.45 |\\n| 11.0 | 0.00 | 0.05 | 0.35 |\\n| 12.0 | 0.00 | 0.02 | 0.28 |\\n| 13.0 | 0.00 | 0.00 | 0.20 |\\n| 14.0 | 0.00 | 0.00 | 0.12 |\\n| 15.0 | 0.00 | 0.00 | 0.05 |\\n| 16.0 | 0.00 | 0.00 | 0.02 |\\n| 17.0 | 0.00 | 0.00 | 0.00 |\"},{\"panel_id\":\"b\",\"text\":\"| Z (Å) | O Number density (nm⁻³) | Al Number density (nm⁻³) | H Number density (nm⁻³) |\\n| --- | --- | --- | --- |\\n| 3.0 | 13.7 | 17.0 | 0.3 |\\n| 4.0 | 14.3 | 3.7 | 0.4 |\\n| 5.0 | 9.4 | 7.7 | 1.3 |\\n| 6.0 | 15.0 | 9.9 | 1.4 |\\n| 7.0 | 12.7 | 7.6 | 2.0 |\\n| 8.0 | 10.1 | 5.8 | 3.04 |\\n| 9.0 | 8.0 | 3.4 | 3.9 |\\n| 10.0 | 4.2 | 1.7 | 3.7 |\\n| 11.0 | 2.6 | 0.76 | 2.4 |\\n| 12.0 | 1.1 | 0.13 | 1.3 |\\n| 13.0 | 0.42 | 0.02 | 0.6 |\\n| 14.0 | 0.10 | 0.00 | 0.05 |\\n| 15.0 | 0.00 | 0.00 | 0.00 |\\n| 16.0 | 0.00 | 0.00 | 0.00 |\\n| 17.0 | 0.00 | 0.00 | 0.00 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"An oxygen-enriched surface region is indicated by the sharp O peak at the smallest Z, consistent with terminal oxygen or hydroxyl groups at the outer surface.​​\\n\\nA subsurface region where Al and O peaks overlap suggests a more stoichiometric Al₂O₃ network with coordinated Al–O polyhedra.​​\\n\\nA broader hydrogen profile peaking slightly beyond the main Al region implies surface and near-surface hydroxylation, with hydrogen bonded to oxygen at or just beneath the interface.​​\\n\\nThe decay of all three profiles to near zero at larger Z indicates that the film terminates within the plotted range, with negligible atomic population beyond about 13–14 Å.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Low bulk hydrogen indicates a pure Al₂O₃ lattice with minimal impurity-based charge traps. Hydrogen in oxides can create electronic defects that increase leakage current and reduce dielectric reliability. Therefore, its confinement to the interfaces suggests the functional bulk of the film has high electrical quality.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Initial substrate interface (0 cycles, blue).\\n\\n, Film nucleation and early growth (4 cycles, red).\\n\\n, Thickening and development of a uniform bulk region (8 cycles, green).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The central density is lower and less uniform at 4 cycles, but reaches a stable, high plateau (~1.0 g/cm³) at 8 cycles. The trade-off is between process speed (fewer cycles) and material quality (uniform, dense film). Achieving device-grade films requires more cycles to move past the non-uniform interfacial growth phase.\"}]}]","bbox":[{"panel_id":"b","x":673,"y":8,"width":658,"height":511},{"panel_id":"a","x":4,"y":8,"width":640,"height":509}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/David Sibanda et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1336,"height":520,"image_format":"jpeg","image_sha256":"defbfb81439cbf309103cf62738af71fbf66bd9447fefb392c80ffd5217f15c8","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_17_figure_21.jpg","caption":"Figure 21: Heat capacity values  $(C_{\\mathrm{v}})$  of various size FNPs [161].","id":"train/atomic-layer-deposition/simulation-usecase/17/figure_21","sample_id":"atomic-layer-deposition/simulation-usecase/17/figure_21","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multiple line chart represents the Specific heat capacity (Cv) against temperature for various iron nanoparticles (FNP) with the dimensions (2 nm, 2.5 nm, 3.0 nm, 3.5 nm, and 4.0 nm). Sharp peaks were observed at specific temperatures indicating melting points of the respective particles. This graph illustrates the melting point depression phenomenon, where the melting temperature drops significantly (from 2100 K to 1525 K) as the nanoparticle diameter decreases from 4.0 nm to 2.0 nm. It confirms that smaller particles are thermodynamically less stable and undergo phase transitions at lower thermal energies compared to larger ones.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (K) | 2 nm | 2.5 nm | 3.0 nm | 3.5 nm | 4.0 nm |\\n|---|---|---|---|---|---|\\n| 200 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 400 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 600 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 800 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1000 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1200 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1400 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1525 | 0.4| 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1600 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1775 | 0.0 | 0.8 | 0.0 | 0.0 | 0.0 |\\n| 1800 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 1925 | 0.0 | 0.0 | 0.9 | 0.0 | 0.0 |\\n| 2000 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 2050 | 0.0 | 0.0 | 0.0 | 0.6 | 0.0 |\\n| 2100 | 0.0 | 0.0 | 0.0 | 0.0 | 1.6 |\\n| 2200 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 2400 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies the catalyst must operate at a significantly lower temperature than expected for bulk iron to prevent deactivation. If the process temperature approaches the depressed melting point of the nanoparticles, they will become mobile and rapidly sinter, destroying the high-surface-area catalyst structure. The operational window is constrained by this size-dependent thermal instability.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Larger nanoparticles have a higher melting point. The sharp Cv peak shifts to higher temperatures as diameter increases from 2 nm to 4 nm. The cause is the surface-to-volume ratio: smaller particles have a larger fraction of surface atoms with lower coordination and higher energy, making the entire lattice less stable and melting at a lower temperature.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"2 nm\\n, 2.5 nm\\n, 3.0 nm\\n, 3.5 nm\\n, 4.0 nm\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"For computational efficiency in screening, the potential energy method is often preferred. It requires fewer data points to identify the trend break and avoids the extra step of calculating the derivative, saving time for each simulation run.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":4,"width":636,"height":489}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_21.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/images/figure_21.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/17/David Sibanda et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":644,"height":494,"image_format":"jpeg","image_sha256":"c1aeeef03300ca4011c69c4c0f67d75c06df2274ceca9b2d33e7841d56d874ec","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_2_fig_4.jpg","caption":"Fig. 4. Plot of the changes in selected bond lengths and distances (Hf-Cl, Hf-O, O-H and Cl-H) along the calculated IRC for the elimination of HCl from  $\\mathrm{Si(OH)_3 - OH - HfCl_4}$  complex. The adsorbed complex is on the left, products are on the right, and the TS is set to  $S = 0$  amu $^{1 / 2}$  bohr.","id":"train/atomic-layer-deposition/simulation-usecase/2/fig_4","sample_id":"atomic-layer-deposition/simulation-usecase/2/fig_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows bond distances (Å) against S((amu)^{1/2}bohr), with four distinct bond types: Hf-Cl, Hf-O, O-H, and Cl-H. Hf-Cl shows a slight increase as the S value increases, while H-O and Cl-H decrease. On the other hand, O-H increases above an S value of -0.5.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| S((amu)^{1/2}bohr) | Hf-Cl | Hf-O | O-H | Cl-H |\\n|---|---|---|---|---|\\n|-2.5 | 2.7 | 2.1 | 1.0 | 2.0 |\\n|-2.0 | 2.7 | 2.1 | 1.0 | 1.9 |\\n|-1.5 | 2.7 | 2.1 | 1.0 | 1.9 |\\n|-1.0 | 2.7 | 2.1 | 1.0 | 1.8 |\\n|-0.5 | 2.8 | 2.1 | 1.1 | 1.7 |\\n|0.0 | 2.8 | 2.1 | 1.4 | 1.5 |\\n|0.5 | 2.8 | 2.1 | 1.6 | 1.4 |\\n|1.0 | 2.8 | 2.0 | 1.7 | 1.3 |\\n|1.5 | 2.8 | 2.0 | 1.9 | 1.3 |\\n|2.0 | 2.8 | 2.0 | 2.0 | 1.3 |\\n|2.5 | 2.9 | 2.0 | 2.1 | 1.3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Changes in bond lenghts are closely related to the transition state (TS).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the S value increases from -2.5 to 2.5, the bond distance decreases for Hf-O and Cl-H, and increases for Hf-Cl and O-H.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Hf-Cl bond has the highest bond distance values close to 2.7 Å.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":657,"height":506}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/2/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/2/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/2/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/2/Jie Ren et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"2","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":661,"height":511,"image_format":"jpeg","image_sha256":"76d755fe27bd88785d2ef9448a9c1eb783e89ddb3e66242a8cdeea5acfe19f1d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_25_figure_4.jpg","caption":"Figure 4. ALD simulations. (A) Height of film growth per cycle and (B) roughness per cycle. Present results compared to Hu et al. The range of growth per cycle commonly observed experimentally is highlighted in yellow.","id":"train/atomic-layer-deposition/simulation-usecase/25/figure_4","sample_id":"atomic-layer-deposition/simulation-usecase/25/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the film growth per ALD cycle in Å for two studies,  comparing the present study simulation (blue) with the results from Hu et al (red) over 12 cycles. The yellow band marks the experimentally reported growth range (~0.9–1.1 Å/cycle).\"},{\"panel_id\":\"b\",\"text\":\"The line chart displays the surface roughness  in Å vs. cycle number for two studies, 'Present Study' and 'Hu et al.', over 12 cycles. The roughness from the present study is less rough.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle | Growth per cycle, Å : Present Study | Growth per cycle, Å: Hu et al. |\\n| --- | --- | --- | \\n| 1 | 1.04 | 0.49 |\\n| 2 | 0.88 | 0.48 |\\n| 3 | 0.83 | 0.54 |\\n| 4 | 0.89 | 0.38 |\\n| 5 | 1.06 | 0.48 |\\n| 6 | 0.76 | 0.59 |\\n| 7 | 1.17 | 0.68 |\\n| 8 | 1.17 | 0.46 |\\n| 9 | 1.00 | 0.49 |\\n| 10 | 1.07 | 0.46 |\\n| 11 | 1.19 | 0.57 |\\n| 12 | 0.97 | 0.45 |\"},{\"panel_id\":\"b\",\"text\":\"| Cycle | Surface Roughness, Å: Present Study | Surface Roughness, Å: Hu et al. |\\n| --- | --- | --- | \\n| 1 | 0.78 | 1.4 |\\n| 2 | 1.13 | 1.5 |\\n| 3 | 1.28 | 1.56; 1.66 |\\n| 4 | 1.31 | 1.73 |\\n| 5 | 1.44 | 1.80 |\\n| 6 | 1.33 | 2.0 |\\n| 7 | 1.35 | 2.0 |\\n| 8 | 1.42 | 1.89 |\\n| 9 | 1.44 | 1.89 |\\n| 10 | 1.36 | 1.70 |\\n| 11 | 1.51 | 1.68 |\\n| 12 | 1.58 | 1.65 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Present study (the values are better aligned with experimental ones, shown as yellow band).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The present study shows higher cycle-to-cycle variability. You can see it from the spread of values: for the present study it ranges from 0.76 Å/cycle (cycle 6) up to 1.19 Å/cycle (cycle 11), a spread of about 0.43 Å/cycle. Hu et al. ranges from 0.38 Å/cycle (cycle 4) to 0.68 Å/cycle (cycle 7), a smaller spread of about 0.30 Å/cycle. From one cycle to the next, the present study has a big jump from 0.76 (cycle 6) to 1.17 (cycle 7) (≈ +0.41), whereas the largest step change in Hu et al. is smaller (e.g., 0.68 → 0.46, ≈- 0.22).  So overall, the values fluctuate more in the present study.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Present study (the values for roughness are generally lower than in the study of Hu et al.).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They converge overall (the gap between the two data series decreases).\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":565,"height":434},{"panel_id":"b","x":572,"y":2,"width":549,"height":434}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/Kristopher S. Brown et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1125,"height":439,"image_format":"jpeg","image_sha256":"7f28e01c0cf2e544c4216dcdcb105d531da2545ec697af84729cbcb2ecc2fa54","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_25_figure_6.jpg","caption":"Figure 6. Predicted and measured alumina ALD coating growths on an alumina (blue) and silica (red) surface. Predictions are depicted by squares and dotted lines. Measurements of Al added on silica $^{47}$  are depicted by red circles and solid lines. Experimental values for Al added over alumina (blue circles and solid lines) inferred from Puurunen's empirical model, $^{45,46}$  which related the number of Al atoms added to surface OH coverage.","id":"train/atomic-layer-deposition/simulation-usecase/25/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/25/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the number of Al atoms added per unit area (Al/nm²) over cycles for both Al<sub>2</sub>O<sub>3</sub> experiment and model, and SiO<sub>2</sub> experiment and model. The Al<sub>2</sub>O<sub>3</sub> experiment and model exhibit a more similar trend, while the SiO<sub>2</sub> experiment and model deviate more from each other.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle | Al<sub>2</sub>O<sub>3</sub> Experiment | SiO<sub>2</sub> Experiment | Al<sub>2</sub>O<sub>3</sub> Model | SiO<sub>2</sub> Model |\\n|---|---|---|---|---|\\n| 1 | 2.5 | 4.5 | 2.8 | 2.4 |\\n| 2 | 3.2 | 2.4 | 3.4 | 2.4 |\\n| 3 | 3.4 | 2.7 | 3.2 | 2.1 |\\n| 4 | 3.2 |- | 3.4 | 2.4 |\\n| 5 | 3.6 | - | 3.9 | 2.6 |\\n| 6 | 3.3 | 2.9 | 3.7 | 2.5 |\\n| 7 | 3.3 | - | 3.7 | 2.7 |\\n| 8 | 3.4 | 3.2 | 4.1 | 2.9 |\\n| 9 | 3.4 | - | 3.9 | 3.0 |\\n| 10 | 3.3 | - | 3.9 | 3.1 |\\n| 11 | - | 3.8 | - | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Al<sub>2</sub>O<sub>3</sub> model generally overestimates the predicted values of added Al atoms.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO<sub>2</sub> model generally underestimates the predicted values of added Al atoms.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Considering only the experimental data from cycles 1–3, SiO₂ has the greater total Al atoms deposited per area. Even though its rate drops in cycles 2 and 3, the  initial deposition of 4.5 Al/nm² in Cycle 1 outweighs the steady but lower start of the Al₂O₃ surface. Summing those cycles gives SiO₂ = 4.5 + 2.4 + 2.7 = 9.6, while Al₂O₃ = 2.5 + 3.2 + 3.4 = 9.1.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":560,"height":448}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/Kristopher S. Brown et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":564,"height":452,"image_format":"jpeg","image_sha256":"f5a407a091c127d66cc06e86c5cb3ed5aded9b8b5e6a58cd66b97f46ad9482a9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_25_figure_7.jpg","caption":"Figure 7. CLD simulations. (A) Cumulative height growth and (B) roughness per cycle.","id":"train/atomic-layer-deposition/simulation-usecase/25/figure_7","sample_id":"atomic-layer-deposition/simulation-usecase/25/figure_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The  chart shows the growth per cycle for two methods, CLD (blue) and ALD (red), over ten cycles. CLD growth is less uniform than ALD under the conditions shown.\"},{\"panel_id\":\"b\",\"text\":\"The line chart displays the surface roughness for two methods, CLD and ALD, over cycles. CLD (blue) exhibits a progressive roughening trend, ALD (red) remains roughly constant .\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle | Growth per cycle, A: ALD | Growth per cycle, A: CLD |\\n|---|---|---|\\n| 1 | 1.0 | 0.2 |\\n| 2 | 0.9 | 0.9 |\\n| 3 | 0.8 | 1.5 |\\n| 4 | 0.9 | 0.2 |\\n| 5 | 1.1 | 1.0 |\\n| 6 | 0.8 | -0.8 |\\n| 7 | 1.1 | 0.5 |\\n| 8 | 1.1 | 4.7 |\\n| 9 | 0.9 | 0.4 |\\n| 10 | 1.1 | 1.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Cycle | Surface Roughness, A:  ALD| Surface Roughness, A: CLD|\\n|---|---|---|\\n| 0 | - | 0.7 |\\n| 1 | 0.8 | 1.2 |\\n| 2 | 1.1 | 2.2 |\\n| 3 | 1.2 | 3.5 |\\n| 4 | 1.3 | 3.8 |\\n| 5 | 1.5 | 5.0 |\\n| 6 | 1.3 | 3.7 |\\n| 7 | 1.2 | 4.2 |\\n| 8 | 1.4 | 8.5 |\\n| 9 | 1.4 | 7.3 |\\n| 10 | 1.3 | 8.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The CLD process eventually switches from deposition to etching mode. The growth per cycle for CLD drops below zero at cycle 6, indicating a negative growth rate. A negative value for growth per cycle physically means the removal of material from the surface rather than addition.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ALD (GPC remains roughly constant).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ALD (surface roughness is relatively constant).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ALD (surface roughness is lower).\"}]}]","bbox":[{"panel_id":"a","x":0,"y":1,"width":509,"height":322},{"panel_id":"b","x":519,"y":0,"width":472,"height":323}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/Kristopher S. Brown et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":995,"height":325,"image_format":"jpeg","image_sha256":"53961ccd7e23d3eeafbc5a5ac870874c0ee15308bdf7ade0fb1a4b0f59e2691a","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_25_figure_8.jpg","caption":"Figure 8. Effect of different injection stoichiometries. (A) Cumulative height growth per cycle and (B) roughness per cycle. Blue: base case. Red: half amount of water injected per cycle.","id":"train/atomic-layer-deposition/simulation-usecase/25/figure_8","sample_id":"atomic-layer-deposition/simulation-usecase/25/figure_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the growth per cycle over 10 cycles for two conditions: the base case and a process using half the H₂O injected per cycle. The base case remains low and fairly steady. In contrast, halving the H₂O dose produces highly unstable growth, with large cycle-to-cycle fluctuations.\"},{\"panel_id\":\"b\",\"text\":\"The chart illustrates the surface roughness over 10 cycles for two conditions: the base case and a process using half the H₂O injected per cycle. The base case remains relatively smooth, showing a slow increase in roughness over cycling. The half-H₂O condition exhibits significant increases in surface roughness.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle | Growth per cycle, Å (Base case) | Growth per cycle, Å (Half H₂O amount injected per cycle) |\\n|---|---|---|\\n| 1 | 0.98 | 3.07 |\\n| 2 | 1.22 | 5.64 |\\n| 3 | 1.96 | 2.57 |\\n| 4 | 0.98 | 18.77 |\\n| 5 | 1.59 | 25.15 |\\n| 6 | 0 | 3.31 |\\n| 7 | 1.22 | 47.8 |\\n| 8 | 5.27 | 9.44 |\\n| 9 | 0.85 | 3.55 |\\n| 10 | 1.47 | 32.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Cycle | Surface Roughness, Å (Base case) | Surface Roughness, Å (Half H₂O amount injected per cycle) |\\n|---|---|---|\\n| 0 | 1.17 | 1.17 |\\n| 1 | 1.76 | 4.5 |\\n| 2 | 2.35 | 10 |\\n| 3 | 4.11 | 12.35 |\\n| 4 | 4.51 | 29.01 |\\n| 5 | 5.49 | 38.23 |\\n| 6 | 3.72 | 43.13 |\\n| 7 | 4.51 | 67.84 |\\n| 8 | 8.63 | 64.90 |\\n| 9 | 7.45 | 67.25 |\\n| 10 | 8.43 | 72.15 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Halving the H₂O dose makes the growth much less stable from cycle to cycle. In the base case, the growth per cycle stays low and fairly consistent, with only small fluctuations across the 10 cycles. With half H₂O, the growth becomes unstable, showing large swings.  So the reduced H₂O dose breaks the steady, repeatable growth behavior.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"To obtain smooth films, the base-case  (stoichiometric) dosing  should be selected. The base case maintains a consistently low surface roughness (below 10 Å ) throughout the sequence. By contrast, the half-H₂O condition exhibits a rapid and substantial increase in roughness, reaching over 70 Å by the end of the measurement. Overall, this behavior is consistent with the development of a rough surface.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Using substoichiometric (insufficient) amounts of H₂O leads to surface roughening.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":1,"width":476,"height":345},{"panel_id":"b","x":497,"y":2,"width":487,"height":342}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/Kristopher S. Brown et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"25","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":986,"height":347,"image_format":"jpeg","image_sha256":"176d23f8cc67aee84c18387b5ab1e29898052e19146dbb289a09c646d2bfd722","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_38_figure_11.jpg","caption":"Figure 11. Coverage (\\%) versus rotation speed (rad/s) for (a) Hacac step, (b) BDEAS step, and (c) ozone step.  $x_{A}$ ,  $x_{B}$ , and  $x_{C}$  represents the mole fraction of Hacac, BDEAS, and ozone, respectively.","id":"train/atomic-layer-deposition/simulation-usecase/38/figure_11","sample_id":"atomic-layer-deposition/simulation-usecase/38/figure_11","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Hacac surface coverage (%) as a function of wafer rotation speed at different reactant mole fractions, showing a general decrease with increasing rotation speed. Coverage remains nearly full at low speeds even for lower mole fractions.\"},{\"panel_id\":\"b\",\"text\":\"BDEAS surface coverage (%) as a function of wafer rotation speed at different reactant mole fractions, showing sharper decreases at higher rotation speeds due to its noncompetitive kinetic mechanism, producing steeper and somewhat discontinuous curves.\"},{\"panel_id\":\"c\",\"text\":\"Ozone surface coverage (%) as a function of wafer rotation speed at different reactant mole fractions, generally decreasing with increasing rotation speed, with nearly full coverage at low speeds even for lower mole fractions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Rotation speed (rad/s)|xHacac = 0.05|xHacac = 0.1|xHacac = 0.3|\\n|-----------------------|-------------|------------|------------|\\n|0.2|100|100|100|\\n|0.3|98.8|99.5|100|\\n|0.4|97.5|98.7|100|\\n|0.5|96.2|97.8|99.5|\\n|0.6|95.1|96.5|100|\\n|0.7|94.0|95.2|99.8|\\n|0.8|92.8|94.0|99.5|\"},{\"panel_id\":\"b\",\"text\":\"| Rotation speed (rad/s)|xBDEAS = 0.05|xBDEAS = 0.1|xBDEAS = 0.3|\\n|-----------------------|-------------|------------|------------|\\n|0.2|100|100|100|\\n|0.3|97.5|98.5|100|\\n|0.4|95.0|97.0|99.2|\\n|0.5|92.0|95.0|98.0|\\n|0.6|89.5|93.0|97.5|\\n|0.7|87.0|91.5|97.0|\\n|0.8|85.0|90.0|96.5|\"},{\"panel_id\":\"c\",\"text\":\"| Rotation speed (rad/s)|xO3 = 0.05|xO3 = 0.1|xO3 = 0.3|\\n|-----------------------|----------|---------|---------|\\n|0.2|100|100|100|\\n|0.3|98.8|99.5|100|\\n|0.4|97.5|98.7|100|\\n|0.5|96.2|97.8|99.5|\\n|0.6|95.1|96.5|100|\\n|0.7|94.0|95.2|99.8|\\n|0.8|92.8|94.0|99.5|\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The coverage versus rotation speed data helps identify optimal operating conditions for the ASALD process. By selecting lower rotation speeds or higher reagent mole fractions, full surface coverage can be achieved for Hacac, BDEAS, and ozone steps. Understanding the steep coverage decrease for BDEAS allows engineers to adjust the mole fraction or rotation speed to avoid incomplete coverage. These insights ensure uniform thin film deposition, efficient reagent use, and controlled reaction kinetics, which are critical for producing high-quality films with predictable thickness and surface properties.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The BDEAS step shows the steepest decrease in coverage, attributed to its noncompetitive kinetic mechanism.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. Higher rotation speeds decrease the exposure time of the wafer to the reagents, resulting in lower surface coverage.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Wafer rotation speed (lower speeds increase coverage), Mole fraction of the reagent (higher mole fractions increase coverage), Kinetic mechanism of the reaction (e.g., noncompetitive behavior of BDEAS), Exposure time of the wafer to the reagent\"}]}]","bbox":[{"panel_id":"a","x":5,"y":9,"width":500,"height":388},{"panel_id":"b","x":550,"y":9,"width":507,"height":393},{"panel_id":"c","x":273,"y":480,"width":508,"height":382}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/38/Sungil Yun et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1058,"height":906,"image_format":"jpeg","image_sha256":"ff4736023581f073bdff3e17588470107b22883f6601c3ba0a35027261eb65ec","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_53_figure_11.jpg","caption":"Figure 11. Energetics for the bound reactants (BR), transition state (TS), bound products (BP) and unbound products (UP) relative to the unbound reactants (UR) for the reaction of  $\\mathrm{SiH}_2(\\mathrm{NH}_2)_2$  (triangles, dashed lines),  $\\mathrm{SiH}_2\\mathrm{DMA}_2$  (squares, solid lines) with the cluster models  $(\\mathrm{Si}_3\\mathrm{N}_4)_4(\\mathrm{NH}_3)_{12}$  (red) and  $(\\mathrm{Si}_3\\mathrm{N}_4)_4(\\mathrm{H}_2\\mathrm{O})_{12}$  (green). The table inset presents the  $E_{\\mathrm{act}}$  values for the precursors with the  $\\mathrm{NH}_2$  and OH covered surfaces.  $\\Delta E$  values are given in  $\\mathrm{kJ / mol}$  and were determined from BP86/SV(P) DFT calculations.","id":"train/atomic-layer-deposition/simulation-usecase/53/figure_11","sample_id":"atomic-layer-deposition/simulation-usecase/53/figure_11","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the energy changes ($\\\\Delta E$ in kJ/mol) for two compounds, SiH<sub>2</sub>(NH<sub>2</sub>)<sub>2</sub> and SiH<sub>2</sub>DMA<sub>2</sub>, across four stages: UR, BR, TS, BP, and UP. The chart also includes surface types (NH<sub>2</sub> and OH) for comparison.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Step | SiH₂(NH₂)₂ on NH₂ Surface | SiH₂DMA₂ on NH₂ Surface | SiH₂(NH₂)₂ on OH Surface | SiH₂DMA₂ on OH Surface |\\n|------|---------------------------|--------------------------|---------------------------|--------------------------|\\n| UR   | 0                         | 0                        | 0                         | 0                        |\\n| BR   | -10                     | -30                    | -65                     | -50                    |\\n| TS   | +70                     | +70                   | +10                    | 0                        |\\n| BP   | 20                       | 20                      | -30                    | -80                    |\\n| UP   | 5                         | 5                         | -90                    | -50                      |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"The activation energies for the precursors are as follows\\n1. SiH₂(NH₂)₂  on NH₂ Surface =    98.29   kJ/mol\\n2.  SiH₂(NH₂)₂ on   OH Surface  = 79.13   kJ/mol\\n3. SiH₂DMA₂  on NH₂ Surface =    79.97   kJ/mol\\n4.  SiH₂DMA₂ on   OH Surface  = 51.74   kJ/mol\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The larger steric interactions experienced by SiH2DMA2 with the amine-covered surface compared with the hydroxyl-covered surface destabilize BR. The same steric interactions explain the weaker adsorption by SiH2DMA2 relative to SiH2(NH2)2 on both substrates\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The higher energy of BR reflects a reduced bond strength of the precursor to the surface. This may increase the probability of the precursor returning unreacted to the gas phase at ALD temperatures and may prevent the remaining reaction steps occurring. This initial BR step is therefore crucial in determining the deposition rate of a\\nprecursor and was chosen as a metric for ALD reactivity of other potential precursors.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Overall, SiH₂(NH₂)₂  and SiH₂DMA₂ on NH₂ Surface have higher energy for all the states compared to OH surface\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":664,"height":492}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/Ciaran A. Murray et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"53","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":670,"height":497,"image_format":"jpeg","image_sha256":"f3508ff53dc22c869775ead69ef7bd79099d9c9bf2f24f14eb43883fcfa8a8cc","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_53_figure_6.jpg","caption":"Figure 6.  $\\Delta E$  for the hydrolysis  $\\mathrm{(YH = H_2O)}$  and amination  $\\mathrm{(YH =}$ $\\mathrm{NH_3}$  ) of various silicon precursors calculated using eqs 1a and 1b from BP86/TZVPP total energies.","id":"train/atomic-layer-deposition/simulation-usecase/53/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/53/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure plots the change in energy (ΔE) in kJ/mol for various compounds when interacting with H2O and NH3.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Precursor       | ΔE (kJ/mol) with H₂O | ΔE (kJ/mol) with NH₃ |\\n|------------------|------------------------|------------------------|\\n| SiH₂Cl₂         | 0                      | +45                   |\\n| SiH₂DMA₂        | -30                    | 0                     |\\n| SiH₂HFMA₂       | -60                    | -20                   |\\n| SiH₂EMA₂        | -50                    | 0                     |\\n| BDEAS           | -55                    | -2                    |\\n| BTBAS           | -60                    | -10                   |\\n| SiH₂CH₃         | -110                   | -70                   |\\n| SiH₄            | -80                    | -30                   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ΔEhyd was significantly more negative than ΔEamin by 47.14 kJ/mol. This value |ΔE| corresponds to the reaction SiH2(NH2)2 + 2H2O ↔ SiH2(OH)2 + 2NH3).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The difference ΔE between hydration and amination means that functional group elimination from difunctionalized silane precursors is thermodynamically more favorable on OH covered SiO2 surfaces than on NH2/NH-covered Si3N4 surfaces. This appears to correlate with the much slower ALD growth rates for silicon nitride compared with silicon dioxide.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. SiH₂Cl₂       \\n2. SiH₂DMA₂      \\n3.  SiH₂HFMA₂    \\n4. SiH₂EMA₂       \\n5. BDEAS       \\n6.  BTBAS        \\n7. SiH₂CH₃      \\n8.  SiH₄\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":1,"y":0,"width":657,"height":500}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/Ciaran A. Murray et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"53","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":661,"height":502,"image_format":"jpeg","image_sha256":"6b6043be3c97e6ceb854c5dfac75a54805d0742486eabe502b80192f78bfe65e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_53_figure_8.jpg","caption":"Figure 8. Energetics for the bound reactants (BR), transition state (TS), bound products (BP) and unbound products (UP) relative to the unbound reactants (UR) for the reaction of (a)  $\\mathrm{SiH}_2(\\mathrm{NH}_2)_2$  and (b)  $\\mathrm{SiH}_2\\mathrm{DMA}_2$  precursor with the surface group models  $\\mathrm{SiH}_3\\mathrm{-OH}$  (red diamonds),  $\\mathrm{SiH}_3\\mathrm{-NH}_2$  (green squares), and  $\\mathrm{SiH}_3\\mathrm{-NH - SiH}_3$  (blue triangles).  $\\Delta E$  values are given in  $\\mathrm{kJ / mol}$  and were determined using BP86/SV(P) DFT calculations.","id":"train/atomic-layer-deposition/simulation-usecase/53/figure_8","sample_id":"atomic-layer-deposition/simulation-usecase/53/figure_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart displays changes in ΔE (kJ/mol) for  SiH2(NH2)2 on surface of three different species (SiH3-OH, SiH3-NH2, SiH3-NH-SiH3) across five states (UR, BR, TS, BP, UP).\"},{\"panel_id\":\"b\",\"text\":\"The line chart shows changes in ΔE (kJ/mol) for  SiH2DMA2 precursor on surface of three different species (SiH3-OH, SiH3-NH2, SiH3-NH-SiH3) across five states (UR, BR, TS, BP, UP).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Step | SiH₃–OH | SiH₃–NH₂ | SiH₃–NH–SiH₃ |\\n|------|---------|----------|--------------|\\n| UR   | 0       | 0        | 0            |\\n| BR   | -40     | -20      | -15          |\\n| TS   | +80     | +60      | +70          |\\n| BP   | -60     | -30      | -25          |\\n| UP   | -80     | -10      | -5           |\"},{\"panel_id\":\"b\",\"text\":\"| Step | SiH₃–OH | SiH₃–NH₂ | SiH₃–NH–SiH₃ |\\n|------|---------|----------|--------------|\\n| UR   | 0       | 0        | 0            |\\n| BR   | -35     | -18      | -15          |\\n| TS   | +75     | +55      | +65          |\\n| BP   | -55     | -28      | -25          |\\n| UP   | -75     | -12      | -8           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The reactions leading to BR, BP and UP (optimized to local minima) on the SiH3−OH model surface were calculated to be more exothermic than those on the SiH3−NH2 and SiH3−NH−SiH3 systems, consistent with the results of the thermodynamic model\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"These calculations ignored temperature effects that may reduce the significance of these energy differences.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ΔE values for the SiH3−NH2 and SiH3−NH−SiH3 models were almost identical, indicating the similar chemistry of primary (NH2) and secondary (NH) amine\\ngroups. The transition state energy barriers or activation energies for SiH2(NH2)2 and SiH2DMA2 are quite similar to Eact for both precursors,  with the SiH3−OH substrate lower than with SiH3−NH2/SiH3−NH−SiH3 substrates.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":0,"width":370,"height":458},{"panel_id":"b","x":367,"y":0,"width":300,"height":456}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/Ciaran A. Murray et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"53","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":672,"height":458,"image_format":"jpeg","image_sha256":"03233352f6788668f7bf9ed6569b5e68946b530baaf5fa095e66b178171bebe6","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_10.jpg","caption":"Figure 10. The evolution of the growth rate during the  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD starting with two different initial  $-\\mathrm{OH}$  concentrations at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_10","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Growth per cycle is plotted against ALD cycles. Higher initial OH gives higher growth from the start. Lower OH shows slow early growth and gradual increase.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | Growth per cycle (Å), H: 6% | Growth per cycle (Å), H: 0.5% |\\n|---|---|---|\\n| 1 | ~0.50 | ~0.10 |\\n| 2 | ~0.50 | ~0.10 |\\n| 3 | ~0.55 | ~0.15 |\\n| 4 | ~0.40 | ~0.15 |\\n| 5 | ~0.50 | ~0.20 |\\n| 6 | ~0.60 | ~0.20 |\\n| 7 | ~0.70 | ~0.30 |\\n| 8 | ~0.46 | ~0.40 |\\n| 9 | ~0.50 | ~0.30 |\\n| 10 | ~0.46 | ~0.45 |\\n| 11 | ~0.56 | ~0.30 |\\n| 12 | ~0.46 | ~0.40 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Fewer OH groups limit precursor adsorption in early cycles. Reactive sites must form gradually through repeated ALD reactions. Thus, low initial OH lead to suppressed early-cycle growth.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The 6% OH condition.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It implies a higher density of reactive OH sites. More sites allow more precursor reactions per cycle, increasing growth efficiency.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure indicates that increasing or stabilizing initial OH coverage can reduce incubation effects and improve early-cycle growth uniformity. Process steps such as surface pre-treatment or controlled exposure conditions could be used to achieve this. Understanding the cycle-dependent fluctuations also helps in selecting suitable cycle counts to obtain consistent film thickness.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":6,"width":625,"height":558}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":633,"height":567,"image_format":"jpeg","image_sha256":"a08b54146b90f05f2ab9da5f67e697a503619e712774d06376f2104107a67b7f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_11.jpg","caption":"Figure 11. The evolution of the surface roughness during the  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD starting with two different initial  $-\\mathrm{OH}$  concentrations at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_11","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_11","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows surface roughness during Al₂O₃ ALD at 300 °C. High initial OH gives an early roughness peak and later smoothing. Low initial OH shows steady roughness increase with cycle number.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | Roughness (Å), H: 6% | Roughness (Å), H: 0.5% |\\n|---|---|---|\\n| 1 | ~1.40 | ~0.95 |\\n| 3 | ~1.55 | ~1.00 |\\n| 5 | ~1.80 | ~1.30 |\\n| 6 | ~2.00 | ~1.40 |\\n| 8 | ~1.90 | ~1.85 |\\n| 9 | ~1.88 | ~1.85 |\\n| 11 | ~1.70 | ~1.95 |\\n| 12 | ~1.65 | ~2.05 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"When starting from a high OH concentration, surface roughness increases rapidly during early ALD cycles due to active growth and then decreases as the surface becomes more uniformly covered. In contrast, a low OH concentration leads to a slower but more continuous buildup of roughness as reactive sites gradually form. This indicates that initial surface chemistry strongly influences how roughness develops over time.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The high-OH case shows a non-monotonic trend with a clear maximum in roughness at intermediate cycles followed by a decline. By comparison, the low-OH case shows a largely monotonic increase in roughness that continues into later cycles. These contrasting trends suggest different dominant growth and relaxation mechanisms depending on initial OH coverage.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The relationship implies that higher OH coverage initially promotes more uniform nucleation and surface reactions, eventually smoothing the surface as growth proceeds. Lower OH coverage delays uniform nucleation, leading to rougher and more heterogeneous surface structures over time. Thus, surface hydroxylation is closely tied to morphological evolution during ALD.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The trends suggest that higher initial OH concentrations can help limit long-term surface roughness by promoting more uniform growth after early cycles. Process strategies such as surface pre-treatment or controlled OH generation could therefore improve film smoothness. Monitoring roughness evolution also helps determine optimal cycle numbers to avoid roughness buildup at later stages.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":5,"width":620,"height":534}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":625,"height":542,"image_format":"jpeg","image_sha256":"78ee70080700c7a570b35d68c97d8fd9be3ab4419b0d3a7e783f1c8400d7fd25","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_12.jpg","caption":"Figure 12. The evolution of the surface  $-\\mathrm{OH}$  concentration during the ALD process at two different temperatures.","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_12","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_12","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows how surface −OH concentration increases during ALD. The increase is faster at 300 °C than at 150 °C. Higher temperature leads to much higher −OH coverage.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | –OH concentration (nm⁻²), 150 °C | –OH concentration (nm⁻²), 300 °C |\\n|---|---|---|\\n| 0 | 0.10 | ~0.10 |\\n| 1 | 0.18 | ~0.20 |\\n| 2 | 0.20 | ~0.25 |\\n| 3 | 0.30 | ~0.40 |\\n| 4 | 0.48 | ~0.60 |\\n| 5 | 0.60 | ~0.90 |\\n| 6 | 0.70 | ~1.10 |\\n| 7 | 0.95 | ~1.50 |\\n| 8 | 1.20 | ~1.80 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Cycle 1.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The strong temperature dependence indicates that surface hydroxylation is facilitated by thermally activated mechanisms such as ligand removal and reactive site formation. At 300 °C, these processes proceed more efficiently, creating a higher density of reactive –OH groups that can participate in subsequent ALD reactions. Thus, the evolving surface structure at elevated temperature enhances the reactivity and continuity of the ALD growth process.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.6 nm⁻².\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":638,"height":602}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":642,"height":609,"image_format":"jpeg","image_sha256":"56a10d61b3c9508f33c678f6119e03049873774035d463c4a2d8f96a94f0360d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_13.jpg","caption":"Figure 13. The evolution of the surface roughness during the ALD process at two different temperatures.","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_13","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_13","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows growth per cycle during ALD at 150 °C and 300 °C. Growth decreases during the first few cycles at both temperatures. After this, growth increases with cycle number. The increase is stronger at 300 °C than at 150 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | Growth per cycle (Å), 150 °C | Growth per cycle (Å), 300 °C |\\n|---|---|---|\\n| 1 | ~0.11 | ~0.08 |\\n| 2 | ~0.04 | ~0.07 |\\n| 3 | ~0.03 | ~0.15 |\\n| 4 | ~0.10 | ~0.13 |\\n| 5 | ~0.10 | ~0.18 |\\n| 6 | ~0.14 | ~0.17 |\\n| 7 | ~0.23 | ~0.30 |\\n| 8 | ~0.21 | ~0.38 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The incubation (or nucleation) stage.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around cycle 7.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"-Slower formation of reactive surface sites\\n-Reduced reaction kinetics at lower temperature\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure suggests that higher temperatures can be used to achieve faster growth once incubation effects are overcome, which may be beneficial for throughput-sensitive applications. Conversely, lower temperatures may be chosen when finer control over early growth is required. Understanding the trade-off between incubation behavior and steady-state growth helps optimize temperature selection for specific film quality and performance goals.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":5,"width":627,"height":513}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":633,"height":520,"image_format":"jpeg","image_sha256":"9638180a763924e0e914e26a392303ede66688acbceff86abca8e2544eb82067","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_3.jpg","caption":"Figure 3. Total density profiles after zero (blue), four (red) and eight (green) ALD cycles as a function of the distance along the  $Z$  direction of the  $\\mathrm{Al}_2\\mathrm{O}_3$  film at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_3","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents total density profiles along the Z direction of an Al₂O₃ film after 0, 4, and 8 ALD cycles. All three profiles show a high initial density near the substrate interface, followed by a rapid decrease and a secondary peak around 5–7 Å. With increasing ALD cycles, the density extends further into the film, producing broader and more gradual decay profiles. The L8 curve maintains higher density values at larger Z distances, indicating thicker and more fully developed film growth compared with L0 and L4.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z (Å) | Density L0 (g/cm³) | Density L4 (g/cm³) | Density L8 (g/cm³) |\\n|---|---|---|---|\\n| 3  | 1.15 | 1.15 | 1.15 |\\n| 4  | 0.58 | 0.58 | 0.58 |\\n| 5  | 0.65 | 0.63 | 0.61 |\\n| 6  | 0.80 | 0.87 | 0.85 |\\n| 7  | 0.56 | 0.70 | 0.70 |\\n| 8  | 0.25 | 0.58 | 0.68 |\\n| 9  | 0.10 | 0.38 | 0.58 |\\n| 10 | 0.03 | 0.20 | 0.50 |\\n| 11 | 0.00 | 0.10 | 0.40 |\\n| 12 | 0.00 | 0.05 | 0.30 |\\n| 13 | 0.00 | 0.00 | 0.18 |\\n| 14 | 0.00 | 0.00 | 0.10 |\\n| 15 | 0.00 | 0.00 | 0.00 |\\n| 17 | 0.00 | 0.00 | 0.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing the number of ALD cycles systematically shifts the density profiles outward along the Z direction and broadens their shapes. L0 shows the steepest decay, indicating a thinner initial film, whereas L4 and especially L8 retain significant density at larger Z values. The secondary peak present in all curves becomes more pronounced and moves slightly outward as the film grows. Overall, the progression from L0 to L8 reflects the accumulation of material and the development of a thicker, more homogeneous oxide layer.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A locally dense atomic layer or sublayer within the growing Al₂O₃ film.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A broader density distribution suggests a more extended and uniformly packed film, which may enhance barrier properties such as moisture resistance, dielectric strength, and mechanical stability. Greater density retention at larger Z distances indicates fewer voids or low-density regions, contributing to improved film continuity. These characteristics generally lead to better long-term performance in applications where conformal coverage and structural integrity are critical.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"- The steepness of the initial density decay, indicating early nucleation efficiency.\\n- The height and location of the secondary peak, reflecting sublayer formation.\\n- The extent of density at larger Z values, showing how far the film has grown.\\n- Differences between L0, L4, and L8 that reveal transition from initial nucleation to steady-state growth.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":614,"height":497}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":619,"height":500,"image_format":"jpeg","image_sha256":"e9d597c02aa42f1af5fdb12e95a58860d6cf89293f561d81b15ffb96324d3cbc","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_4.jpg","caption":"Figure 4. Number density profiles for O, Al and H as a function of the distance along the  $Z$  direction of the  $\\mathrm{Al}_2\\mathrm{O}_3$  film at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_4","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows number-density profiles of O, Al, and H atoms across the Z direction of an Al2O3 film at 300 °C. Oxygen exhibits the highest and most structured density distribution near the film surface, with prominent peaks indicating layered arrangements. Aluminum shows a similar but slightly shifted density profile with a lower overall magnitude. Hydrogen appears in much smaller concentrations and peaks closer to the mid-film region before tapering off. Together these profiles describe the atomic layering and composition variation within the film thickness.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z (Å) | O density (nm⁻³) | Al density (nm⁻³) | H density (nm⁻³) |\\n|----|----|----|----|\\n| 3  | 14.0 | 17.0 | 0.5 |\\n| 4  | 15.0 | 4.0  | 0.5 |\\n| 5  | 10.0 | 8.0  | 1.5 |\\n| 6  | 15.5| 10.0 | 1.5 |\\n| 7  | 13.0 | 8.0  | 2.0 |\\n| 8  | 10.0 | 6.0  | 3.0 |\\n| 9  | 8.0  | 4.0  | 4.0 |\\n| 10 | 4.0  | 2.0  | 4.0 |\\n| 11 | 3.0  | 1.0  | 3.0 |\\n| 12 | 2.0  | 0.0  | 2.0 |\\n| 13 | 0.5| 0.0  | 1.0 |\\n| 14 | 0.0  | 0.0  | 0.0 |\\n| 15 | 0.0  | 0.0  | 0.0 |\\n| 16 | 0.0  | 0.0  | 0.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 6 Å.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, after approximately Z = 11 Å, the H density remains positive while O drops toward zero.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"None; all densities have reached 0 nm⁻³ by this distance.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"- The relative heights and positions of the O and Al peaks, indicating layered ordering.\\n- The full width and decay behavior of the density tails as Z increases.\\n- The presence and magnitude of hydrogen density, as it reflects realistic surface hydroxylation or defect levels.\\n- The overall stoichiometric consistency inferred from integrated densities.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":614,"height":494}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":620,"height":497,"image_format":"jpeg","image_sha256":"bf21daf84048bc60fab6ead4ea4c3ebefe2f416ad1bc1eb97dda0f35b4de9dde","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_5.jpg","caption":"Figure 5.Ratio of Al[4]/Al[3] after zero (blue), four (red) and eight (green) ALD cycles as a function of the distance along the  $Z$  direction of the  $\\mathrm{Al}_2\\mathrm{O}_3$  film at  $300^{\\circ}\\mathrm{C}$","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the ratio of four-fold to three-fold coordinated aluminum (Al[4]/Al[3]) as a function of depth (Z direction) in an Al2O3 film after 0, 4, and 8 ALD cycles at 300 °C. The ratio is lowest and rapidly decays with depth for L0, indicating limited tetrahedral coordination. With increasing ALD cycles (L4 and L8), the Al[4]/Al[3] ratio increases significantly near the surface and exhibits pronounced oscillations, followed by a sharp decrease toward the film interior. This suggests cycle-dependent structural evolution of aluminum coordination across the film thickness.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z (Å) | Al[4]/Al[3] (L0) | Al[4]/Al[3] (L4) | Al[4]/Al[3] (L8) |\\n|---|---|---|---|\\n| 5 | 5.0 | 5.0 | 10 |\\n| 6 | 3.0 | 6.2 | 13 |\\n| 7 | 2.5 | 7.5 | 16 |\\n| 8 | 2.0 | 10.0 | 14.5 |\\n| 9 | 1.5 | 8.5 | 21 |\\n| 10 | 0.0 | 7.5 | 16 |\\n| 11 | 0.0 | 7.5 | 11.5 |\\n| 12 | - | 10.0 | 18.5 |\\n| 13 | — | 0.0 | 6.0 |\\n| 14–16 | — | — | 3 → ~0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Eight ALD cycles (L8).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"- Increased tetrahedral aluminum coordination\\n- Higher network connectivity\\n- Surface regions closer to bulk-like Al2O3 structure\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Depth-dependent coordination provides insight into how film structure develops during ALD and how surface reactions propagate inward. Variations in Al coordination affect film density, chemical stability, and dielectric behavior. Understanding these trends helps optimize ALD cycle numbers and processing conditions for applications requiring controlled structural and functional properties.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":5,"width":630,"height":541}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":637,"height":547,"image_format":"jpeg","image_sha256":"b9a4f191183b7b67300980b21366e05261a28ed9045e05f6a48be0ff27af0198","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_6.jpg","caption":"Figure 6.Ratio of O[3]/O[2] after zero (blue), four (red) and eight (green) ALD cycles as a function of the distance along the  $Z$  direction of the  $\\mathrm{Al}_2\\mathrm{O}_3$  film at  $300^{\\circ}\\mathrm{C}$","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the ratio of three-fold to two-fold coordinated oxygen (O[3]/O[2]) as a function of depth in an Al2O3 film after 0, 4, and 8 ALD cycles at 300 °C. For all cases, the ratio is highest near the surface and decreases rapidly toward the film interior. Increasing ALD cycles shift the O[3]/O[2] distribution to higher values and extend the depth range over which three-fold coordination persists, indicating progressive structural densification near the surface.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z (Å) | O[3]/O[2] (L0) | O[3]/O[2] (L4) | O[3]/O[2] (L8) |\\n|---|---|---|---|\\n| 5 | 1.0 | 1.20 | 1.00 |\\n| 6 | 1.1 | 1.55 | 1.90 |\\n| 7 | 1.0 | 1.70 | 2.10 |\\n| 8 | 0.7 | 1.20 | 1.65 |\\n| 9 | 0.35 | 0.75 | 1.60 |\\n| 10 | 0.2 | 0.80 | 1.30 |\\n| 11 | 0.0 | 0.30 | 1.00 |\\n| 12 | 0.0 | 0.20 | 0.80 |\\n| 13 | 0.0 | 0.00 | 0.50 |\\n| 14–16 | — | 0.00 | 0.1 → 0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Eight ALD cycles (L8).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"- Higher prevalence of three-fold coordinated oxygen\\n- Increased network connectivity\\n- More bulk-like oxide bonding configuration\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Oxygen coordination directly influences film density, mechanical stability, and dielectric properties of Al2O3. Variations in O[3]/O[2] with depth reveal how surface reactions evolve during ALD and how structural relaxation progresses into the film. Such insight is essential for designing ALD processes that produce uniform, high-quality oxide layers for electronic and protective applications.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":633,"height":547}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":639,"height":553,"image_format":"jpeg","image_sha256":"9ad86218129a364f324af35cd994404f7671dd5fc6017440b973c2bd8cf39998","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_55_figure_9.jpg","caption":"Figure 9. The evolution of the surface  $-\\mathrm{OH}$  concentration during the  $\\mathrm{Al}_2\\mathrm{O}_3$  ALD starting with two different initial  $-\\mathrm{OH}$  concentrations at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/55/figure_9","sample_id":"atomic-layer-deposition/simulation-usecase/55/figure_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the evolution of surface –OH concentration during Al2O3 ALD at 300 °C for two different initial –OH surface coverages. For the high initial –OH case (6.0%), the surface concentration increases rapidly during early cycles and then approaches a plateau near 5 nm⁻². In contrast, the low initial –OH case (0.5%) exhibits a slower but continuous increase throughout the ALD cycles, eventually reaching much lower values than the high-OH surface. This highlights the strong influence of initial surface chemistry on hydroxyl regeneration behavior.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | –OH concentration (nm⁻²), H: 6.0% | –OH concentration (nm⁻²), H: 0.5% |\\n|---|---|---|\\n| 1 | 2.8 | 0.2 |\\n| 3 | 3.8 | 0.4 |\\n| 5 | 4.0 | 0.9 |\\n| 7 | 4.6 | 1.4 |\\n| 8 | 4.9 | 1.8 |\\n| 9 | 4.9 | 2.0 |\\n| 10 | 4.9 | 2.6 |\\n| 12 | 5.1 | 3.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The surface starting with 6.0% initial –OH concentration.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"- Rapid increase during early ALD cycles\\n- Reduced slope at intermediate cycles\\n- Nearly constant –OH concentration at later cycles\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Surface –OH groups act as key reactive sites for precursor adsorption during ALD. Knowing how their concentration evolves helps in predicting growth rates, incubation behavior, and surface reactivity over multiple cycles. This information is critical for selecting surface pre-treatments and process conditions that lead to uniform and controlled film growth.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":5,"width":629,"height":555}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/images/figure_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/55/Zheng Hu et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"55","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":634,"height":562,"image_format":"jpeg","image_sha256":"adabac16ae58d7da622ad22875ab5d96f7bb59d475f7a0f7b083623f2f392ef1","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_8_figure_15.jpg","caption":"Figure 15. Fraction of lattice site vacant at the end of the simulation after  $2 \\mathrm{~s}$  of deposition (10 cycles). It is evident that the film growth is slow at low temperature (blue) because the lattice sites are occupied only up to a  $9 \\mathrm{\\AA}$  thickness. At slightly higher temperatures, the film grows faster but the deposited film contains many defects and has a lower density of atoms than a bulk  $\\mathrm{ZnO}$ . The fraction of defects decreases with increasing temperature.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_15","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_15","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the fraction of vacant lattice sites along the z-axis at various temperatures. Where the different temperatures result in different profiles also related to GPC.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| z-axis / angstrom | 27 °C | 52 °C | 77 °C | 127 °C | 177 °C |\\n|---|---|---|---|---|---|\\n| 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |\\n| 5.00 | 0.60 | 0.30 | 0.28 | 0.25 | 0.20 |\\n| 10.00 | 0.75 | 0.28 | 0.25 | 0.15 | 0.15 |\\n| 15.00 | 0.95 | 0.40 | 0.25 | 0.15 | 0.15 |\\n| 20.00 | 1.00 | 0.70 | 0.25 | 0.15 | 0.15 |\\n| 25.00 | 1.00 | 0.95 | 0.50 | 0.15 | 0.15 |\\n| 30.00 | 1.00 | 1.00 | 0.95 | 0.75 | 0.75 |\\n| 35.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |\\n| 40.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It seems that if the line reaches the value 1 this represents that the growth has stopped, indicating that for higher temperature the film thickness increases.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At the interface between the starting surface and at the surface.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"About 10 angstrom for the process' at 127 and 177 degrees. The rest of the material shows little vacancies and can be seen as the bulk.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Vacancies in the lattice structure can act as charge traps, resulting in lower performances.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":4,"width":612,"height":482}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_15.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_15.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/Timo Weckman et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":616,"height":486,"image_format":"jpeg","image_sha256":"5eb0d1d2b31432b74b78f39d38773470c734a6935353c7fe278a4b518561a630","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_8_figure_16.jpg","caption":"Figure 16. Fraction of lattice sites in the film that contain an ethyl-ligand. The fraction of sites containing an ethyl-ligand peaks at the beginning of the deposition. The peak then levels down to about 20 to  $25\\%$  of sites containing an ethyl-ligand.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_16","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_16","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the fraction of sites containing a ligand along the z-axis at various temperatures (27°C, 77°C, 127°C, and 177°C). The highest fraction occurs around 5 angstroms for all temperatures, with a subsequent decline.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| z-axis / angstrom | 27 °C | 77 °C | 127 °C | 177 °C |\\n|---|---|---|---|---|\\n| 0 | 0.00 | 0.00 | 0.00 | 0.00 |\\n| 3 | 0.48 | 0.48 | 0.35 | 0.30 |\\n| 5 | 0.20 | 0.20 | 0.20 | 0.20 |\\n| 10 | 0.15 | 0.24 | 0.24 | 0.24 |\\n| 15 | 0.00 | 0.24 | 0.24 | 0.24 |\\n| 20 | 0.00 | 0.24 | 0.24 | 0.24 |\\n| 25 | 0.00 | 0.10 | 0.15 | 0.15 |\\n| 30 | 0.00 | 0.00 | 0.00 | 0.00 |\\n| 35 | 0.00 | 0.00 | 0.00 | 0.00 |\\n| 40 | 0.00 | 0.00 | 0.00 | 0.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Deep in the film, right at the interface between the substrate and the film. This is irrespective of the deposition temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ABout 24 degrees.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The GPC is roughly equal as for both temperatures the film extends equally far in the z-direction.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1:0.50:0.50\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":615,"height":481}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_16.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_16.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/Timo Weckman et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":617,"height":486,"image_format":"jpeg","image_sha256":"95fc7f7b2835bd7cea0b33fe8b5bfe15703927e7fb4663274a8ef6b50ca2f5ae","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_8_figure_6.jpg","caption":"Figure 6. Change in energy as a function of water molecules as well as the averaged adsorption energy per adsorbed water molecule. At the ethyl-coverage of 7.1 ethyl  $\\mathrm{nm}^{-2}$ , the adsorption bond of a single water molecule is strong. However, inclusion of more water molecules weakens the average adsorption bond because of steric hindrance. After three water molecules, no further water can adsorb to the surface.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the trend of overall adsorption energy and averaged adsorption energy per H₂O molecule as the number of H₂O molecules increases, where the adsorption energy minimizes after the adsorption of three water molecules.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of H₂O molecules in the system | Overall adsorption energy | Averaged adsorption energy per H₂O |\\n|---|---|---|\\n| 1 | -1.25 | -1.25 |\\n| 2 | -1.75 | -0.80 |\\n| 3 | -2.25 | -0.75 |\\n| 4 | -2.1 | -0.5 |\\n| 5 | -2.25 | -0.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This can be concluded by the fact that the overall adsorption energy does not further decrease with increasing number of adsorbed water molecules.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Linear\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Water reacts with the remaining ligands at the surface. If these are all removed the water can no longer adsorb.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"-2.25 eV.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":625,"height":513}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/Timo Weckman et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":625,"height":514,"image_format":"jpeg","image_sha256":"414692eb2d13d30765643bc8d179d6b18f895c1cd223c97748026334c76186f2","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_8_figure_9.jpg","caption":"Figure 9. Simulated QCM data from the kMC model. Initially, the growth rate is the same for all temperatures when the surface becomes saturated with DEZ. However, after ethyl-saturation, the different temperature graphs diverge as the elimination of ethyl-ligands is low at low temperatures.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_9","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the mass gain over time at different temperatures (27°C, 77°C, 127°C, 177°C). The mass gain increases stepwise with time, indicating a cumulative effect, characteristic of ALD.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time / seconds | Mass-gain / ng cm⁻² 27 | Mass-gain / ng cm⁻² 77 | Mass-gain / ng cm⁻² 127 | Mass-gain / ng cm⁻² 177 |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 1 | 120 | 120 | 120 | 120 |\\n| 2 | 225|250|300|300|\\n| 4 | 300 |350|450|450|\\n| 6 | 300 |450|600|600|\\n| 8 | 300 |740|900|900|\\n| 10 | 375 |850|1050|1050|\\n| 12 | 375 |975|1200|1200|\\n| 14 | 375 |1100|1350|1350|\\n| 16 | 375 |1250|1500|1500|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The duration of single cycle is 1.6 seconds. This can be concluded from the graph that there are 10 cycles in a time span of 16 seconds in the figure.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No there is no cyclic growth. For ALD the increase every cycle should be the same, this is not the case fpr 27 degrees Celcius.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"127, 177 degrees Celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Water can condensate below a 100 degrees Celcius, so a process temperature of 77 degrees can result in liquid water in the reactor.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":609,"height":482}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/Timo Weckman et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":613,"height":483,"image_format":"jpeg","image_sha256":"f2913c5a31f3df34688b510d0fc06ebd46c077157640860d623693c5e51089a4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_11_figure_4.jpg","caption":"Figure 4.  $\\mathrm{ZnO}$  thickness as a function of the number of ALE cycles for temperatures between  $100$  and  $250^{\\circ}\\mathrm{C}$ , as measured by in situ SE.","id":"train/atomic-layer-etching/experimental-usecase/11/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/11/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The plot shows change in ZnO thickness as a function of number of ALE cycles at various temperatures\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of ALE Cycles | 100 °C (Å) | 150 °C (Å) | 200 °C (Å) | 250 °C (Å) |\\n|----------------------|-----------|-----------|-----------|-----------|\\n| 0                   | 585       | 585       | 585       | 585       |\\n| 10                  | 580       | 575       | 570       | 565       |\\n| 20                  | 575       | 565       | 555       | 545       |\\n| 30                  | 570       | 555       | 540       | 530       |\\n| 40                  | 565       | 545       | 530       | 525       |\\n| 50                  | 560       | 540       | 525       | 520       |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No, in-situ SE\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The EPCs at different temperatures are as follows, \\n1.  100 °C, 0.54 ± 0.05 Å\\n2.  150 °C, 0.97 ± 0.07  Å\\n3.   200 °C, 1.25 ± 0.08  Å\\n4.   250 °C, 1.31 ± 0.08 Å\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ZnO thickness can be seen to decrease linearly with the number of ALE cycles for each  temperature, in line with a layer-by-layer etching mechanism.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As seen in image, the thicknesses decrease more at higher temperatures, thus etch per cycle and the ALE increases with increase in temperature\"}]}]","bbox":[{"panel_id":"a","x":459,"y":3,"width":365,"height":303}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/11/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/11/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/11/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/11/Isotropic Atomic Layer Etching of ZnO Using Acetylacetone and O2 Plasma.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"11","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":427,"height":348,"image_format":"jpeg","image_sha256":"3a1b8e63e53530dfb0a821d70946069cf7a6f85b7ad132e21004c37923e038a0","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_3.jpg","caption":"Figure 3. Etch rate per cycle of the atomic layer etching (ALE) for (a) varying  $\\mathrm{O_2}$  and fixed  $\\mathrm{BCl_3}$  plasma times per cycle and (b) varying  $\\mathrm{BCl_3}$  and fixed  $\\mathrm{O_2}$  plasma times.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch rate increases linearly with increasing O₂ time/cycle for both BCl₃ time/cycles of 60 sec and 30 sec.\"},{\"panel_id\":\"b\",\"text\":\"The etch rate increases linearly with increasing BCl₃ time/cycle for both O₂ time/cycles of 60 sec and 30 sec.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| O₂ time/cycle (sec) | Etch rate [nm/cycle] BCl₃ time/cycle 30 sec | Etch rate [nm/cycle] BCl₃ time/cycle 60 sec |\\n|---|---|---|\\n| 0 | 0.0 | 0.0 |\\n| 30 | 0.4 | 0.4 |\\n| 45 | 0.5 | 0.6 |\\n| 50 | 0.5 | 0.6 |\\n| 60 | 0.55 | 0.7 |\\n| 90 | 0.55 | 0.7 |\"},{\"panel_id\":\"b\",\"text\":\"| BCl₃ time/cycle (sec) | Etch rate [nm/cycle] O₂ time/cycle 30 sec | Etch rate [nm/cycle] O₂ time/cycle 60 sec |\\n|---|---|---|\\n| 0 | 0.0 | 0.0 |\\n| 30 | 0.4 | 0.6 |\\n| 45 | 0.4 | 0.7 |\\n| 50 | 0.4 | 0.7 |\\n| 90 | 0.4 | 0.7 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Figure 3 shows that the etch rate per ALE cycle increases with increasing O₂ plasma exposure up to 60 s when the BCl₃ plasma time is held constant, indicating that longer oxidation enhances the formation of a removable surface layer. Beyond 60 s, the etch rate plateaus, suggesting that the surface becomes fully oxidized and that further O₂ exposure does not increase the amount of material available for removal, consistent with self-limiting behaviour. Similarly, when the O₂ plasma time is fixed at 60 s and the BCl₃ exposure is increased, the etch rate rises until 45 s, after which it also saturates, implying that the chlorination and removal step has become complete.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.4 nm/cycle.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.7 nm/cycle.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.6 nm/cycle.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":4,"width":549,"height":409},{"panel_id":"b","x":595,"y":8,"width":553,"height":404}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/Low-Damage and Self-Limiting (Al)GaN Etching Process.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"13","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1150,"height":414,"image_format":"jpeg","image_sha256":"ae2956d4683a4f7835b97ae8ba70f6c9f771226f6fa4224b643393f63fbed642","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_4.jpg","caption":"Figure 4. Etch rate when using  $\\mathrm{BCl}_3$  plasma under a fixed ICP power of  $400 \\mathrm{W}$  and varying bias powers.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the etch rate of GaN and AlGaN as a function of bias power. The etch rate increases linearly with bias power for both materials.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Bias power [W] | Etch rate [nm/min] GaN | Etch rate [nm/min] GaN |\\n|---|---|---|\\n| 0 | 0.0 | 0.0 |\\n| 5 | 1.0 | 1.0 |\\n| 10 | 4.0 | 2.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 mTorr.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"400 W.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"BCl3 plasma.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the bias power increases, initially no etching is observed. Only at a bias power higher than 2 W is any etching observed. This means that above 2 W, the BCl3 plasma is energetic enough to etch the substrates without the need of O2 plasma, which also means that above 2 W, the reactions during the BCl3 plasma exposure are not self-limiting.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":0,"width":619,"height":475}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/Low-Damage and Self-Limiting (Al)GaN Etching Process.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"13","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":623,"height":478,"image_format":"jpeg","image_sha256":"bf66abcda001fa1b3194d3edb70f4248b0130a5de1ef7a760bfb1318ee30bdf8","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_13_figure_8.jpg","caption":"Figure 8.  $I - V$  curves of the fabricated Schottky diodes. (a) Forward  $I - V$  characteristics (linear scale), (b) forward  $I - V$  characteristics (log scale), (c) reverse  $I - V$  characteristics (log scale), and (d) ideality factors.","id":"train/atomic-layer-etching/experimental-usecase/13/figure_8","sample_id":"atomic-layer-etching/experimental-usecase/13/figure_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between forward voltage and current for two methods: ALE and Digital etching.\"},{\"panel_id\":\"b\",\"text\":\"The line chart displays the logarithmic scale of current against forward voltage for ALE and Digital etching.\"},{\"panel_id\":\"c\",\"text\":\"The line chart illustrates the absolute value of current against reverse voltage for ALE and Digital etching.\"},{\"panel_id\":\"d\",\"text\":\"The line chart depicts the ideality factor n against forward voltage for ALE and Digital etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Forward voltage (V) | Current [mA] ALE | Current [mA] Digital etching |\\n|---|---|---|\\n| 0.0 | 0 | 0 |\\n| 0.5 | 0 | 0 |\\n| 1.0 | 20 | 20 |\\n| 1.5 | 60 | 60 |\\n| 2.0 | 100 | 100 |\"},{\"panel_id\":\"b\",\"text\":\"| Forward voltage (V) | Current [A] ALE | Current [A] Digital etching |\\n|---|---|---|\\n| 0.0 | 10^-11 | 10^-11 |\\n| 0.5 | 10^-5 | 10^-6 |\\n| 1.0 | 10^-2 | 10^-2 |\\n| 1.5 | 10^-2 | 10^-2 |\\n| 2.0 | 10^-1 | 10^-1 |\"},{\"panel_id\":\"c\",\"text\":\"| Reverse voltage (V) | Current [A] ALE | Current [A] Digital etching |\\n|---|---|---|\\n| -10 | 10^-4 | 10^-4 |\\n| -9 | 10^-5 | 10^-5 |\\n| -8 | 10^-5 | 10^-5 |\\n| -7 | 10^-6 | 10^-6 |\\n| -6 | 10^-6 | 10^-6 |\\n| -5 | 10^-6 | 10^-6 |\\n| -4 | 10^-7 | 10^-7 |\\n| -3 | 10^-7 | 10^-7 |\\n| -2 | 10^-8 | 10^-8 |\\n| -1 | 10^-10 | 10^-8 |\\n| 0 | 10^-12 | 10^-10 |\"},{\"panel_id\":\"d\",\"text\":\"| Forward voltage (V) | Ideality factor n Digital etching | Ideality factor n ALE |\\n|---|---|---|\\n| 0.0 | 0.5 | 1.4 |\\n| 0.05 | 1.3 | 1.3 |\\n| 0.1 | 1.8 | 1.2 |\\n| 0.15 | 1.9 | 1.2 |\\n| 0.2 | 2.2 | 1.1 |\\n| 0.25 | 2.4 | 1.2 |\\n| 0.3 | 2.5 | 1.1 |\\n| 0.35 | 1.7 | 1.2 |\\n| 0.4 | 1.5 | 1.2 |\\n| 0.45 | 1.2 | 1.2 |\\n| 0.5 | 1.2 | 1.5 |\\n| 0.55 | 1.2 | 1.7 |\\n| 0.6 | 1.6 | 2.3 |\\n| 0.65 | 2.2 |  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 20 mA.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is about 10^-8 A for the ALE sample and 10^-7 A for the Digital etching sample.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Figure 8d shows that the diode fabricated using ALE exhibits a much more uniform and lower ideality factor in the low forward-bias region compared to the diode processed by digital etching, indicating reduced trap-assisted recombination and fewer defect states at the junction. In contrast, the non-uniform slope observed for the digital etching case suggests greater plasma-induced damage.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Roughly 10^-10 A for the ALE sample and 10^-9 A for the Digital etching sample.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":11,"width":659,"height":432},{"panel_id":"b","x":703,"y":9,"width":663,"height":460},{"panel_id":"c","x":5,"y":535,"width":641,"height":421},{"panel_id":"d","x":721,"y":551,"width":645,"height":402}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/images/figure_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/13/Low-Damage and Self-Limiting (Al)GaN Etching Process.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"13","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1367,"height":964,"image_format":"jpeg","image_sha256":"3398e05ea2f7dd696f5c12129a0583084338d58a0dacc52469c93f2a2674326c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_16_fig_2.jpg","caption":"Fig. 2. a) Film deposition rate on different materials and b) radicals concentration over  $\\nu_{\\mathrm{CF4}}$  in the plasma mixture, while Ar fraction is  $83.3\\%$ . Negative values correspond to etching.","id":"train/atomic-layer-etching/experimental-usecase/16/fig_2","sample_id":"atomic-layer-etching/experimental-usecase/16/fig_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the deposition rate on different materials (Si, Al2O3, HfO2, AlNx, TiN, SiO2) against the fraction of CF4 in the plasma mixture. For all materials, the deposition rate increases rapidly at low CF₄ content and then decreases again at higher CF₄ concentrations.\"},{\"panel_id\":\"b\",\"text\":\"The figure illustrates the concentration of F, CF2, and CF species against the fraction of CF4 in the plasma mixture. For all three species, the concentration increases with CF₄ content, but the increase becomes more gradual at higher CF₄ concentrations.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ν_CF4 (%) | Si (nm/min) | Al2O3 (nm/min) | HfO2 (nm/min) | AlNx (nm/min) | TiN (nm/min) | SiO2 (nm/min) |\\n|-----------|-------------|----------------|---------------|---------------|--------------|---------------|\\n| ~3        | ~1.5        | ~2.5           | ~2.3          | ~2.6          | ~2.2         | ~0.8          |\\n| ~5        | ~4.0        | ~6.0           | ~5.5          | ~6.2          | ~1.5         | ~1.2          |\\n| ~10       | ~3.0        | ~4.5           | ~3.5          | ~4.8          | ~−1.0        | ~−5.5         |\\n| ~13       | ~2.0        | ~3.8           | ~3.2          | ~4.0          | ~−0.3        | ~−5.0         |\"},{\"panel_id\":\"b\",\"text\":\"| ν_CF4 (%) | F (cm⁻³)        | CF2 (cm⁻³)      | CF (cm⁻³)       |\\n|-----------|------------------|------------------|------------------|\\n| ~3        | ~1×10¹³          | ~5×10¹⁰         | ~2×10¹⁰         |\\n| ~5        | ~2×10¹³          | ~8×10¹⁰         | ~3×10¹⁰         |\\n| ~7        | ~5×10¹³          | ~1×10¹¹         | ~5×10¹⁰         |\\n| ~10       | ~8×10¹³          | ~1.6×10¹¹       | ~9×10¹⁰         |\\n| ~13       | ~1×10¹⁴          | ~1.8×10¹¹       | ~1.1×10¹¹       |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Both increase in value as the concentration of CH4 increases. However, F reaches a maximum value close to 10^14, while CF reaches 10^11\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The CF₄ fraction (ν_CF₄) in the plasma mixture.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The plot indicates that as the CF4 fraction in the plasma increases, the F concentration rises extremely rapidly for mixtures above about 5% CF4, growing by several orders of magnitude relative to CF and CF2 (due to the logarithmic scale). While CF and CF2 radicals contribute mainly to film growth, F atoms favor the formation of volatile fluorides and thus promote etching. When F becomes predominant in the plasma, deposition is progressively suppressed and, for some materials, the overall process switches from film growth to etching.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Fluorine radical concentration increases from 10^13 to the order of 10^14  with change in CF4 concentration and is higher than the CFx concentrations.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"With increase of CF4 fraction, the deposition rate increases slowly at first and then reaches a maximum. Further increase of CF4 leads to decrease of deposition rate for Si, Al2O3, HfO2, and AlNx.  In the case of SiO2 and TiN, increase of CF4 causes a change from deposition to etching.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No.\\nFluorine-containing radicals concentration increase with the growth of CF₄ fraction in plasma mixture.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"F (fluorine), with concentration near 10¹⁴ cm⁻³.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As ν_CF₄ increases, higher F and CFₓ radical densities initially enhance surface reactions, leading to higher deposition rates. Beyond the mid-fraction range, excessive fluorination and increased etching probability reduce net growth, producing the observed decline in deposition rates despite rising radical concentrations.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The radical concentration of CF has the highest dependence on the CF4 concentration.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the CF₄ fraction increases, the concentration of fluorine-containing radicals such as F, CF, and CF₂ rises significantly, enhancing chemical reactivity at the surface. At moderate CF₄ levels, polymer-forming species promote deposition, while at higher CF₄ fractions, the high density of reactive fluorine radicals enhances bond breaking and material removal. This shift in plasma chemistry alters surface reaction pathways, causing certain materials, such as SiO₂ and TiN, to transition from net deposition to net etching.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"TiN, SiO₂\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Although the process involves depositing materials onto a film, come materials, such as TiN and SiO2, etch. Therefore, the deposition rate decreases as the CF4 concentration increases.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The deposition rate is lowest for SiO1 and highest for ALNx\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The deposition ratio increases up to around 5% and 6% of CF4, then decreases as the concentration increases. In contrast, the concentration of CF4 and radicals is proportional, and the relationship is stronger in the case of F.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Atomic F has the fastest growth relative to CF and CF2, this implies a kinetic shift towards etching as v_CF4 increases.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Si, Al2O3, HfO2, AINx, TiN, SiO2\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"All materials start in net deposition at low ν_CF₄. As ν_CF₄ increases, the rates climb to a peak at low–mid ν_CF₄ and then drop. For SiO₂ and TiN, the curve crosses zero at higher ν_CF₄, this means they switch from deposition to etching. This is consistent with panel (b): the F radical grows fastest relative to CF/CF₂, pushing the surface balance toward etching.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The graph shows how the deposition rate changes with CF4 fraction for different materials. Based only on this data, AlNx looks like the best choice for industrial use. It has the highest deposition rate and stays relatively high across the whole CF4 concentration range, which would enable faster processing.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"6.6% (fastest deposition).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Below 5% CF4, as the concentration of CF4 increases, so too does the deposition rate on all substrates. Above 5%, the deposition rate generally decreases. On SiO2 however, etching occurs instead above 5% CF4 and this also occurs for TiN above 10% CF4.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN and SiO2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The chamber pressure was kept at 10 mTorr; The plasma was always composed of 83.3% Ar the molecular fraction of CF4 varied between 3.3 and 13.3%, with the remaining fraction consisting of H2.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At about 7% CF4 concentration, the deposition rates of Al2O3, HfO2, Si and AlN are highest.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the CF₄ fraction rises further, the concentration of highly reactive fluorine (F) radicals increases much more rapidly. This significant rise in 'F' radicals shifts the surface kinetics from a deposition-dominated process to an etching-dominated one, where 'F' atoms aggressively etch the deposited film.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"For higher CF₄ fractions:-\\n\\n1) On SiO₂ the deposition rate becomes negative, indicating an etching process.\\n\\n2) On Si, Al₂O₃, HfO₂, and AlNₓ, the deposition rate decreases but remains positive, indicating net deposition.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, but only for specific materials.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO₂ and TiN\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Low CF₄ fraction (<5%) because: (1) provides polymer deposition on target materials (Al₂O₃, AlNx, HfO₂), (2) avoids etching of potential mask materials (SiO₂, TiN)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes (approximately—most materials peak around 5-6.6% CF₄).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Highest: [ALNx] (~[6.1]); net etch: SiO2 (negative).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The materials are TiN and SiO2\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The material that should be used as substrate is AlNx\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"An experiment can be designed such that initial thickness and final thickness can be determined after exposure to the plasma. Some materials thicknesses will decrease, others will increase.\"}]}]","bbox":[{"panel_id":"a","x":43,"y":35,"width":470,"height":319},{"panel_id":"b","x":529,"y":36,"width":434,"height":318}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/Selective atomic layer etching of Al2O3, AlNx and HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":964,"height":355,"image_format":"jpeg","image_sha256":"7032718f0be80a9e6064ebb2bc165f223e723a2d2408bb3bc7ba3011334b70cf","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_16_fig_4.jpg","caption":"Fig. 4. The etch rate over modification step duration a) for all studied materials, and b) enlarged for  $\\mathrm{Al_2O_3}$ ,  $\\mathrm{AlN_x}$ , and  $\\mathrm{HfO_2}$ .","id":"train/atomic-layer-etching/experimental-usecase/16/fig_4","sample_id":"atomic-layer-etching/experimental-usecase/16/fig_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Etch rate as a function of modification step duration shows material-dependent behavior. SiO₂ increases nearly monotonically, reaching ~0.8 nm/cycle at long times. AlNₓ and Al₂O₃ exhibit mid-range maxima (~6–10 s) followed by a decline at longer durations. HfO₂ peaks later (~10–12 s) at a lower maximum rate. TiN remains near zero or slightly negative across the full duration range.\"},{\"panel_id\":\"b\",\"text\":\"The enlarged view emphasizes distinct optimal modification windows: AlNₓ peaks earliest (~6–8 s), Al₂O₃ slightly later (~8–10 s), and HfO₂ latest (~10–12 s). Beyond these windows, etch rates decrease, indicating over-modification effects.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Modification step duration (s) | Al₂O₃ (nm/cyc) | HfO₂ (nm/cyc) | AlNₓ (nm/cyc) | TiN (nm/cyc) | SiO₂ (nm/cyc) |\\n|---|---|---|---|---|---|\\n| 0  | ~0.02 | ~0.01 | ~0.02 | ~0.00 | ~0.05 |\\n| 4  | ~0.15 | ~0.06 | ~0.17 | ~0.00 | ~0.30 |\\n| 6  | ~0.17 | ~0.08 | ~0.18 | ~0.00 | ~0.45 |\\n| 10 | ~0.16 | ~0.09 | ~0.17 | ~0.00 | ~0.62 |\\n| 12 | ~0.15 | ~0.09 | ~0.16 | ~0.00 | ~0.65 |\\n| 14 | ~0.09 | ~0.07 | ~0.10 | ~−0.02 | ~0.82 |\"},{\"panel_id\":\"b\",\"text\":\"| Modification step duration (s) | Al₂O₃ (nm/cyc) | HfO₂ (nm/cyc) | AlNₓ (nm/cyc) |\\n|---|---|---|---|\\n| 0  | ~0.02 | ~0.01 | ~0.02 |\\n| 4  | ~0.15 | ~0.07 | ~0.17 |\\n| 6  | ~0.17 | ~0.09 | ~0.19 |\\n| 10 | ~0.16 | ~0.10 | ~0.19 |\\n| 12 | ~0.16 | ~0.10 | ~0.18 |\\n| 14 | ~0.08 | ~0.06 | ~0.10 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The etch ratio of different materials such as Al2O3, HfO2, AlNx, TiN, SiO2, is analyzed as a function of modification step duration.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The etch rate increases for all materials except for SiO2 up to approximately 5s, after which remains constant until 12s before finally decreasing. In contrast, the etch ratio of SiO2 depend on the duration of the modification step.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO2, and TiN\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Using SiO2 enables higher etch ratios to be achieved in a shorter amount of tiem.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For AL2O3, HFO2, and AINx, the etch rate rises rapidly and peaks around 6-8 seconds. It then falls off. TiN's etch rate remains low/inactive across the whole duration. SiO2's etc rate rises throughout the whole duration range.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"a2: TiN\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO2 (highest etch rate by modification step t = 0)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Al2O3, HfO2, and AlNx show an ALE window. Their etch rate rises with increasing modification time and then becomes nearly constant, indicating self-limiting, surface-saturation behavior characteristic of ALE. SiO2 does not show a clear window because its etch rate keeps increasing with time, and TiN does not because it is essentially not etched.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Al2O3/TiN, HfO2/TiN, AlNx/TiN.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 5 and 12 seconds; Al2O3, HfO2 and AlNx.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"According to the authors, this is most likely due to the difference in volatility between the formed products: the sublimation point of titanium fluoride is at 284 ◦C and the boiling point of silicon-fluoride is -86 ◦C. Titanium fluoride is much less volatile, which makes etching more difficult and the etch rate slower. The sputtering rate of argon ions on SiO2 is also two to three times higher than that of the other studies materials, which adds to the etch rate.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Although insignificant, the minimal etching is attributed to sputtering by argon ions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For SiO2 the etch rate increases linearly with modification step duration, for other materials it is more complex\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"With the increase of modification step duration, the etch rate of Al2O3, AlNx and HfO2 reaches saturation, providing the ALE window. Further rise of modification step leads to the decrease of the etch rate due to too thick film deposition\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing the modification time gives the activating chemicals more time to react with the surface, breaking more bonds and making the material easier to remove. This creates a thicker or more reactive modified layer, which leads to a higher etch rate. However, this effect eventually levels off once the surface is fully saturated.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"HfO₂\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SiO2 exhibits continuously increasing etch rate because silicon fluoride (SiF4) desorbs easily at lower temperatures, making it easily removed even with relatively low ion energy bombardment. In contrast, TiN shows virtually no etching because titanium fluoride has a much higher sublimation point, making desorption of reaction products extremely difficult at the process conditions used. This difference in etch behavior provides good selectivity for etching Al2O3, AlNx, and HfO2 over a TiN hard mask.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"AlNx\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 5-6 seconds\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Highest: SiO₂ (~0.82). \\nLowest: TiN (~0).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SiO2 has a way higher etch rate than for example HfO2, at 15 s of modification duration. This indicates that SiO2, at least partially, can selectively be etch with respect to HfO2.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN does not etch for any modification duration.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"5 s of modification duration is most likely used as this would have similar etch results as longer times, but reduces the processing time.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No there is continuous etching. For all three materials the etch rate is non-zero at 0 s modification duration.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At short modification durations, the surface is only partially covered by the fluorocarbon film, resulting in very low etch rates due to insufficient surface activation. As the modification time increases, a more complete reactive layer forms, enabling efficient ion-assisted reactions during the activation step and causing the etch rate to rise. Beyond an optimal duration, the etch rate saturates, reflecting self-limiting behavior, and may slightly decrease at longer times due to excessive film thickness hindering ion penetration and surface reaction efficiency.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Al2O3, HfO2, AINx\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At modification step duration of 5–12 seconds.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate of Al2O3, AlNx and HfO2 with rise of modification step to 12s lead to the decrease of the etch rate going out of saturation\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"- AlNx: shows the highest etch-rate increase with duration.\\n- Al₂O₃: displays a moderate increase with duration.\\n- HfO₂: increases more slowly and remains the lowest of the three across the range.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. \\n Al₂O₃ is not suitable for a process that must be highly precise and insensitive to modification-step duration.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"All three materials show similar behavior with initial etch rate increase, followed by a plateau at 5-12 seconds representing the ALE window. After 12 seconds, the etch rate decreases. This self-limiting plateau demonstrates atomic layer etching behavior.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"AlNₓ reaches its maximum etch rate earliest, around 6–8 s, followed by Al₂O₃ at roughly 8–10 s, while HfO₂ peaks later at about 10–12 s. This ordering suggests different kinetics for forming the modified surface state that enables removal, with HfO₂ requiring longer modification to reach maximum effectiveness.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Presence of an etch-rate saturation plateau with increasing modification time, Formation of a clear ALE window at intermediate modification durations, Reduced sensitivity to argon sputtering effects, Ability to form a self-limiting modified surface layer\"}]}]","bbox":[{"panel_id":"a","x":13,"y":33,"width":507,"height":349},{"panel_id":"b","x":547,"y":51,"width":406,"height":317}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/Selective atomic layer etching of Al2O3, AlNx and HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":953,"height":383,"image_format":"jpeg","image_sha256":"8546dbe44307f34383247a219f554da1449ae7012e2c77cb6ccf556bb405e858","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_16_fig_5.jpg","caption":"Fig. 5. The etch rate over activation step duration. Negative values correspond to film formation.","id":"train/atomic-layer-etching/experimental-usecase/16/fig_5","sample_id":"atomic-layer-etching/experimental-usecase/16/fig_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Etch rate as a function of activation step duration shows a common negative starting value near −0.5 nm/cycle at 0 s for all materials, indicating net film formation without activation. As activation time increases, material-dependent removal emerges. SiO₂ exhibits the strongest response, increasing monotonically to ~0.7 nm/cycle at 20 s. AlNₓ and Al₂O₃ show intermediate increases, reaching ~0.3 nm/cycle at 20 s, while HfO₂ increases only modestly. TiN remains near zero across the range, indicating strong resistance to activation-induced removal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Activation step duration (s) | Al2O3 etch rate (nm/cyc) | HfO2 etch rate (nm/cyc) | AlNx etch rate (nm/cyc) | TiN etch rate (nm/cyc) | SiO2 etch rate (nm/cyc) |\\n|---|---|---|---|---|---|\\n| 0  | ~-0.55 | ~-0.55 | ~-0.55 | ~-0.55 | ~-0.55 |\\n| 5  | ~0.00  | ~0.00  | ~0.00  | ~0.00  | ~0.40  |\\n| 10 | ~0.18  | ~0.10  | ~0.18  | ~0.00  | ~0.55  |\\n| 20 | ~0.30  | ~0.15  | ~0.30  | ~0.05  | ~0.72  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They are studying the etch rate  (nm/cyc) as a function of the activation step duration (s) for the different materials used in the process (Al2O3, HfO2, AlNx, TiN, SiO2)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No, because all materials demonstrate film deposition when activation step duration is less than 5s (2.5s for SiO2).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No it does not seem to matter. Both Al2O3 and AlNx have similar etch results.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes it is possible, SiO2 can be selectively etched wrt to all other materials at an activation time of 5 seconds.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, there is an increase in etch rate for increasing activation time\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"20 s for HfO2 and 10 s for Al2O3.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO2 at 20 s with ≈ 0.70nm/cycle.\\nEvidence: Its marker/line is highest among the series at 20 s.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch-per-cycle response depends on material. SiO2 shows the steepest increase with activation time, while AlNx and Al2O3 rise more modestly and TiN remains near zero. By 20s, the ranking is SiO2 ≫ AlNx ~ Al2O3 ≫ HfO2≫ TiN. This indicates [concise implication about selectivity or sensitivity to activation] grounded in the plotted slopes.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Stop layer: TiN — its curve stays ~0 across the range (slope ≈ 0).\\nLongest activation to achieve net removal: TiN— remains negative or near 0 until  10s, then crosses into positive etch.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At zero activation, polymer deposits at -0.5 nm/cycle on all surfaces. Increasing activation duration causes etch rates to shift from negative to positive as ions first remove polymer then activate material etching. The gradual increase in etch rate with activation time reflects the sequential sputtering and surface reaction activation processes.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 0.5 nm/cycle (i.e. negative etch rate of -0.5 nm/cycle)\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is 0.16 nm/cycle\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No. The reason is that the deposited polymer film must first be sputtered away before material etching can begin.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The activation step duration directly controls the etch rate by determining the extent of surface modification. At very short durations, incomplete reactions lead to net film formation, but as duration increases, the surface becomes fully saturated, allowing the etch rate to rise sharply. Once the surface is completely modified, the process reaches a self-limiting plateau where further increases in activation time no longer affect the etch rate.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1.\\tAt very short durations, film formation (negative etch rate) occurs.\\n2.\\tAs duration increases, the process transitions from net deposition to net etching.\\n3.\\tAt intermediate durations, the etch rate rises sharply.\\n4.\\tAt long durations, the etch rate plateaus at a higher value.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes . HfO₂ etches more slowly than Al₂O₃ because its higher sputtering threshold reduces ion-induced material removal.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After 5s of activation steps, the etch rate begins to increase for all materials except TiN\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, the etch rate remains low and only goes up to 0.02nm/cycle after 20s of activation steps\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, Al2O3 and AlN show similar behaviour in etch rate vs activation steps and also have similar etch rates.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SiO2 requires the least number of activation as the etch rate increases after ~2.5s of activation step duration.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A negative etch rate corresponds to film growth/formation, rather than etching.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 seconds.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"HfO2 has a higher sputtering threshold.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is probably due to the non-monoenergetic ion distribution function. The incline observed in the ALE window for the activation step may hamper reproducibility.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A strong time-dependent increase in etch rate indicates that more volatile products are generated as the activation time increases. Almost no volatile species are produced during the activation of TiN, since its etch rate remains essentially zero over the entire activation-time range. In contrast, SiO₂ is the most volatile, exhibiting the highest etch rate and a continuous rise with activation time.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At around 5 seconds.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 1 nm thick (0.5 nm/cycle).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Al₂O₃, HfO₂, and AlNₓ all show a  shallow rise in etch rate as the activation step becomes increases. TiN stays close to zero throughout, only moving slightly upward from a negative value at short times. SiO₂  increases far more strongly than the other materials, becoming the most rapidly etched surface at longer activation durations.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO2\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Al2O3, HfO2, SiO2\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Zero activation: polymer film deposition (~0.5 nm) results in negative etch rate, Early activation (5–10 s): transition from deposition to etching, Moderate activation (10–15 s): onset of positive etch rate and initiation of steady etching, Longer activation (15–20 s): gradual increase in etch rate, no clear plateau\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.0 nm/cycle for both materials (transition point from deposition to etching).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"HfO₂ shows a slower increase in etch rate compared to Al₂O₃, with the slope of the etch rate approximately half that of Al₂O₃. This difference arises from HfO₂’s higher sputtering threshold and stronger resistance to argon ion bombardment, which reduces etching efficiency and delays the onset of steady material removal, highlighting how intrinsic material properties influence ALE performance.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":689,"height":434}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/Selective atomic layer etching of Al2O3, AlNx and HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":689,"height":434,"image_format":"jpeg","image_sha256":"873a06926fe98b1e93ec14cd7b5192cf73ca509a1ee6dd3d58fd48fab3a76bb2","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_16_fig_6.jpg","caption":"Fig. 6. The etch depth over the number of cycles in ALE process (activation step duration is  $10~\\mathrm{s}$ ).","id":"train/atomic-layer-etching/experimental-usecase/16/fig_6","sample_id":"atomic-layer-etching/experimental-usecase/16/fig_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Etch depth increases approximately linearly with the number of ALE cycles for most materials, indicating a near-constant etch-per-cycle under a 10 s activation step. SiO₂ exhibits the steepest slope and the largest accumulated etch depth, reaching ~27 nm after 40 cycles. AlNₓ and Al₂O₃ show moderate linear removal, while HfO₂ etches slowly. TiN shows essentially no measurable etch depth across all cycles, indicating strong resistance to the ALE process. Dashed lines represent linear fits to the experimental data.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycles | Al2O3 etch depth (nm) | HfO2 etch depth (nm) | AlNx etch depth (nm) | TiN etch depth (nm) | SiO2 etch depth (nm) |\\n|---|---|---|---|---|---|\\n| 0  | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |\\n| 15 | ~2.0 | ~1.5 | ~3.0 | ~0.0 | ~6.0 |\\n| 20 | ~3.5 | ~2.5 | ~5.0 | ~0.0 | ~12.0 |\\n| 30 | ~5.0 | ~3.2 | ~6.0 | ~0.0 | ~20.0 |\\n| 40 | ~6.0 | ~4.0 | ~7.5 | ~0.0 | ~27.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The etch depth as a function of the number of cycles, for different materials.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN, because the etch depth value is close to 0, regardless of the number of cycles.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No, because, unlike other materials, the etch depth continues to increase with each cycle. However, after 30 cycles, the etch depth appears to remain constant for Al2O2, HfO3, AlNx, and TiN.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO2 is preferred because its slope is steeper than that of the other materials. This means that a greater etching depth can be achieved in the same number of cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SiO2 does not show that good of a fit. This suggests that, if the characterization is performed well, that the process is not that controlled. This in turn suggests that perfect ALE is not acquired.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"HfO2 shows the highest etch resistivity as for this material, the total etch depth is the lowest.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Al2O3 and AlNx.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"30 cycles are required.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes, approximately linear for Al2O3, HfO2, AlNx, SiO2;  TiN shows flat/near-zero change.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"“SiO2”\\nEvidence: “At 40 cycles it has the largest etch depth  compared with other materials.\\\"\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"EPC depends strongly on material. SiO2 has the highest slope, followed by AlNx and Al2O3, while HfO2 is low and TiN is ≈0. This ranking indicates, how etch depth varies with increasing number of cycles.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TiN is the best candidate for an etch-stop under these ALE conditions. Its etch depth stays zero after 40 cycles, giving an etch-per-cycle of 0 nm/cycle. In the plot, its markers lie near the x-axis and the dashed fit has nearly zero slope, whereas other materials show measurable slopes and larger depths. Therefore, TiN would maintain thickness during etching and serve as an effective mask/stop layer.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Al₂O₃ – black squares\\n\\nHfO₂ – red circles\\n\\nAlNₓ – blue triangles\\n\\nSiO₂ – black asterisks\\n\\nTiN – red crosses\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 seconds.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Roughly, 5 to 6 nm.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process doesn't drift quickly with time. That allows predictable depth control just by choosing the cycle number.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, the data in Figure 6 confirms the process's adherence to the self-limiting mechanism, as evidenced by the perfectly linear relationship between etch depth and the number of cycles. \\nThis linearity is governed by material-specific properties, including the surface's ability to saturate during the activation step and the volatility of the etched byproducts during the removal step. \\nFor the material tested here, these properties are well-suited to the ALE process, whereas materials with poor surface saturation or low byproduct volatility would show a sub-linear, reaction-limited etch rate.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The linear relationship is significant because it confirms the process is self-limiting itself , highly controllable and predictable. \\nIn a perfect ALE process, each cycle removes exactly one layer of atoms and then stops on its own. It cannot continue etching deeper, no matter how long you leave the reactants there. This trend is very well observed for most of the materials studied here.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1) The Etch Per Cycle (EPC) for various different materials\\n2) The total etch depth for a given number of cycles.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes \\n(This trend allows device manufacutures to achieve atomic-scale precision by simply controlling the number of cycles. Also, they can reliably etch to a target depth with minimal risk of damaging the underlying substrate, which is crucial for manufacturing advanced semiconductor devices.)\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"From the graph, SiO2 has higher etch rate compared to other materials\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch depth has a linearly dependence  with the number of cycles\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"From the graph, it is evident that TiN is not etched at all, while SiO2 has high etch rate, therefore with the current process, SiO2 could be selectively etched from a target containing both TiN and SiO2 films.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 seconds for both steps each.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.16, 0.20, and 0.11 nm per cycle, respectively.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"About 30, 40, and 20, respectively.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The measured etch rates for Al2O3, AlNx and HfO2 are comparable to the Al-O bond length of 0.169 nm, which means that the ALE process has close to single-atom precision.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TiN would make the best mask. From the graph, its etch depth stays close to zero even after 40 cycles, meaning it is almost not etched at all. It can protect underlying layers while the other materials are removed.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 24 - 25 cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SiO2, AlNx, Al2O3, HfO2.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiO2 (largest fluctuations in etch depth).\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The five materials show noticeably different slopes. SiO₂ increases the most rapidly and reaches the largest etch depth. AlNₓ, Al₂O₃, and HfO₂ all rise more moderately, with HfO₂ showing the slowest increase of the three. TiN remains nearly flat across the entire range, with only a slight upward shift at the highest cycle count.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SiO2, AINx, Al2O3, HfO2\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Highest: AlNₓ (0.20 nm/cycle), Medium: Al₂O₃ (0.16 nm/cycle), HfO₂ (0.11 nm/cycle), Lowest: TiN (0.006 nm/cycle), Poor repeatability: SiO₂ (0.62 nm/cycle)\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"8 nm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The chart shows that etch depth increases linearly with cycles, but the rate differs strongly between materials. Al₂O₃, AlNₓ, and HfO₂ exhibit controlled, moderate etch rates, indicating effective ALE behavior. TiN shows almost no etching, confirming resistance to the process, while SiO₂ has high variability, indicating poor repeatability. This illustrates that the ALE process can selectively remove material layer-by-layer depending on the chemical and physical properties of each substrate, highlighting the importance of tuning parameters for selective etching.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":698,"height":556}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/16/Selective atomic layer etching of Al2O3, AlNx and HfO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":698,"height":556,"image_format":"jpeg","image_sha256":"26b84b84f0ecb72b7f97dbf259e00332f800ffd8ea39552adb2d40b3d009e63c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_23_figure_11.jpg","caption":"Figure 11 displays the time evolution of the HF exposure together with the  $\\mathrm{H}_2\\mathrm{S}^+$  ion signal at  $m / z$  34 produced by the","id":"train/atomic-layer-etching/experimental-usecase/23/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_11","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multi-line chart shows how the etch rate (Å/cycle) of ZnS ALD changes with temperature for depositions carried out at 100 °C, 200 °C, and 300 °C. As the temperature increases, the etch rate also increases.The film deposited at 100 °C etches the fastest, while the film deposited at 300 °C etches the slowest at any given temperature. This implies that films deposited at higher temperatures are denser or chemically more robust.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Etch Rate (Å/cycle) | Condition        |\\n|------------------|----------------------|------------------|\\n| 225              | 1.40                 | ZnS ALD 100°C    |\\n| 225              | 1.02                 | ZnS ALD 200°C    |\\n| 225              | 0.77                 | ZnS ALD 300°C    |\\n| 250              | 1.80                 | ZnS ALD 100°C    |\\n| 250              | 1.44                 | ZnS ALD 200°C    |\\n| 250              | 1.33                 | ZnS ALD 300°C    |\\n| 275              | 2.00                 | ZnS ALD 100°C    |\\n| 275              | 1.75                 | ZnS ALD 200°C    |\\n| 275              | 1.65                 | ZnS ALD 300°C    |\\n| 300              | 2.11                 | ZnS ALD 100°C    |\\n| 300              | 1.98                 | ZnS ALD 200°C    |\\n| 300              | 1.85                 | ZnS ALD 300°C    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The setpoint should be 300 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The higher etch rate for the 100 °C film (black curve) indicates it is structurally weaker and less dense than the 300 °C film. The \\\"looser\\\" atomic network of the low-temperature film allows the etchant species to diffuse and react more readily, whereas the 300 °C film is more crystalline and compact, offering greater resistance to chemical attack.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Films deposited at lower temperatures (100 °C) typically have lower density and higher impurity content (e.g., unreacted ligands).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At high etch temperatures (300 °C), the etch rates for all films begin to converge (gap narrows to ~0.2 Å/cycle ). This suggests the process is transitioning from a surface-reaction-limited regime (where film density/quality dominates) to a mass-transport-limited regime (or saturation). In this regime, the supply of etchant or the desorption of byproducts becomes the bottleneck, making the intrinsic quality of the film less of a differentiator than it is at lower temperatures.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":647,"height":566}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/Thermal Atomic Layer Etching of Zinc Sulfide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"23","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":650,"height":570,"image_format":"jpeg","image_sha256":"48905067cd54a2fab10fff153a34a7c983a7049a624d7287df01951d22183819","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_23_figure_12.jpg","caption":"Figure 12.  $\\mathrm{ZnS}$  etch rates vs temperature for  $\\mathrm{ZnS}$  ALD films grown at different temperatures.","id":"train/atomic-layer-etching/experimental-usecase/23/figure_12","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_12","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multi-line chart shows how the etch rate (Å/cycle) of ZnS ALD changes with HF pressures for depositions carried out at TMA pressures of 20 m.Torr and 200 m.Torr. As the pressure increases till  50, the etch rate increases and plateus.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| HF Pressure (mTorr) | Etch Rate (Å/cycle) | TMA Pressure |\\n|----------------------|----------------------|--------------|\\n| 0                    | 0.00                 | 20 mTorr     |\\n| 0                    | 0.00                 | 200 mTorr    |\\n| 50                   | 1.65                 | 20 mTorr     |\\n| 50                   | 2.00                 | 200 mTorr    |\\n| 100                  | 1.75                 | 20 mTorr     |\\n| 100                  | 2.24                 | 200 mTorr    |\\n| 150                  | 1.81                 | 20 mTorr     |\\n| 150                  | 2.43                 | 200 mTorr    |\\n| 200                  | 1.83                 | 20 mTorr     |\\n| 200                  | 2.43                 | 200 mTorr    |\\n| 250                  | 1.83                 | 20 mTorr     |\\n| 250                  | 2.46                 | 200 mTorr    |\\n| 300                  | 1.83                 | 20 mTorr     |\\n| 300                  | 2.48                 | 200 mTorr    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"First, the TMA pressure must be set and stabilized at 200 mTorr. Then, the HF pressure must be set and stabilized at 50 mTorr for the duration of the ALE cycle before measuring the etch depth.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"200 mTorr TMA is better for throughput as it gives a higher etch rate. The trade-off is higher precursor consumption and potentially more stringent pumping requirements to maintain process pressure.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The plateau occurs because all available surface reaction sites become saturated with HF-derived species within the pulse time. This reveals the self-limiting nature of the HF half-reaction, a defining characteristic of ideal ALE.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It implies the tool requires excellent pressure control and uniformity for both precursors. Any drift or non-uniformity in pressure will directly translate into etch rate variations, affecting device yield and performance consistency.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":665,"height":554}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/Thermal Atomic Layer Etching of Zinc Sulfide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"23","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":669,"height":558,"image_format":"jpeg","image_sha256":"bcd140015b59ad8c8b4bc8eea4b848b59dcb1ff3dd436ff57ef6d5431f9247eb","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_23_figure_6.jpg","caption":"Figure 6. Etch rate during  $\\mathrm{ZnS}$  ALE vs precursor exposure time at  $300^{\\circ}\\mathrm{C}$ . The HF pressure was  $40\\mathrm{mTorr}$  and the TMA pressure was  $45\\mathrm{mTorr}$ . Initial  $\\mathrm{ZnS}$  ALE films were grown at  $100^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/23/figure_6","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This muliple line chart  shows the etch rate per cycle (Å/cycle) as a function of precursor exposure time for two different ALE process conditions: one varying HF exposure time (x) with fixed TMA time (4 s), and the other varying TMA exposure time (y) with fixed HF time (1 s). Both show a rapid increase in etch rate with exposure time up to ~2 seconds, followed by a saturation plateau, indicating self-limiting surface reactions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Precursor Exposure Time (s) | Etch Rate (Å/cycle) | HF-Purging-TMA-Purging     |\\n|-----------------------------|----------------------|-----------------------------|\\n| 0                           | 0                    | x-30-4-30 (s)               |\\n| 0                           | 0.006                | 1-30-y-30 (s)               |\\n| 1                           | 2.08                 | x-30-4-30 (s)               |\\n| 1                           | 0.57                 | 1-30-y-30 (s)               |\\n| 2                           | 2.02                 | x-30-4-30 (s)               |\\n| 2                           | 1.8                  | 1-30-y-30 (s)               |\\n| 3                           | 2.03                 | x-30-4-30 (s)               |\\n| 3                           | 2.03                 | 1-30-y-30 (s)               |\\n| 4                           | 2.05                 | x-30-4-30 (s)               |\\n| 4                           | 2.05                 | 1-30-y-30 (s)               |\\n| 5                           |                      | x-30-4-30 (s)               |\\n| 5                           |                      | 1-30-y-30 (s)               |\\n| 6                           | 2.04                 | x-30-4-30 (s)               |\\n| 6                           | 2.05                 | 1-30-y-30 (s)               |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the solid red  line (x-30-4-30): 1) HF exposure for ‘x’ seconds → 2) Purge for 30 s → 3) TMA exposure for 4 s → 4) Purge for 30 s.\\nFor the blue line (1-30-y-30): 1) HF exposure for 1 s → 2) Purge for 30 s → 3) TMA exposure for ‘y’ seconds → 4) Purge for 30 s.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TMA requires a longer exposure to reach saturation (etch rate plateaus after ~3 s) compared to HF (~1 s). This implies that the TMA half-reaction is kinetically slower, so process optimization must balance sufficient TMA exposure for complete etching against throughput, as longer TMA times increase cycle time.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The saturation indicates that the surface reaction is self-limiting; after ~2 seconds, all available surface sites have reacted. This reveals that the etching mechanism is governed by monolayer-level surface saturation rather than continuous etching, which is the hallmark of ideal ALE.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It ensures robust uniformity despite inevitable local variations in precursor flux or exposure time. As long as the exposure time exceeds the saturation threshold everywhere on the wafer, the etch rate will be identical, eliminating thickness variations caused by flow non-uniformities.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":662,"height":556}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/Thermal Atomic Layer Etching of Zinc Sulfide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"23","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":666,"height":561,"image_format":"jpeg","image_sha256":"6f88d1b0c4648983c5399bf758d71256a04938a5e55929eef60c4cd4108fd5ec","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_28_fig2.jpg","caption":"FIG.2. (a) SIMS depth profiles of F as a function of the sputtering time after  $\\mathrm{NF}_3$  plasma treatment for  $60~\\mathrm{s}$  at 0, 10, 50, and  $100W$  b) F 1s XPS peak from  $\\mathrm{Al}_2\\mathrm{O}_3$  films; and (c) saturation of atomic fractions as  $\\mathrm{NF}_3$  plasma exposure time at  $10W$  and  $100^{\\circ}C$","id":"train/atomic-layer-etching/experimental-usecase/28/fig2","sample_id":"atomic-layer-etching/experimental-usecase/28/fig2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multi spectra chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the intensity of F-  at different depths for various powers (0, 10, 50, 100 W). Increasing power leads to penetration of the F⁻ deeper into the film.\"},{\"panel_id\":\"b\",\"text\":\"The XP-spectrum t displays the intensity of F 1s peaks at different times (Reference, 5s, 20s, 50s, 60s)  during plasma fluorination. The Al–F peak grows strongly after fluorination and then changes more modestly at longer times.\"},{\"panel_id\":\"c\",\"text\":\"The chart illustrates the atomic percentage changes of O1 s, Al 2p, F 1s, and C 1s with increasing plasma fluorination time. O, Al and C decrease rapidly and then plateau, F increases rapidly and then plateaus.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Depth (nm) | 0W: Intensity (arb.units) | 10W: Intensity (arb.units) | 50W: Intensity (arb.units) | 100W: Intensity (arb.units) |\\n|---|---|---|---|---|\\n| 0 | 2 | 65 | 63 | 66 |\\n| 1 | 2|  43| 52 | 60 |\\n| 2 | 2 | 15 | 15 | 20 |\\n| 3 | 2 | 8 | 10 | 11 |\\n| 4 | 2 | 6 | 8 | 7 |\"},{\"panel_id\":\"b\",\"text\":\"| Measurement | Peak: Intensity (arb.units) | Peak: Binding energy [eV]  | \\n|---|---|---|\\n| Reference | 2.15 | 685 | \\n| 5s |  29 | 686.1 | \\n| 20s| 38 | 686.4 | \\n| 50s| 42 | 686.4 | \\n| 60s| 42 | 686.4 |\"},{\"panel_id\":\"c\",\"text\":\"| Plasma fluorination time (s) | Atomic % O 1s | Atomic % Al 2p | Atomic % F 1s | Atomic % C 1s |\\n|---|---|---|---|---|\\n| 0 | 52.1% | 28.8% | 0% | 18.5% |\\n| 5 | 42% | 27.9% | 21.2% | 9% |\\n| 20 | 38.7% | 27% | 21.8% | 10.8% |\\n| 50 | 37.5% | 27.9% | 24.9% | 8.6% |\\n| 60 | 38.4% | 27.9% | 25.3% | 8.1% |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The depth profiles indicate that increasing RF power leads to greater fluorine incorporation and a deeper fluorinated layer. This is attributed to higher RF power producing a larger flux of reactive fluorine radicals in the NF₃ plasma, which enhances fluorination. The fluorine signal still decreases with depth because the formed fluorinated layer acts as a diffusion barrier and hinders further transport.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approx. 50 s (F 1s peak stops increasing).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Noticeable binding energy shift suggest that the dominant bonding environment might be evolving between 5s and 20s (consistent with the probed surface becoming more uniformly fluorinated).\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Fluorination does not appear to modify the entire region probed by XPS. The F atomic % increases but reaches a plateau while O remains predominant. It is consistent with a picture of a fluorine-enriched surface layers over underlying oxide that is still largely unfluorinated.\"}]}]","bbox":[{"panel_id":"a","x":102,"y":10,"width":556,"height":478},{"panel_id":"b","x":111,"y":522,"width":542,"height":470},{"panel_id":"c","x":72,"y":1026,"width":583,"height":495}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/images/fig2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/Atomic layer etching of Al2O3 with NF3 plasma fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":658,"height":1528,"image_format":"jpeg","image_sha256":"762d9e5088c12f0da2c44d7d29c314d9a78596c427499a5acbe82bea3709f92d","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_28_fig_4.jpg","caption":"FIG. 4. (a) TMA exposure time at 290, 300, and  $330^{\\circ}\\mathrm{C}$  of the removal step; (b) EPC using  $\\mathrm{NF}_3 / \\mathrm{TMA} = 60 / 30\\mathrm{s}$  at the range of  $250 - 480^{\\circ}\\mathrm{C}$  of the removal step; and (c) C 1s XPS peak from  $\\mathrm{Al}_2\\mathrm{O}_3$  films after  $\\mathrm{NF}_3$  plasma exposure time at  $10\\mathrm{W}$  and  $100^{\\circ}\\mathrm{C}$  and after removal step at an elevated temperature range of  $280 - 480^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/28/fig_4","sample_id":"atomic-layer-etching/experimental-usecase/28/fig_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"multi spectra chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch per cycle as a function of TMA exposure time shows a rapid rise and then saturation, with 300 °C   removal-step temperature highest etch rate.\"},{\"panel_id\":\"b\",\"text\":\"The etch per cycle versus removal-step temperature increases sharply from low temperature, peaks around 300 °C, and then decreases at higher temperatures,\"},{\"panel_id\":\"c\",\"text\":\"XP-Spectrum shows that the peak intensity of C-C and C-Fx species increase with temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TMA exposure time (s) | Temperature (°C) | Etch per cycle (nm/cycle) |\\n|---|---|---|\\n| 0 | 290 | 0.0 |\\n| 0 | 300 | 0.0 |\\n| 0 | 330 | 0.0 |\\n| 10 | 290 | 0.097 |\\n| 10 | 300 | 0.17 |\\n| 10 | 330 | 0.14 |\\n| 20 | 290 | 0.17 |\\n| 20 | 300 | 0.25 |\\n| 20 | 330 | 0.23 |\\n| 30 | 290 | 0.20 |\\n| 30 | 300 | 0.30 |\\n| 30 | 330 | 0.27 |\\n| 40 | 290 | 0.21 |\\n| 40 | 300 | 0.30 |\\n| 40 | 330 | 0.27 |\"},{\"panel_id\":\"b\",\"text\":\"| Removal step temperature (°C) | Etch per cycle (nm/cycle) |\\n|---|---|\\n| 250 | 0.0 |\\n| 275 | 0.09 |\\n| 290 | 0.20 |\\n| 300 | 0.3 |\\n| 330 | 0.27 |\\n| 400 | 0.20 |\\n| 480 | 0.0 |\"},{\"panel_id\":\"c\",\"text\":\"| Measurement | Binding energy [eV] | Intensity (a.u.) | Assignment |\\n|---|---|---|---|\\n| NF3 plasma | 284.7 | 22 | C-C |\\n| NF3 plasma | 289.0 | 5 | C-Fx|\\n| 280 °C | 284.7 | 28 | C-C |\\n| 280 °C | 288.9 | 5 | C-Fx|\\n| 290 °C | 284.7 | 26 | C-C |\\n| 290 °C | 288.8 |4 | C-Fx|\\n| 300 °C | 284.7 | 35 | C-C |\\n| 300 °C | 288.8 | 5 | C-Fx|\\n| 400 °C | 284.7 | 51 | C-C |\\n| 400 °C | 288.7 | 9 | C-Fx |\\n| 480 °C | 284.7 | 58 | C-C |\\n| 480 °C | 288.7 | 10 | C-Fx |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"30 s (etch per cycle plateaus after 30 s).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"300 °C (fastest etching)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The EPC curve has a volcano-like shape because different mechanisms dominate at different temperatures. At low removal-step temperature, the etch per cycle is low because the ligand-exchange reaction between TMA and the fluorinated film is likely  kinetically hindered.  As the temperature increases into the intermediate range, this reaction proceeds more efficiently, leading to a higher EPC. At the highest temperatures, the EPC drops back to low values,  because TMA might be thermally decomposing, which hinders the reaction.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At the high removal-step temperatures (400 °C  and 480 °C ), the C 1s spectra show a much stronger C–C peak, compared to lower temperatures. This suggests that non-volatile carbon residues are accumulating on the surface with higher temperatures. Since TMA is the only significant source of carbon in the process, this strongly hints thermal decomposition of TMA at high temperatures.\"}]}]","bbox":[{"panel_id":"a","x":62,"y":5,"width":585,"height":481},{"panel_id":"b","x":72,"y":498,"width":577,"height":492},{"panel_id":"c","x":110,"y":1016,"width":536,"height":472}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/Atomic layer etching of Al2O3 with NF3 plasma fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":650,"height":1502,"image_format":"jpeg","image_sha256":"28f4b09264c023ab82c335565ab0372c7146c21bbfcaf4229bb31cf298e0e24f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_28_fig_5.jpg","caption":"FIG. 5. (a) Depth comparison of each half cycle of the fluorination step for  $60 \\text{s}$ $\\mathrm{NF}_3$  plasma at 10, 50, and  $100 \\text{W}$  and a temperature of  $100^{\\circ}\\text{C}$ , and the removal step with TMA for  $30 \\text{s}$  at an elevated temperature of  $300^{\\circ}\\text{C}$  and (b) surface atomic fraction tendency at each half cycle at  $10 \\text{W}$ .","id":"train/atomic-layer-etching/experimental-usecase/28/fig_5","sample_id":"atomic-layer-etching/experimental-usecase/28/fig_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between RF power and the depth of fluorinated layer and etched depth, demonstrating  both depths increasing with power, with the fluorinated layer always thicker than the removed thickness.\"},{\"panel_id\":\"b\",\"text\":\"The chart displays the atomic percentage of O1s, Al2p, F1s, and C1s across NF3 and TMA half-cycles. Percentage of fluorine is increasing after NF₃, then decreasing after TMA.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| RF power (W) | Fluorinated layer depth (nm) | Etched depth (nm) |\\n|---|---|---|\\n| 0 | 0.0 | 0.0 |\\n| 10 | 0.79 | 0.3 |\\n| 50 | 0.95 | 0.38 |\\n| 100 | 1.14 | 0.43 |\"},{\"panel_id\":\"b\",\"text\":\"| Cycle number | Step |O 1s | Al 2p | F 1s | C 1s |\\n|---|---|---|---|---|---|\\n| 0 | Start NF3  | 51 | 29 | 0 |18 |\\n| 0.5 | Start TMA  | 38 | 29 | 26 | 9 |\\n| 1 | Start NF3  | 44 | 30 |  15 | 10 |\\n| 1.5 | Start TMA | 39 | 26 | 25 | 9 |\\n| 2 | Start NF3  | 48 | 29 | 14 | 9 |\\n| 2.5 | Start TMA | 38 | 28 | 25 | 9 |\\n| 3 | Start NF3  | 48 | 30 | 13 | 9 |\\n| 3.5 | Start TMA | 38 | 28 | 26 | 9 |\\n| 4 | End TMA | 46 | 30 | 13 | 11 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TMA is not removing fluorine completely, as the etched depth is always smaller than the fluorinated-layer depth.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"50 W.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"NF₃ acts as the fluorination step: during the NF₃ half-cycles the F1s signal rises, showing that the surface is converted to an F-rich layer. TMA then serves as the removal step: in the TMA half-cycles the F1s signal decreases. This is consistent with TMA consuming surface fluorine to form volatile Al-containing products.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":72,"y":10,"width":588,"height":476},{"panel_id":"b","x":72,"y":522,"width":587,"height":477}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/28/Atomic layer etching of Al2O3 with NF3 plasma fluorination.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":663,"height":1000,"image_format":"jpeg","image_sha256":"73f6812f2da020ff72aa0b331d4f2c26db5e979ac5658118e5653d901eb2894f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_29_fig_7.jpg","caption":"FIG. 7. Normalized NBE signal intensity of GaN underneath the AlGaN film: (a) after the BT step, (b) after exposure to  $\\mathrm{Cl}_2$  plasma, and (c) after exposure to Ar plasma as a function of plasma exposure time.","id":"train/atomic-layer-etching/experimental-usecase/29/fig_7","sample_id":"atomic-layer-etching/experimental-usecase/29/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The normalized NBE signal intensity decreases with increasing plasma exposure time for both Ar plasma and Cl2 plasma.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Plasma Exposure Time (s) | Normalized NBE Signal Intensity (a.u.) Cl2 plasma| Normalized NBE Signal Intensity (a.u.) Ar plasma|\\n|---|---|---|\\n| 0 | 1.0 | 1.0|\\n| 50 | 0.5 |0.7|\\n| 100 | 0.45 |0.6|\\n| 200 | 0.35 |0.5|\\n| 300 | 0.25 |0.35|\\n| 350 | 0.25 |0.3|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both plasmas inflict ion bombardment, but only the Cl2 plasma also has a chemical effect on the substrate.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Especially for the Cl2 plasma curve there seems to be saturating behaviour, so yes.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 70 %.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.3\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":540,"height":528}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/Atomic layer etching of AlGaN using Cl2 and Ar gas.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":539,"height":528,"image_format":"jpeg","image_sha256":"366ab560be5b634b9b37c977a8aa6d66a3f236ca38bd633acec48c147a00d9df","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_33_fig_7.jpg","caption":"FIG. 7. Depth profiles of atomic concentrations of Ti, N, O, and C in the pristine TiN film. The thickness of the native oxide was approximately  $3\\mathrm{nm}$ .","id":"train/atomic-layer-etching/experimental-usecase/33/fig_7","sample_id":"atomic-layer-etching/experimental-usecase/33/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The atomic concentration of Ti, N, O, and C changes over sputtering time, with Ti and N reaching a plateau at 50% while O and C decrease significantly.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputtering Time (min.) | Atomic Concentration (%) |\\n|---|---|\\n| 0 | Ti: 25, N: 25, O: 28, C: 22 |\\n| 1 | Ti: 44, N: 44, O: 12, C: 0 |\\n| 2 | Ti: 49, N: 46, O: 5, C: 0 |\\n| 3 | Ti: 49, N: 46, O: 5, C: 0 |\\n| 4 | Ti: 49, N: 46, O: 5, C: 0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The native oxide is 3 nm thick, the oxygen levels out at 5% after 2 min of etching, so the etch rate is 0.25 angstrom/s.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The measured carbon is advantitious carbon and likely comes from storing in ambient conditions.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Ti:N ratio is 1:1.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, there is a constant level of oxygen at around 5%, indicating that there are impurities.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":0,"width":665,"height":535}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/33/Atomic layer etching of titanium nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":670,"height":538,"image_format":"jpeg","image_sha256":"dd16d306d8b6b91f7f56a8fbdf350b7acea677583ef15f49ba04f48bda9c387f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_39_fig_2.jpg","caption":"FIG. 2. (a) EPC vs cycle number with  $40 \\text{s} H_{2}$  plasma exposures only (triangles),  $40 \\text{s} SF_{6} / Ar$  plasma exposures only (squares), and both half-cycles (circles). All processes occur at  $0^{\\circ}C$ . The dashed lines are guides to the eye. (b) EPC versus  $H_{2}$  plasma exposure time with  $SF_{6} / Ar$  plasma exposure time fixed at  $30 \\text{s}$ . (c) EPC versus  $SF_{6} / Ar$  plasma exposure time with  $H_{2}$  plasma exposure time fixed at  $30 \\text{s}$ . The etch rates are observed to saturate with exposure time, demonstrating the self-limiting nature of the process.","id":"train/atomic-layer-etching/experimental-usecase/39/fig_2","sample_id":"atomic-layer-etching/experimental-usecase/39/fig_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Thickness change remains near zero over 50 cycles for H₂-only and SF₆/Ar-only plasma exposures, while the combined H₂+SF₆/Ar sequence produces an approximately linear thickness decrease reaching ~80 nm by 50 cycles.\"},{\"panel_id\":\"b\",\"text\":\"With SF₆/Ar exposure fixed at 30 s, the etch per cycle increases with H₂ plasma exposure time and saturates near ~1.45–1.5 nm/cycle beyond ~40 s.\"},{\"panel_id\":\"c\",\"text\":\"With H₂ exposure fixed at 30 s, the etch per cycle similarly increases with SF₆/Ar exposure time and approaches a plateau near ~1.5–1.6 nm/cycle, indicating saturation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Curve | Number of cycles | Thickness change (nm) |\\n|---|---|---|\\n| H₂ (triangles) | 50 | ~ -1.0 |\\n| SF₆ (squares) | 50 | ~ -2.0 |\\n| H₂ + SF₆ (circles) | 10 | ~ -12.0 |\\n| H₂ + SF₆ (circles) | 20 | ~ -32.0 |\\n| H₂ + SF₆ (circles) | 30 | ~ -51.0 |\\n| H₂ + SF₆ (circles) | 40 | ~ -67.0 |\\n| H₂ + SF₆ (circles) | 50 | ~ -80.0 |\"},{\"panel_id\":\"b\",\"text\":\"| H₂ exposure time (s) | Etch rate (nm/cycle) | \\n|---|---| \\n| 0 | ~0.0 | \\n| 10 | ~1.00 | \\n| 20 | ~1.30 | \\n| 30 | ~1.40 | \\n| 40 | ~1.45 | \\n| 50 | ~1.50 |\"},{\"panel_id\":\"c\",\"text\":\"| SF₆ exposure time (s) | Etch rate (nm/cycle) |\\n|---|---|\\n| 0 | ~0.00 |\\n| 10 | ~1.20 |\\n| 20 | ~1.30 |\\n| 30 | ~1.40 |\\n| 40 | ~1.50 |\\n| 50 | ~1.60 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The H₂-only and SF₆-only conditions show little to no net thickness change over 50 cycles, with both traces staying close to 0 nm. In contrast, the combined H₂+SF₆ sequence produces a large, cumulative thickness decrease that reaches ~80 nm by 50 cycles. The comparison is consistent with effective etching occurring only when both half-cycles are applied in sequence.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Saturation ensures that additional exposure does not increase material removal beyond a fixed amount per cycle. This makes the process insensitive to small timing variations and enables precise, repeatable etch control. Such behavior is a defining requirement for true atomic layer etching.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At approximately 40 s.\"}]}]","bbox":[{"panel_id":"a","x":8,"y":13,"width":583,"height":616},{"panel_id":"b","x":646,"y":8,"width":437,"height":305},{"panel_id":"c","x":647,"y":340,"width":436,"height":290}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/39/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/39/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/39/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/39/Isotropic atomic layer etching of MgO-doped lithium niobate.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"39","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1083,"height":630,"image_format":"jpeg","image_sha256":"e248092abbb544cd9536a98d9918d58d32c70c4a2c53470bade1ddb18d596e9f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_4_fig_2.jpg","caption":"FIG. 2. (a) Energy scan and (b) synergy scan of the ITO ALE process at  $150^{\\circ}\\mathrm{C}$ . Error bars in (a) and (b) show standard deviation from the mean, when multiple measurements are done with the same parameters.","id":"train/atomic-layer-etching/experimental-usecase/4/fig_2","sample_id":"atomic-layer-etching/experimental-usecase/4/fig_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between DC bias and EPC (Å) for two conditions: ALE and Ar milling only.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates the synergy percentage at different DC bias voltages.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| DC bias (-V) | EPC (Å) ALE | EPC (Å) Ar milling only |\\n|---|---|---|\\n| 0 | 0.1 | 0 |\\n| 6 | 0.5 |  |\\n| 9 | 1 | 0.1 |\\n| 12 | 1.5 | 0.25 |\\n| 18 | 1.5 | 0.5 |\"},{\"panel_id\":\"b\",\"text\":\"| DC bias (-V) | Synergy (%) |\\n|---|---|\\n| 0 |  |\\n| 6 |  |\\n| 9 | 80 |\\n| 12 | 75 |\\n| 18 | 55 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The resulting EPC as a function of RF power starts to plateau at 20-30 W. Meanwhile, when observing just the effect of the Ar ions, which is the case for \\\"Ar milling only\\\", the Ar ions become energetic enough to significantly etch the surface above 15 W. This means that extra care should be taken when analysing the process above 15 W, as Ar etching might influence the observed EPC.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.5 Å.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.5 Å.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"80%.\"}]}]","bbox":[{"panel_id":"a","x":37,"y":39,"width":647,"height":444},{"panel_id":"b","x":760,"y":41,"width":650,"height":441}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/4/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/4/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/4/Atomic layer etching of indium tin oxide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1414,"height":486,"image_format":"jpeg","image_sha256":"86fdde7580780b06a0adc43fd206b0299be6e2cf53fa7681dc7077a7aafc209c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_42_figure_10.jpg","caption":"Figure 10. Etch rates of  $\\mathrm{Al}_2\\mathrm{O}_3$  at different temperatures using (a)  $\\mathrm{Sn(acac)}_2,$  b) DMAC and c) TMA as the metal precursor.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_10","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Etch rate versus temperature is plotted for Al2O3 using TMA, DMAC, Sn(acac)2\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Sn(acac)₂ Etch Rate (Å/cycle) | DMAC Etch Rate (Å/cycle) | TMA Etch Rate (Å/cycle) |\\n|------------------|--------------------------------|---------------------------|--------------------------|\\n| 150              | 0.15                          | —                         | —                        |\\n| 175              | 0.20                          | —                         | —                        |\\n| 200              | 0.28                          | —                         | —                        |\\n| 225              | 0.40                          | 0.10                     | —                        |\\n| 250              | 0.50                          | 0.35                     | 0.15                    |\\n| 275              | —                              | 0.50                     | 0.30                    |\\n| 300              | —                              | 0.65                     | 0.45                    |\\n| 325              | —                              | —                         | 0.65                    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Al2O3 etching has different threshold temperatures for each metal precursor,  hence the etch rate changes for TMA and DMAC. The Al2O3 etch rates all increase at higher temperatures after exceeding the threshold temperature.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. Sn(acac)2 etched Al2O3 at temperatures ≥150 °C\\n2. In comparison, DMAC etched Al2O3 at higher temperatures ≥225 °C\\n3. TMA etched Al2O3 at even higher temperatures ≥250 °C\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The different temperature thresholds may be attributed to varying activation barriers for the ligand-exchange reactions. The\\ndifferent temperature thresholds for the etching may provide another method for obtaining selective thermal ALE.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":0,"width":668,"height":495}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/Selectivity in Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"42","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":672,"height":497,"image_format":"jpeg","image_sha256":"049df95b8761b42c5d6772a56c5d83b5e025ee93b049c770b7fc32044668ec3a","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_42_figure_9.jpg","caption":"Figure 9. Etch rates of  $\\mathrm{ZrO_2}$ $\\mathrm{HfO_2}$  and  $\\mathrm{Al}_2\\mathrm{O}_3$  versus temperature using  $\\mathrm{SiCl_4}$  as the metal precursor.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_9","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart plots the etch rate of SiCl4  versus temperature for ZrO2, HfO2, and Al2O3.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | ZrO₂ Etch Rate (Å/cycle) | HfO₂ Etch Rate (Å/cycle) | Al₂O₃ Etch Rate (Å/cycle) |\\n|------------------|---------------------------|---------------------------|---------------------------|\\n| 250              | 0.00                    | 0.00                    | 0.00                    |\\n| 300              | 0.055                 | 0.015                   | 0.005                   |\\n| 350              | 0.15                    | 0.06                  | 0.01                    |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. At 250 °C, Etch rate is almost 0\\n2. At 300 °C, Etch rate is 0.055 Å/cycle\\n3. At 350 °C, Etch rate is 0.15 Å/cycle\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The ligand-exchange reaction between SiCl4 and AlF3 is unfavorable at all the temperatures hence the etch rate is nil.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ligand-exchange reaction between SiCl4 and ZrF4 or HfF4 is feasible at >200 °C, hence increasing etch rates observed at >250 °C . The ZrO2 etch rate is higher than the HfO2 etch rate even though their predicted ΔG values are nearly equivalent. There must be additional kinetic factors that determine the actual etch rates\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":663,"height":496}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/Selectivity in Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"42","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":667,"height":503,"image_format":"jpeg","image_sha256":"8da4c1d4f1b1d5c74ac382dd05417d75bc130c1bffb5b37dc80f4575fe985212","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_43_fig_6.jpg","caption":"Fig. 6. ToF-SIMS depth profiles of  $\\mathrm{Al}_2\\mathrm{O}_3$  film after ALE process. The ALE cycle was repeated 5 times, and the samples were analyzed after the fluorination and removal steps. The  $\\mathrm{AlF_2^-}$  concentration profiles were plotted for the different ALE process temperatures of (a)  $200^{\\circ}\\mathrm{C}$ , (b)  $250^{\\circ}\\mathrm{C}$ , and (c)  $300^{\\circ}\\mathrm{C}$ . The  $\\mathrm{AlOF_2^-}$  concentration profiles were also plotted for (d)  $200^{\\circ}\\mathrm{C}$ , (e)  $250^{\\circ}\\mathrm{C}$ , and (f)  $300^{\\circ}\\mathrm{C}$  respectively.","id":"train/atomic-layer-etching/experimental-usecase/43/fig_6","sample_id":"atomic-layer-etching/experimental-usecase/43/fig_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multiple line chart"},{"panel_id":"f","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure shows Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) depth profiles of Al₂O₃ films after five Atomic Layer Etching (ALE) cycles, measured after the fluorination and removal steps at different process temperatures.AlF₂⁻ depth profile at 200 °C, showing fluorine incorporation near the surface after fluorination and its partial removal during the etching step.\"},{\"panel_id\":\"b\",\"text\":\"AlF₂⁻ depth profile at 250 °C, indicating enhanced fluorine penetration and modified removal behavior compared to 200 °C.\"},{\"panel_id\":\"c\",\"text\":\"AlF₂⁻ depth profile at 300 °C, exhibiting deeper fluorine incorporation and more pronounced temperature-dependent effects.\"},{\"panel_id\":\"d\",\"text\":\"AlOF₂⁻ depth profile at 200 °C, reflecting the formation of mixed oxyfluoride species confined near the surface.\"},{\"panel_id\":\"e\",\"text\":\"AlOF₂⁻ depth profile at 250 °C, showing increased formation and redistribution of oxyfluoride species.\"},{\"panel_id\":\"f\",\"text\":\"AlOF₂⁻ depth profile at 300 °C, indicating stronger oxyfluoride formation and deeper distribution at higher temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputter Time (s) | Intensity (c/s) | Method |\\n|---|---|---|\\n| 0 | 1E+02 | Fluorination |\\n| 0 | 1E+02 | Removal |\\n| 1 | 1.5E+02 | Fluorination |\\n| 1 | 1.5E+02 | Removal |\\n| 2 | 1E+02 | Fluorination |\\n| 2 | 1E+02 | Removal |\\n| 3 | 1.5E+01 | Fluorination |\\n| 3 | 1.5E+01 | Removal |\\n| 4 | 1E+01 | Fluorination |\\n| 4 | 1E+01 | Removal |\\n| 5 | 1.5E+00 | Fluorination |\\n| 5 | 1.5E+00 | Removal |\\n| 6 | 1E+00 | Fluorination |\\n| 6 | 1E+00 | Removal |\"},{\"panel_id\":\"b\",\"text\":\"| Sputter Time (s) | Intensity (c/s) | Method |\\n|---|---|---|\\n| 0 | 1E+02 | Fluorination |\\n| 0 | 1E+02 | Removal |\\n| 1 | 1.5E+02 | Fluorination |\\n| 1 | 1.4E+02 | Removal |\\n| 2 | 1E+02 | Fluorination |\\n| 2 | 9E+01 | Removal |\\n| 3 | 5E+01 | Fluorination |\\n| 3 | 4E+01 | Removal |\\n| 4 | 1E+01 | Fluorination |\\n| 4 | 8E+00 | Removal |\\n| 5 | 1.5E+00 | Fluorination |\\n| 5 | 1.4E+00 | Removal |\\n| 6 | 1.2E+00 | Fluorination |\\n| 6 | 1.1E+00 | Removal |\"},{\"panel_id\":\"c\",\"text\":\"| Sputter Time (s) | Intensity (c/s) | Method |\\n|---|---|---|\\n| 0 | 1E+02 | Fluorination |\\n| 0 | 8E+01 | Removal |\\n| 1 | 1.5E+02 | Fluorination |\\n| 1 | 1.3E+02 | Removal |\\n| 2 | 1E+02 | Fluorination |\\n| 2 | 8E+01 | Removal |\\n| 3 | 5E+01 | Fluorination |\\n| 3 | 3E+01 | Removal |\\n| 4 | 1E+01 | Fluorination |\\n| 4 | 8E+00 | Removal |\\n| 5 | 1.5E+00 | Fluorination |\\n| 5 | 1.3E+00 | Removal |\\n| 6 | 1.2E+00 | Fluorination |\\n| 6 | 1E+00 | Removal |\"},{\"panel_id\":\"d\",\"text\":\"| Sputter Time (s) | Intensity (c/s) | Method |\\n|---|---|---|\\n| 0 | 1E+02 | Fluorination |\\n| 0 | 1E+02 | Removal |\\n| 1 | 1.5E+02 | Fluorination |\\n| 1 | 1.5E+02 | Removal |\\n| 2 | 1E+02 | Fluorination |\\n| 2 | 1E+02 | Removal |\\n| 3 | 1.5E+01 | Fluorination |\\n| 3 | 1.5E+01 | Removal |\\n| 4 | 1E+01 | Fluorination |\\n| 4 | 1E+01 | Removal |\\n| 5 | 1.5E+00 | Fluorination |\\n| 5 | 1.5E+00 | Removal |\\n| 6 | 1E+00 | Fluorination |\\n| 6 | 1E+00 | Removal |\"},{\"panel_id\":\"e\",\"text\":\"| Sputter Time (s) | Intensity (c/s) | Method |\\n|---|---|---|\\n| 0 | 1E+02 | Fluorination |\\n| 0 | 1E+02 | Removal |\\n| 1 | 1.5E+02 | Fluorination |\\n| 1 | 1.4E+02 | Removal |\\n| 2 | 1E+02 | Fluorination |\\n| 2 | 9E+01 | Removal |\\n| 3 | 5E+01 | Fluorination |\\n| 3 | 4E+01 | Removal |\\n| 4 | 1E+01 | Fluorination |\\n| 4 | 8E+00 | Removal |\\n| 5 | 1.5E+00 | Fluorination |\\n| 5 | 1.4E+00 | Removal |\\n| 6 | 1.2E+00 | Fluorination |\\n| 6 | 1.1E+00 | Removal |\"},{\"panel_id\":\"f\",\"text\":\"| Sputter Time (s) | Intensity (c/s) | Method |\\n|---|---|---|\\n| 0 | 1E+02 | Fluorination |\\n| 0 | 8E+01 | Removal |\\n| 1 | 1.5E+02 | Fluorination |\\n| 1 | 1.3E+02 | Removal |\\n| 2 | 1E+02 | Fluorination |\\n| 2 | 8E+01 | Removal |\\n| 3 | 5E+01 | Fluorination |\\n| 3 | 3E+01 | Removal |\\n| 4 | 1E+01 | Fluorination |\\n| 4 | 8E+00 | Removal |\\n| 5 | 1.5E+00 | Fluorination |\\n| 5 | 1.3E+00 | Removal |\\n| 6 | 1.2E+00 | Fluorination |\\n| 6 | 1E+00 | Removal |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 200°C, the overlapping curves indicate that the TMA cannot overcome the activation barrier to effectively react with or volatilize the AlF_x layer, leaving the surface structure largely unchanged. At 300°C, the significant drop in signal confirms that the thermal energy enables the ligand-exchange mechanism to function, restoring the surface closer to the underlying bulk oxide composition by stripping away the fluorinated layer.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The drop is caused by the Ligand-Exchange Reaction with Trimethylaluminum (TMA).\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process is not viable at 200°C. The trade-off for the low thermal budget is a complete failure of the etch mechanism; the temperature must be raised to at least 250°C to activate the ligand exchange.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"NO, While the fluorination is shallow at 200°C, the AlOF₂⁻ profile (d) shows the removal step is incomplete, leaving a modified interface. For a 2 nm film, this residual modified layer constitutes a significant fraction of the total thickness, drastically altering the interface properties. A higher temperature (like 250°C) with complete removal, despite slightly deeper penetration, would better preserve the electrical properties of the underlying interface.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":6,"width":491,"height":436},{"panel_id":"b","x":491,"y":2,"width":365,"height":438},{"panel_id":"c","x":854,"y":4,"width":366,"height":440},{"panel_id":"d","x":5,"y":512,"width":483,"height":442},{"panel_id":"e","x":487,"y":505,"width":365,"height":445},{"panel_id":"f","x":848,"y":510,"width":381,"height":442}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/43/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/43/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/43/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/43/Surface reaction during thermal atomic layer etching of aluminum oxide.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"43","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1230,"height":955,"image_format":"jpeg","image_sha256":"374d7ca7bbd1ec7f3331f23410f29b55c0aa9b7ed35d30dc6b3cc7fed85874e9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_45_figure_2.jpg","caption":"Figure 2. Mass changes during 10 consecutive HF exposures on  $\\mathrm{TiO}_2$  ALD film at 200, 250, and  $300^{\\circ}\\mathrm{C}$","id":"train/atomic-layer-etching/experimental-usecase/45/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/45/figure_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the mass change over time for 10 HF exposures on ALD TiO₂ at different temperatures (200°C, 250°C, and 300°C). The mass decreases over time, indicating a loss of material\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | 200°C (Mass Change, ng/cm^2) | 250°C (Mass Change, ng/cm^2) | 300°C (Mass Change, ng/cm^2) |\\n|---|---|---|---|\\n| 0 | 0 | 0 | 0 |\\n| 50 | -20 | -25 | -28 |\\n| 100 | -35 | -45 | -55 |\\n| 150 | -50 | -75 | -85 |\\n| 200 | -60 | -95 | -120 |\\n| 250 | -75 | -105 | -125 |\\n| 300 | -85 | -125 | -155 |\\n| 350 | -95 | -150 | -185 |\\n| 400 | -105 | -175 | -190 |\\n| 450 | -115 | -200 | -240 |\\n| 500 | -125 | -220 | -250 |\\n| 550 | -135 | -245 | -280 |\\n| 600 | -160 | -255 | -310 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Each step represents one discrete HF exposure pulse, with the flat regions corresponding to N₂ purge periods between exposures.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Higher temperatures increase the mass loss per HF exposure: −17 ng/cm² at 200°C, −25 ng/cm² at 250°C, and −32 ng/cm² at 300°C. This corresponds to TiO₂ removal of 0.45–0.85 Å per exposure, based on TiO₂ density of 3.76 g/cm³.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This result validates the core mechanism: if HF can spontaneously etch TiO₂ to form volatile TiF₄ and H₂O, then oxidizing TiN to TiO₂ first enables its removal. Without this spontaneous etching, the oxidation-fluorination ALE approach would not work.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The reaction has ΔG° = −6.1 kcal/mol at 25°C (favorable) but becomes +6.3 kcal/mol at 250°C (unfavorable under standard conditions). However, etching still occurs because excess HF and rapid removal of volatile products (TiF₄, H₂O) by N₂ purge drive the reaction forward.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":2,"width":667,"height":521}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/Thermal Atomic Layer Etching of Titanium Nitride.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"45","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":667,"height":523,"image_format":"jpeg","image_sha256":"3722465db5d82bf7c8f79e7e0640c7b16e0303179a870adf8e7b55730136ab39","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_534d04cd0f885c3fdd5f4aaca3bde4aea20cf962e714ca66eae14e7e8af77611.jpg","caption":"","id":"train/atomic-layer-etching/experimental-usecase/46/534d04cd0f885c3fdd5f4aaca3bde4aea20cf962e714ca66eae14e7e8af77611","sample_id":"atomic-layer-etching/experimental-usecase/46/534d04cd0f885c3fdd5f4aaca3bde4aea20cf962e714ca66eae14e7e8af77611","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents quartz crystal microbalance (QCM) data showing mass changes during chemical exposures at 170 °C. In panel (a), only BCl₃ or WF₆ dosing leads to mass gain or negligible change, and WF₆/Sn(acac)₂ cycling results in net deposition, whereas alternating WF₆ and BCl₃ produces a continuous linear mass loss, confirming atomic layer etching (ALE) of TiO₂.\"},{\"panel_id\":\"b\",\"text\":\"The image zooms into the ALE process, revealing a sawtooth profile in which the WF₆ pulse increases mass due to surface fluorination, followed by a sharp mass decrease during the BCl₃ pulse from ligand exchange and volatilization, yielding net material removal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass loading (ng/cm²) | Sample |\\n|---------:|----------------------:|--------|\\n| 16000 | 4228.46 | WF₆/BCl₃ |\\n| 16000 | 4664.16 | WF₆/Sn(acac)₂ |\\n| 16000 | 5358.22 | WF₆ doses |\\n| 16000 | 6120.32 | BCl₃ doses |\\n| 17000 | 4483.84 | WF₆/BCl₃ |\\n| 17000 | 4919.33 | WF₆/Sn(acac)₂ |\\n| 17000 | 5722.26 | WF₆ doses |\\n| 17000 | 6341.46 | BCl₃ doses |\\n| 18000 | 4773.03 | WF₆/BCl₃ |\\n| 18000 | 5211.92 | WF₆/Sn(acac)₂ |\\n| 18000 | 6011.45 | WF₆ doses |\\n| 18000 | 6776.95 | BCl₃ doses |\\n| 19000 | 4555.29 | WF₆/BCl₃ |\\n| 19000 | 5647.41 | WF₆/Sn(acac)₂ |\\n| 19000 | 6341.46 | WF₆ doses |\\n| 19000 | 6668.08 | BCl₃ doses |\\n| 20000 | 4300.12 | WF₆/BCl₃ |\\n| 20000 | 5613.38 | WF₆/Sn(acac)₂ |\\n| 20000 | 6375.48 | WF₆ doses |\\n| 20000 |  | BCl₃ doses |\\n| 21000 | 3827.21 | WF₆/BCl₃ |\\n| 21000 | 5501.11 | WF₆/Sn(acac)₂ |\\n| 21000 | 6304.04 | WF₆ doses |\\n| 21000 |  | BCl₃ doses |\\n| 22000 | 3279.45 | WF₆/BCl₃ |\\n| 22000 | 5245.94 | WF₆/Sn(acac)₂ |\\n| 22000 | 6266.16 | WF₆ doses |\\n| 22000 |  | BCl₃ doses |\\n| 23000 | 2877.99 | WF₆/BCl₃ |\\n| 23000 | 5137.07 | WF₆/Sn(acac)₂ |\\n| 23000 |  | WF₆ doses |\\n| 23000 |  | BCl₃ doses |\\n| 24000 | 2333.63 | WF₆/BCl₃ |\\n| 24000 | 4956.75 | WF₆/Sn(acac)₂ |\\n| 24000 |  | WF₆ doses |\\n| 24000 |  | BCl₃ doses |\"},{\"panel_id\":\"b\",\"text\":\"| ALE cycle number | Mass loading (ng/cm²) | Pulse |\\n|-----------------:|----------------------:|-------|\\n| 20   | 4559.39 | WF₆ |\\n| 20   | 4629.66 | WF₆ |\\n| 20.5 | 4526.56 | BCl₃ |\\n| 20.5 | 4615.65 | BCl₃ |\\n| 21   | 4507.95 | WF₆ |\\n| 21   | 4542.97 | WF₆ |\\n| 21.5 | 4468.11 | BCl₃ |\\n| 21.5 | 4521.96 | BCl₃ |\\n| 22   | 4453.88 | WF₆ |\\n| 22   | 4491.53 | WF₆ |\\n| 22.5 | 4414.04 | BCl₃ |\\n| 22.5 | 4468.11 | BCl₃ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It would fail. The black line shows a positive slope (mass increase) over time, indicating that combining WF₆ and Sn(acac)₂ results in chemical vapor deposition (CVD) or film growth rather than etching. Substituting BCl₃ with Sn(acac)₂ would therefore add material instead of removing it.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 170 °C, WF₆ can fluorinate the TiO₂ surface, creating non-volatile metal fluorides or oxyfluorides and adding mass. However, without a second reagent such as BCl₃ to drive a ligand-exchange reaction and convert these surface species into volatile products (e.g., TiClₓ or volatile oxychlorides), the modified surface layer remains solid, blocking further etching.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Modification step: WF₆ exposure (mass increase).\\nRemoval step: BCl₃ exposure (mass decrease).\\nIdentification: The BCl₃ exposure is the removal step, indicated by the sharp drop in mass below the starting baseline of that cycle.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Cycles: Approximately 167 cycles (100 Å / 0.6 Å·cycle⁻¹). It is preferred for its digital controllability. A continuous wet etch is time-dependent and prone to under- or over-etching due to diffusion and timing errors, whereas ALE removes a fixed amount per cycle independent of time, enabling precise thickness control by simply counting cycles.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":6,"width":468,"height":370},{"panel_id":"b","x":481,"y":6,"width":469,"height":369}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/534d04cd0f885c3fdd5f4aaca3bde4aea20cf962e714ca66eae14e7e8af77611.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/534d04cd0f885c3fdd5f4aaca3bde4aea20cf962e714ca66eae14e7e8af77611.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"not_found"},"width":950,"height":375,"image_format":"jpeg","image_sha256":"61d4c4054b1aad81ed67b40ce28db4469fe2bb82d1ab1cd749ae613b468327ad","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_1.jpg","caption":"Figure 1. QCM analysis at  $220^{\\circ}\\mathrm{C}$  of (a) 200 cycles of  $\\mathrm{Al}_2\\mathrm{O}_3$  or  $\\mathrm{TiO}_2$  followed immediately by  $50\\mathrm{WF}_6 / \\mathrm{Ar}$  exposures of  $1 / 60\\mathrm{s}$  each and (b) initial loading of  $1\\mathrm{WF}_6$  exposure of  $50\\mathrm{s}$  versus  $50\\mathrm{WF}_6 / \\mathrm{Ar}$  exposures of  $1 / 60\\mathrm{s}$  on  $\\mathrm{TiO}_2$  (200 cycles).","id":"train/atomic-layer-etching/experimental-usecase/46/figure_1","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_1","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents QCM data demonstrating material selectivity during WF₆ exposure at 220 °C. The TiO₂ film (red line) shows continuous mass loss, indicating etching, whereas the Al₂O₃ film (black line) exhibits an initial mass gain followed by saturation, indicating no etching.\"},{\"panel_id\":\"b\",\"text\":\"QCM data comparing WF₆ dosing strategies on TiO₂. Pulsed WF₆ dosing (red line) results in cumulative mass loss (etching), while the initial exposure (black line) is dominated by strong adsorption and fluorination, seen as a large mass gain before etching kinetics dominate.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass loading (ng/cm²) | Sample | Condition |\\n|---------:|----------------------:|--------|-----------|\\n| 8000  | 3713.81 | Al₂O₃ | Metal Oxide ALD |\\n| 8000  | 2935.80 | TiO₂ | Metal Oxide ALD |\\n| 8500  | 3920.90 | Al₂O₃ | Metal Oxide ALD |\\n| 8500  | 3025.36 | TiO₂ | Metal Oxide ALD |\\n| 9000  | 4130.79 | Al₂O₃ | Metal Oxide ALD |\\n| 9000  | 3084.13 | TiO₂ | Metal Oxide ALD |\\n| 9500  | 4729.68 | Al₂O₃ | WF₆ Doses |\\n| 9500  | 3294.02 | TiO₂ | WF₆ Doses |\\n| 10000 | 4788.45 | Al₂O₃ | WF₆ Doses |\\n| 10000 | 3145.70 | TiO₂ | WF₆ Doses |\\n| 10500 | 4819.24 | Al₂O₃ | WF₆ Doses |\\n| 10500 | 3025.36 | TiO₂ | WF₆ Doses |\\n| 11000 | 4788.45 | Al₂O₃ | WF₆ Doses |\\n| 11000 | 2756.70 | TiO₂ | WF₆ Doses |\\n| 11500 | 4757.67 | Al₂O₃ | WF₆ Doses |\\n| 11500 | 2216.57 | TiO₂ | WF₆ Doses |\\n| 12000 | 4757.67 | Al₂O₃ | WF₆ Doses |\\n| 12000 | 1351.82 | TiO₂ | WF₆ Doses |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Mass loading (ng/cm²) | Condition |\\n|---------:|----------------------:|-----------|\\n| 18200 | 1546.58 | TiO₂ ALD |\\n| 18200 | 816.87  | 220 °C, 1.75 Torr |\\n| 18300 | 1566.23 | TiO₂ ALD |\\n| 18300 | 857.10  | 220 °C, 1.75 Torr |\\n| 18400 | 1566.23 | TiO₂ ALD |\\n| 18400 | 887.04  | 220 °C, 1.75 Torr |\\n| 18500 | 1806.66 | 50 WF₆ doses (1 s) |\\n| 18500 | 1106.89 | 1 WF₆ dose (50 s) |\\n| 18600 | 1796.37 | 50 WF₆ doses (1 s) |\\n| 18600 | 1117.18 | 1 WF₆ dose (50 s) |\\n| 18700 | 1776.72 | 50 WF₆ doses (1 s) |\\n| 18700 | 1136.82 | 1 WF₆ dose (50 s) |\\n| 18800 | 1726.21 | 50 WF₆ doses (1 s) |\\n| 18800 | 1136.82 | 1 WF₆ dose (50 s) |\\n| 18900 | 1706.56 | 50 WF₆ doses (1 s) |\\n| 18900 | 1147.11 | 1 WF₆ dose (50 s) |\\n| 19000 | 1676.72 | 50 WF₆ doses (1 s) |\\n| 19000 | 1156.47 | 1 WF₆ dose (50 s) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Selective pattern transfer. The TiO₂ film will be etched away, as indicated by the continuous mass loss (negative slope) in the red curve. The Al₂O₃ mask will remain intact and essentially unetched, as indicated by the black curve, which shows only stable mass gain due to surface fluorination. This confirms that WF₆ provides excellent selectivity for removing TiO₂ relative to Al₂O₃.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etchability vs. inertness. TiO₂ (red line) is superior as a sacrificial spacer because it spontaneously etches in WF₆ at 220 °C, whereas Al₂O₃ (black line) is inert and resistant, meaning it would not be removed, failing the sacrificial requirement.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Surface fluorination/adsorption. The WF₆ molecules react with surface oxygen to form a heavy layer of non-volatile metal fluorides (TiFₓ) or tungsten oxyfluorides (WOₓFᵧ). This accumulation of heavy F and W atoms adds mass to the crystal before the slow volatilization step can remove any material.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Continuous vapor etching (CVE). True ALE requires a self-limiting mechanism where the etch stops after removing one layer. The red curve shows a steep, continuous linear drop in mass as long as WF₆ is pulsed, indicating that at 220 °C, the thermal energy is sufficient to drive the reaction continuously without stopping. This is a CVE process, not a self-limiting ALE process.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":525,"height":416},{"panel_id":"b","x":506,"y":2,"width":542,"height":414}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1050,"height":422,"image_format":"jpeg","image_sha256":"10dae58b0704ae6d7f16fdf1e4f49ff207483d2b585d8b96b6528f342a31cd17","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_10.jpg","caption":"Figure 10. (a) QCM analysis at  $170^{\\circ}\\mathrm{C}$  and 1.75 Torr of  $\\mathrm{TiO_2}$  ALD followed by  $25\\mathrm{BCl}_3 / \\mathrm{Ar}$  doses (pink),  $50\\mathrm{WF}_6 / \\mathrm{Ar}$  doses (red),  $50\\mathrm{WF}_6 / \\mathrm{Ar} / \\mathrm{Sn}(\\mathrm{acac})_2 / \\mathrm{Ar}$  cycles (black), or  $50\\mathrm{WF}_6 / \\mathrm{Ar} / \\mathrm{BCl}_3 / \\mathrm{Ar}$  cycles (blue). (b) Enlarged view of the  $\\mathrm{WF}_6 / \\mathrm{Ar} / \\mathrm{BCl}_3 / \\mathrm{Ar}$  etch sequence. (c) QCM of  $\\mathrm{WF}_6 / \\mathrm{Ar} / \\mathrm{BCl}_3 / \\mathrm{Ar}$  sequence at 110, 130, 150, 170, and  $190^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/46/figure_10","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"line chart"},{"panel_id":"c","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents quartz crystal microbalance (QCM) data showing mass changes during chemical exposures at 170 °C. In panel (a), only BCl₃ or WF₆ dosing leads to mass gain or negligible change, and WF₆/Sn(acac)₂ cycling results in net deposition, whereas alternating WF₆ and BCl₃ produces a continuous linear mass loss, confirming atomic layer etching (ALE) of TiO₂.\"},{\"panel_id\":\"b\",\"text\":\"The image zooms into the ALE process, revealing a sawtooth profile in which the WF₆ pulse increases mass due to surface fluorination, followed by a sharp mass decrease during the BCl₃ pulse from ligand exchange and volatilization, yielding net material removal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass loading (ng/cm²) | Sample |\\n|---------:|----------------------:|--------|\\n| 16000 | 4228.46 | WF₆/BCl₃ |\\n| 16000 | 4664.16 | WF₆/Sn(acac)₂ |\\n| 16000 | 5358.22 | WF₆ doses |\\n| 16000 | 6120.32 | BCl₃ doses |\\n| 17000 | 4483.84 | WF₆/BCl₃ |\\n| 17000 | 4919.33 | WF₆/Sn(acac)₂ |\\n| 17000 | 5722.26 | WF₆ doses |\\n| 17000 | 6341.46 | BCl₃ doses |\\n| 18000 | 4773.03 | WF₆/BCl₃ |\\n| 18000 | 5211.92 | WF₆/Sn(acac)₂ |\\n| 18000 | 6011.45 | WF₆ doses |\\n| 18000 | 6776.95 | BCl₃ doses |\\n| 19000 | 4555.29 | WF₆/BCl₃ |\\n| 19000 | 5647.41 | WF₆/Sn(acac)₂ |\\n| 19000 | 6341.46 | WF₆ doses |\\n| 19000 | 6668.08 | BCl₃ doses |\\n| 20000 | 4300.12 | WF₆/BCl₃ |\\n| 20000 | 5613.38 | WF₆/Sn(acac)₂ |\\n| 20000 | 6375.48 | WF₆ doses |\\n| 20000 |  | BCl₃ doses |\\n| 21000 | 3827.21 | WF₆/BCl₃ |\\n| 21000 | 5501.11 | WF₆/Sn(acac)₂ |\\n| 21000 | 6304.04 | WF₆ doses |\\n| 21000 |  | BCl₃ doses |\\n| 22000 | 3279.45 | WF₆/BCl₃ |\\n| 22000 | 5245.94 | WF₆/Sn(acac)₂ |\\n| 22000 | 6266.16 | WF₆ doses |\\n| 22000 |  | BCl₃ doses |\\n| 23000 | 2877.99 | WF₆/BCl₃ |\\n| 23000 | 5137.07 | WF₆/Sn(acac)₂ |\\n| 23000 |  | WF₆ doses |\\n| 23000 |  | BCl₃ doses |\\n| 24000 | 2333.63 | WF₆/BCl₃ |\\n| 24000 | 4956.75 | WF₆/Sn(acac)₂ |\\n| 24000 |  | WF₆ doses |\\n| 24000 |  | BCl₃ doses |\"},{\"panel_id\":\"b\",\"text\":\"| ALE cycle number | Mass loading (ng/cm²) | Pulse |\\n|-----------------:|----------------------:|-------|\\n| 20   | 4559.39 | WF₆ |\\n| 20   | 4629.66 | WF₆ |\\n| 20.5 | 4526.56 | BCl₃ |\\n| 20.5 | 4615.65 | BCl₃ |\\n| 21   | 4507.95 | WF₆ |\\n| 21   | 4542.97 | WF₆ |\\n| 21.5 | 4468.11 | BCl₃ |\\n| 21.5 | 4521.96 | BCl₃ |\\n| 22   | 4453.88 | WF₆ |\\n| 22   | 4491.53 | WF₆ |\\n| 22.5 | 4414.04 | BCl₃ |\\n| 22.5 | 4468.11 | BCl₃ |\"},{\"panel_id\":\"c\",\"text\":\"| Time (s) | Mass loading (ng/cm²) | Condition                     | Temperature |\\n|---------:|----------------------:|-------------------------------|-------------|\\n| 3000 | 696.55  | TiO₂ ALD (TiCl₄/H₂O) | 110 °C |\\n| 3000 | 759.00  | TiO₂ ALD (TiCl₄/H₂O) | 130 °C |\\n| 3000 | 905.00  | TiO₂ ALD (TiCl₄/H₂O) | 150 °C |\\n| 3000 | 821.45  | TiO₂ ALD (TiCl₄/H₂O) | 170 °C |\\n| 3000 | 852.67  | TiO₂ ALD (TiCl₄/H₂O) | 190 °C |\\n| 4000 | 1051.71 | TiO₂ ALD (TiCl₄/H₂O) | 110 °C |\\n| 4000 | 1062.44 | TiO₂ ALD (TiCl₄/H₂O) | 130 °C |\\n| 4000 | 1114.16 | TiO₂ ALD (TiCl₄/H₂O) | 150 °C |\\n| 4000 | 1093.67 | TiO₂ ALD (TiCl₄/H₂O) | 170 °C |\\n| 4000 | 1093.67 | TiO₂ ALD (TiCl₄/H₂O) | 190 °C |\\n| 5000 | 1260.51 | TiO₂ ALE (WF₆/BCl₃) | 110 °C |\\n| 5000 | 1240.02 | TiO₂ ALE (WF₆/BCl₃) | 130 °C |\\n| 5000 | 1166.85 | TiO₂ ALE (WF₆/BCl₃) | 150 °C |\\n| 5000 | 1093.67 | TiO₂ ALE (WF₆/BCl₃) | 170 °C |\\n| 5000 | 1020.49 | TiO₂ ALE (WF₆/BCl₃) | 190 °C |\\n| 6000 | 1135.62 | TiO₂ ALE (WF₆/BCl₃) | 110 °C |\\n| 6000 | 1082.93 | TiO₂ ALE (WF₆/BCl₃) | 130 °C |\\n| 6000 | 947.31  | TiO₂ ALE (WF₆/BCl₃) | 150 °C |\\n| 6000 | 727.78  | TiO₂ ALE (WF₆/BCl₃) | 170 °C |\\n| 6000 | 518.98  | TiO₂ ALE (WF₆/BCl₃) | 190 °C |\\n| 7000 |          | TiO₂ ALE (WF₆/BCl₃) | 110 °C |\\n| 7000 | 925.85  | TiO₂ ALE (WF₆/BCl₃) | 130 °C |\\n| 7000 | 800.96  | TiO₂ ALE (WF₆/BCl₃) | 150 °C |\\n| 7000 | 414.58  | TiO₂ ALE (WF₆/BCl₃) | 170 °C |\\n| 7000 | -55.70  | TiO₂ ALE (WF₆/BCl₃) | 190 °C |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"WF₆ dose\\n\\nAr purge\\n\\nBCl₃ dose\\n\\nAr purge\\n\\nTemperature control is critical because it directly governs the reaction kinetics and ensures the process remains self-limiting.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The self-limiting ALE process at 170°C is required for perfect conformality in high-aspect-ratio structures. Its layer-by-layer nature ensures uniform etching down the trench sidewalls. The trade-off is a much slower etch rate (~0.6 Å/cycle) compared to the rapid, isotropic material removal of the continuous CVE process at 220°C.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate (mass loss per cycle) increases significantly with temperature. This is due to the Arrhenius behavior of the surface reactions: higher thermal energy accelerates the fluorination of TiO₂ by WF₆ and the subsequent removal of the modified layer by BCl₃, leading to more complete reactions per cycle.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This high selectivity means the ALE process can precisely and completely remove the TiO₂ spacer while using the Al₂O₃ layer as a reliable etch stop. This enables the fabrication of more scaled transistor architectures by allowing the selective etching of one oxide without damaging other critical oxide layers in the stack.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":513,"height":410},{"panel_id":"b","x":534,"y":0,"width":509,"height":414},{"panel_id":"c","x":217,"y":449,"width":576,"height":428}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1044,"height":878,"image_format":"jpeg","image_sha256":"e80895aae58796e05f33c82fdf19aac9a1d900379e57147db803f5e64f57dafb","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_11.jpg","caption":"Figure 11. Thermodynamic modeling results showing the expected equilibrium species concentrations from 25 to  $400^{\\circ}C$  for (a) reactions 2 and 3, exposing  $\\mathbb{N}_2$  -diluted  $\\mathrm{BCl}_3$  at 1.5 Torr to  $\\mathrm{TiF_4 / WO_3}$  ..  $3\\mathrm{mol}$  of  $\\mathrm{BCl}_3(\\mathrm{g}) + 1$  mol of  $\\mathrm{TiF_4(s) + 1}$  mol of  $\\mathrm{WO}_3(\\mathrm{s})$  ; and (b) the analogous reactions for the  $\\mathrm{Al}_2\\mathrm{O}_3$  system: 3 mol of  $\\mathrm{BCl}_3(\\mathrm{g}) + 1$  mol of  $\\mathrm{AlF}_3(\\mathrm{s}) + 1$  mol of  $\\mathrm{WO}_3(\\mathrm{s})$  . At equilibrium at  $170^{\\circ}\\mathrm{C},$ $\\mathrm{BCl}_3 + \\mathrm{TiF}_4 / \\mathrm{WO}_3$  produces predominantly volatile  $\\mathrm{BF}_3,$ $\\mathrm{TiCl}_4,$  and  $\\mathrm{WOCl_4}$  and solid  $\\mathrm{B}_2\\mathrm{O}_3$","id":"train/atomic-layer-etching/experimental-usecase/46/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_11","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"TiO₂ System: At 170 °C, the reaction of BCl₃ with fluorinated TiO₂ surface species produces exclusively volatile products such as TiCl₄ (black line) and WOCl₄ (red line), along with solid B₂O₃. The absence of solid titanium species confirms that etching is thermodynamically favorable.\"},{\"panel_id\":\"b\",\"text\":\"Al₂O₃ System: At 170 °C, aluminum remains as solid AlF₃ (purple line), and significant formation of volatile AlCl₃ (black line) occurs only above 300 °C. This explains the selectivity: Ti-fluorides volatilize at low temperatures, whereas Al-fluorides do not.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Equilibrium Phase Concentration (mol) | Material |\\n|-----------------:|---------------------------------------:|----------|\\n| 0   | 1.30 | BF₃ |\\n| 0   | 1.00 | TiCl₄ |\\n| 0   | 0.65 | B₂O₃ |\\n| 0   | 0.30 | BCl₃ |\\n| 0   | 0.05 | WO₃ |\\n| 0   | 0.90 | WOCl₄ |\\n| 0   | 0.00 | WO₂Cl₂ |\\n| 100 | 1.30 | BF₃ |\\n| 100 | 1.00 | TiCl₄ |\\n| 100 | 0.65 | B₂O₃ |\\n| 100 | 0.30 | BCl₃ |\\n| 100 | 0.05 | WO₃ |\\n| 100 | 0.90 | WOCl₄ |\\n| 100 | 0.00 | WO₂Cl₂ |\\n| 200 | 1.30 | BF₃ |\\n| 200 | 1.00 | TiCl₄ |\\n| 200 | 0.65 | B₂O₃ |\\n| 200 | 0.32 | BCl₃ |\\n| 200 | 0.05 | WO₃ |\\n| 200 | 0.85 | WOCl₄ |\\n| 200 | 0.05 | WO₂Cl₂ |\\n| 300 | 1.30 | BF₃ |\\n| 300 | 1.00 | TiCl₄ |\\n| 300 | 0.60 | B₂O₃ |\\n| 300 | 0.40 | BCl₃ |\\n| 300 | 0.05 | WO₃ |\\n| 300 | 0.60 | WOCl₄ |\\n| 300 | 0.30 | WO₂Cl₂ |\\n| 350 | 1.30 | BF₃ |\\n| 350 | 1.00 | TiCl₄ |\\n| 350 | 0.45 | B₂O₃ |\\n| 350 | 0.55 | BCl₃ |\\n| 350 | 0.05 | WO₃ |\\n| 350 | 0.35 | WOCl₄ |\\n| 350 | 0.60 | WO₂Cl₂ |\\n| 400 | 1.30 | BF₃ |\\n| 400 | 1.00 | TiCl₄ |\\n| 400 | 0.35 | B₂O₃ |\\n| 400 | 0.80 | BCl₃ |\\n| 400 | 0.05 | WO₃ |\\n| 400 | 0.15 | WOCl₄ |\\n| 400 | 0.90 | WO₂Cl₂ |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (°C) | Equilibrium Phase Concentration (mol) | Material |\\n|-----------------:|---------------------------------------:|----------|\\n| 0   | 1.65 | BCl₃ |\\n| 0   | 1.00 | AlF₃ |\\n| 0   | 0.65 | B₂O₃ |\\n| 0   | 0.05 | AlCl₃ |\\n| 0   | 0.85 | WOCl₄ |\\n| 0   | 0.00 | WO₂Cl₂ |\\n| 0   | 0.00 | BF₃ |\\n| 100 | 1.65 | BCl₃ |\\n| 100 | 1.00 | AlF₃ |\\n| 100 | 0.65 | B₂O₃ |\\n| 100 | 0.05 | AlCl₃ |\\n| 100 | 0.85 | WOCl₄ |\\n| 100 | 0.00 | WO₂Cl₂ |\\n| 100 | 0.00 | BF₃ |\\n| 200 | 1.65 | BCl₃ |\\n| 200 | 1.00 | AlF₃ |\\n| 200 | 0.65 | B₂O₃ |\\n| 200 | 0.05 | AlCl₃ |\\n| 200 | 0.85 | WOCl₄ |\\n| 200 | 0.05 | WO₂Cl₂ |\\n| 200 | 0.05 | BF₃ |\\n| 300 | 1.68 | BCl₃ |\\n| 300 | 1.00 | AlF₃ |\\n| 300 | 0.60 | B₂O₃ |\\n| 300 | 0.05 | AlCl₃ |\\n| 300 | 0.60 | WOCl₄ |\\n| 300 | 0.30 | WO₂Cl₂ |\\n| 300 | 0.20 | BF₃ |\\n| 350 | 1.70 | BCl₃ |\\n| 350 | 1.00 | AlF₃ |\\n| 350 | 0.45 | B₂O₃ |\\n| 350 | 0.05 | AlCl₃ |\\n| 350 | 0.35 | WOCl₄ |\\n| 350 | 0.60 | WO₂Cl₂ |\\n| 350 | 0.60 | BF₃ |\\n| 400 | 1.50 | BCl₃ |\\n| 400 | 1.00 | AlF₃ |\\n| 400 | 0.30 | B₂O₃ |\\n| 400 | 0.05 | AlCl₃ |\\n| 400 | 0.20 | WOCl₄ |\\n| 400 | 0.75 | WO₂Cl₂ |\\n| 400 | 0.90 | BF₃ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiCl₄ (Titanium Tetrachloride).Temperature: It is stable and maximized across the entire temperature range shown (from 25 °C upwards), indicating very favorable thermodynamics.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Thermodynamic selectivity. At 200 °C, the TiCl₄ concentration in panel (a) is high (≈1.0 mol), indicating that titanium is efficiently driven into the gas phase. In contrast, in panel (b), the AlCl₃ concentration is effectively zero, with aluminum remaining trapped as solid AlF₃, providing a robust thermodynamic etch stop.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"From the tungsten oxide (WO₃). The WO₃ acts as an oxygen source, which is scavenged by boron from BCl₃ to form stable solid B₂O₃, while tungsten forms volatile oxychlorides (WOCl₄).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies BCl₃ acts as the coreactant that removes the modified surface layer. After WF₆ fluorinates TiO₂ to form a TiF₄-like surface, BCl₃ exchanges fluorine for chlorine, producing volatile TiCl₄ and regenerating the surface for the next WF₆ dose. The thermodynamic drive for this exchange exists for TiO₂ but not for Al₂O₃, enabling selective ALE.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":2,"width":498,"height":403},{"panel_id":"b","x":507,"y":5,"width":502,"height":400}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1011,"height":408,"image_format":"jpeg","image_sha256":"4e750a2adac5f0a2abb9c48a053711a9a25472604d5de62359c63c7307ba32ab","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_12.jpg","caption":"Figure 12. (a) QCM analysis of  $\\mathrm{Al}_2\\mathrm{O}_3$  and  $\\mathrm{TiO}_2$  ALD followed by  $50\\mathrm{WF}_6 / \\mathrm{BCl}_3$  ALE cycles at  $170^{\\circ}\\mathrm{C}$ . The ALE sequence follows  $\\mathrm{WF}_6 / \\mathrm{Ar} / \\mathrm{BCl}_3 / \\mathrm{Ar}$ $(0.2 / 45 / 2.5 / 60\\mathrm{s})$ . (b) Enlarged view of (a).","id":"train/atomic-layer-etching/experimental-usecase/46/figure_12","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_12","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure uses Quartz Crystal Microbalance (QCM) data to demonstrate the selective etching of Titanium Dioxide (TiO₂) over Aluminum Oxide (Al₂O₃) at 170 °C. The image shows the full process timeline. Both materials are first deposited via ALD (staircase growth). Upon switching to the ALE recipe (WF₆/BCl₃), TiO₂ (red line) immediately begins to lose mass linearly (etching), while Al₂O₃ (black line) gains a small amount of mass and then remains stable (resistant).\"},{\"panel_id\":\"b\",\"text\":\"A magnified view of the first few ALE cycles. TiO₂ loses mass in every cycle (sawtooth down), whereas Al₂O₃ gains mass during the WF₆ pulse and barely loses it during the BCl₃ pulse, confirming the formation of a stable, non-volatile fluoride layer on aluminum.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass loading (ng cm⁻²) | Material | Process condition |\\n|---------:|-----------------------:|----------|-------------------|\\n| 8000  | 1500 | Al₂O₃ | Metal oxide ALD |\\n| 8000  | 950  | TiO₂  | Metal oxide ALD |\\n| 8500  | 1700 | Al₂O₃ | Metal oxide ALD |\\n| 8500  | 1050 | TiO₂  | Metal oxide ALD |\\n| 9000  | 2100 | Al₂O₃ | Metal oxide ALD |\\n| 9000  | 1200 | TiO₂  | Metal oxide ALD |\\n| 9500  | 2500 | Al₂O₃ | WF₆/BCl₃ ALE |\\n| 9500  | 1100 | TiO₂  | WF₆/BCl₃ ALE |\\n| 10000 | 2500 | Al₂O₃ | WF₆/BCl₃ ALE |\\n| 10000 | 900  | TiO₂  | WF₆/BCl₃ ALE |\\n| 10500 | 2500 | Al₂O₃ | WF₆/BCl₃ ALE |\\n| 10500 | 750  | TiO₂  | WF₆/BCl₃ ALE |\\n| 11000 | 2500 | Al₂O₃ | WF₆/BCl₃ ALE |\\n| 11000 | 650  | TiO₂  | WF₆/BCl₃ ALE |\\n| 11500 | 2500 | Al₂O₃ | WF₆/BCl₃ ALE |\\n| 11500 | 500  | TiO₂  | WF₆/BCl₃ ALE |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Mass loading (ng cm⁻²) | Material | Cyclic condition |\\n|---------:|-----------------------:|----------|------------------|\\n| 9000 | 1500 | Al₂O₃ | ALD  |\\n| 9000 | 1200 | TiO₂  | ALD  |\\n| 9100 | 1550 | Al₂O₃ | ALD |\\n| 9100 | 1250 | TiO₂  | ALD |\\n| 9200 | 1850 | Al₂O₃ | WF₆  |\\n| 9200 | 1400 | TiO₂  | WF₆  |\\n| 9250 | 1800 | Al₂O₃ | BCl₃ |\\n| 9250 | 1250 | TiO₂  | BCl₃ |\\n| 9300 | 1850 | Al₂O₃ | WF₆  |\\n| 9300 | 1300 | TiO₂  | WF₆  |\\n| 9350 | 1800 | Al₂O₃ | BCl₃  |\\n| 9350 | 1200 | TiO₂  | BCl₃  |\\n| 9400 | 1850 | Al₂O₃ | WF₆  |\\n| 9400 | 1250 | TiO₂  | WF₆  |\\n| 9450 | 1800 | Al₂O₃ | BCl₃ |\\n| 9450 | 1150 | TiO₂  | BCl₃  |\\n| 9500 | 1850 | Al₂O₃ | WF₆ |\\n| 9500 | 1200 | TiO₂  | WF₆ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Mass jump / no etch. No, it does not lose mass. Instead, it shows a distinct upward jump (mass gain) of ~300 ng/cm² initially, followed by a stable plateau. This indicates surface fluorination (adsorption) rather than material removal.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The thermal WF₆/BCl₃ ALE process is superior. Its purely chemical, self-limiting nature ensures perfect conformality in high-aspect-ratio structures without the directional ion damage inherent to plasma etching, which can degrade sidewall quality and introduce defects in sensitive nanostructures.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TiO₂ mass decreases because its fluorination/chlorination products (e.g., TiF₄ and TiCl₄) are volatile at 170°C, allowing material removal. Al₂O₃ mass remains stable because its potential reaction products (e.g., AlF₃, AlCl₃) are non-volatile at this temperature, causing the surface reactions to self-passivate rather than etch.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies the process can precisely remove the TiO₂ spacer while using the Al₂O₃ layer as a perfect etch stop. This enables the creation of self-aligned, scaled transistor architectures by selectively etching one material without damaging the other, a critical capability for advanced manufacturing.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":527,"height":417},{"panel_id":"b","x":512,"y":4,"width":538,"height":411}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_12.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1053,"height":422,"image_format":"jpeg","image_sha256":"2b60cb3d505de57d297c9a99cad071bcb6c4dc1f3c35120e88066d9e6421b77c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_2.jpg","caption":"Figure 2. Thermodynamic modeling results showing the expected equilibrium species concentrations from 25 to  $400^{\\circ}\\mathrm{C}$  for  $\\mathrm{N}_2$ -diluted  $\\mathrm{WF}_6$  exposed to  $\\mathrm{TiO}_2$  or  $\\mathrm{Al}_2\\mathrm{O}_3$  at  $P = 1.5$  Torr: (a)  $1\\mathrm{mol}$ $\\mathrm{WF}_6 + 1\\mathrm{mol}$ $\\mathrm{TiO}_2$  and (b)  $1\\mathrm{mol}$ $\\mathrm{WF}_6 + 0.667\\mathrm{mol}$ $\\mathrm{Al}_2\\mathrm{O}_3$ .","id":"train/atomic-layer-etching/experimental-usecase/46/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Equilibrium analysis for the WF₆–TiO₂ system shows a temperature-dependent transition from solid TiF₄(s) to gaseous TiF₄(g) above ~120 °C, enabling etching; at higher temperatures, tungsten oxide byproducts such as WO₃(s) convert to volatile species (e.g., WF₂O₂(g)).\"},{\"panel_id\":\"b\",\"text\":\"Equilibrium analysis for the WF₆–Al₂O₃ system shows that AlF₃(s) remains solid and stable over the entire temperature range, indicating fluorination without volatilization or etching.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Equilibrium Concentration (mol) | Material        |\\n|------------------|----------------------------------|----------------|\\n| 0   | 1.00 | TiF₄ (s) |\\n| 0   | 0.00 | TiF₄ (g) |\\n| 0   | 0.65 | WO₃ (s)  |\\n| 0   | 0.30 | WF₆ (g)  |\\n| 0   | 0.00 | WF₂O₂ (g) |\\n| 50  | 1.00 | TiF₄ (s) |\\n| 50  | 0.00 | TiF₄ (g) |\\n| 50  | 0.65 | WO₃ (s)  |\\n| 50  | 0.30 | WF₆ (g)  |\\n| 50  | 0.00 | WF₂O₂ (g) |\\n| 100 | 0.00 | TiF₄ (s) |\\n| 100 | 1.00 | TiF₄ (g) |\\n| 100 | 0.65 | WO₃ (s)  |\\n| 100 | 0.30 | WF₆ (g)  |\\n| 100 | 0.00 | WF₂O₂ (g) |\\n| 150 | 0.00 | TiF₄ (s) |\\n| 150 | 1.00 | TiF₄ (g) |\\n| 150 | 0.60 | WO₃ (s)  |\\n| 150 | 0.25 | WF₆ (g)  |\\n| 150 | 0.10 | WF₂O₂ (g) |\\n| 200 | 0.00 | TiF₄ (s) |\\n| 200 | 1.00 | TiF₄ (g) |\\n| 200 | 0.40 | WO₃ (s)  |\\n| 200 | 0.20 | WF₆ (g)  |\\n| 200 | 0.40 | WF₂O₂ (g) |\\n| 250 | 0.00 | TiF₄ (s) |\\n| 250 | 1.00 | TiF₄ (g) |\\n| 250 | 0.10 | WO₃ (s)  |\\n| 250 | 0.05 | WF₆ (g)  |\\n| 250 | 0.90 | WF₂O₂ (g) |\\n| 300 | 0.00 | TiF₄ (s) |\\n| 300 | 1.00 | TiF₄ (g) |\\n| 300 | 0.00 | WO₃ (s)  |\\n| 300 | 0.00 | WF₆ (g)  |\\n| 300 | 1.00 | WF₂O₂ (g) |\\n| 350 | 0.00 | TiF₄ (s) |\\n| 350 | 1.00 | TiF₄ (g) |\\n| 350 | 0.00 | WO₃ (s)  |\\n| 350 | 0.00 | WF₆ (g)  |\\n| 350 | 1.00 | WF₂O₂ (g) |\\n| 400 | 0.00 | TiF₄ (s) |\\n| 400 | 1.00 | TiF₄ (g) |\\n| 400 | 0.00 | WO₃ (s)  |\\n| 400 | 0.00 | WF₆ (g)  |\\n| 400 | 1.00 | WF₂O₂ (g) |\"},{\"panel_id\":\"b\",\"text\":\"| Temperature (°C) | Equilibrium Concentration (mol) | Material |\\n|------------------|----------------------------------|----------------|\\n| 0   | 1.30 | AlF₃ (s) |\\n| 0   | 0.65 | WO₃ (s)  |\\n| 0   | 0.30 | WF₆ (g)  |\\n| 0   | 0.00 | WF₂O₂ (g) |\\n| 50  | 1.30 | AlF₃ (s) |\\n| 50  | 0.65 | WO₃ (s)  |\\n| 50  | 0.30 | WF₆ (g)  |\\n| 50  | 0.00 | WF₂O₂ (g) |\\n| 100 | 1.30 | AlF₃ (s) |\\n| 100 | 0.65 | WO₃ (s)  |\\n| 100 | 0.30 | WF₆ (g)  |\\n| 100 | 0.00 | WF₂O₂ (g) |\\n| 150 | 1.30 | AlF₃ (s) |\\n| 150 | 0.60 | WO₃ (s)  |\\n| 150 | 0.25 | WF₆ (g)  |\\n| 150 | 0.10 | WF₂O₂ (g) |\\n| 200 | 1.30 | AlF₃ (s) |\\n| 200 | 0.40 | WO₃ (s)  |\\n| 200 | 0.20 | WF₆ (g)  |\\n| 200 | 0.40 | WF₂O₂ (g) |\\n| 250 | 1.30 | AlF₃ (s) |\\n| 250 | 0.10 | WO₃ (s)  |\\n| 250 | 0.05 | WF₆ (g)  |\\n| 250 | 0.90 | WF₂O₂ (g) |\\n| 300 | 1.30 | AlF₃ (s) |\\n| 300 | 0.00 | WO₃ (s)  |\\n| 300 | 0.00 | WF₆ (g)  |\\n| 300 | 1.00 | WF₂O₂ (g) |\\n| 350 | 1.30 | AlF₃ (s) |\\n| 350 | 0.00 | WO₃ (s)  |\\n| 350 | 0.00 | WF₆ (g)  |\\n| 350 | 1.00 | WF₂O₂ (g) |\\n| 400 | 1.30 | AlF₃ (s) |\\n| 400 | 0.00 | WO₃ (s)  |\\n| 400 | 0.00 | WF₆ (g)  |\\n| 400 | 1.00 | WF₂O₂ (g) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 120°C.Reason: Below this temperature, the byproduct TiF4 is solid. Instead of removing material (etching), the process would simply convert the surface into a crust of solid fluoride salts, potentially increasing the film thickness and creating particles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The equilibrium model may predict low volatility at lower temperatures (e.g., 170°C). However, ALE bypasses this by using a sequential, non-equilibrium process. The BCl₃ step in ALE actively removes the fluorinated surface layer (e.g., as TiCl₄), shifting the local equilibrium and driving the reaction forward kinetically, enabling etching at temperatures where a continuous, equilibrium-driven process would stall.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The modeling predicts that the TiO₂ system forms volatile products like TiF₄(g) and WOFₓ(g), while the Al₂O₃ system forms non-volatile solids like AlF₃(s) or Al₂O₃(s) remains. This difference in product volatility is the thermodynamic origin of etching selectivity: TiO₂ can be removed as gases, whereas Al₂O₃ forms a stable, non-volatile layer that halts the reaction.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This provides a built-in etch stop capability. Engineers can design a stack where a thin Al₂O₃ layer is placed beneath or adjacent to the TiO₂ layer intended for removal. The WF₆-based etch will spontaneously stop on the Al₂O₃, allowing for precise, self-aligned patterning without the need for complex real-time endpoint detection, simplifying integration and improving yield.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":527,"height":423},{"panel_id":"b","x":529,"y":4,"width":522,"height":420}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1053,"height":428,"image_format":"jpeg","image_sha256":"0be6dc1b88b9796a265dbd90fc2cb612b9ab88b6174c0af8f83688e69f57103c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_3.jpg","caption":"Figure 3. (a) QCM analysis of  $\\mathrm{TiO_2}$  ALD followed by  $50~\\mathrm{WF}_6$  doses at 120 and  $220^{\\circ}C$  and (b) an enlarged view of the mass loading during the  $\\mathrm{WF}_6$  doses, showing net mass loss during each  $\\mathrm{WF}_6$  step at  $220^{\\circ}C$  and little to no change at  $120^{\\circ}C$","id":"train/atomic-layer-etching/experimental-usecase/46/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart represents the QCM data showing TiO₂ mass changes during ALD and subsequent WF₆ doses at 120°C and 220°C; only the 220°C sample exhibits noticeable mass loss per dose, indicating active etching.\"},{\"panel_id\":\"b\",\"text\":\"Magnified view of the WF₆ dosing period; 220°C shows stepwise mass loss (etching), while 120°C shows negligible change, confirming the reaction is thermally activated.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Loading (ng/cm²) | Temperature (°C) | Condition |\\n|---------:|-----------------------|------------------|-----------|\\n| 18000 | 5200 | 120 |TiO₂ ALD |\\n| 18000 | 4000 | 220 |TiO₂ ALD |\\n| 18500 | 5500 | 120 | TiO₂ ALD |\\n| 18500 | 4200 | 220 | TiO₂ ALD |\\n| 19000 | 5650 | 120 | WF₆ dose  |\\n| 19000 | 4100 | 220 | WF₆ dose  |\\n| 19500 | 5750 | 120 |  WF₆ dose |\\n| 19500 | 3950 | 220 | WF₆ dose  |\\n| 20000 | 5800 | 120 | WF₆ dose  |\\n| 20000 | 3700 | 220 | WF₆ dose  |\\n| 20500 | 5800 | 120 | WF₆ dose  |\\n| 20500 | 3300 | 220 | WF₆ dose  |\\n| 21000 | 5800 | 120 | WF₆ dose  |\\n| 21000 | 2600 | 220 | WF₆ dose  |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Mass Loading (ng/cm²) | Temperature (°C) | Condition |\\n|---------:|-----------------------|------------------|-----------|\\n| 19920 | 5820 | 120 | |\\n| 19920 | 5720 | 220 | |\\n| 19940 | 5820 | 120 | WF₆ |\\n| 19940 | 5736 | 220 | WF₆|\\n| 19960 | 5820 | 120 | |\\n| 19960 | 5679 | 220 | |\\n| 20000 | 5820 | 120 | WF₆ |\\n| 20000 | 5700 | 220 | WF₆ |\\n| 20040 | 5820 | 120 | |\\n| 20040 | 5650 | 220 | |\\n| 20060 | 5820 | 120 | WF₆ |\\n| 20060 | 5658 | 220 | WF₆ |\\n| 20080 | 5820 | 120 | |\\n| 20080 | 5600 | 220 | |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Deposit a TiO₂ film on the QCM sensor via ALD.\\n\\nStabilize the system at 220°C and the desired pressure.\\n\\nExpose the film to a sequence of 50 WF₆ doses, each followed by an inert purge.\\n\\nContinuously record the QCM frequency (converted to mass loading) throughout the dose sequence.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The lower temperature (120°C) is forced to protect the polymers. The trade-off is that no etching occurs at this temperature, so the desired material removal cannot be achieved with thermal WF₆ alone. An alternative approach, such as using a more reactive chemistry or a non-thermal activation method (like plasma), would be required, introducing complexity or potential damage.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies that a purely thermal WF₆ etch process is not compatible with such a low thermal budget. Since significant etching only occurs at 220°C, exceeding the 150°C limit, integration would require a fundamental process change. Options include performing the etch earlier in the flow (before temperature-sensitive layers) or developing a plasma-assisted WF₆ process that could lower the required substrate temperature.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Temperature drastically affects the etching rate. At 220°C, etching proceeds readily, while at 120°C, it is negligible. This is because the reaction is thermally activated; sufficient thermal energy is required to overcome the activation barrier for breaking Ti–O bonds and forming volatile titanium fluoride (TiF₄) and tungsten oxyfluoride products.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":530,"height":403},{"panel_id":"b","x":523,"y":7,"width":521,"height":394}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1050,"height":408,"image_format":"jpeg","image_sha256":"fe41672f1fc68795d2966f97ed76e1eb88a99d414fc1971156e50480ef554a6e","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_46_figure_7.jpg","caption":"Figure 7. (a) QCM analysis of  $\\mathrm{WF}_6$  etching following 50, 100, and  $200\\mathrm{TiO}_2$  ALD cycles on a bare Au QCM crystal at  $220^{\\circ}C$  and 1.75 Torr. (b) Enlarged view of  $\\mathrm{WF}_6$  etching showing an \"incubation\" period for the vapor etching that is dependent on the initial  $\\mathrm{TiO}_2$  film thickness.","id":"train/atomic-layer-etching/experimental-usecase/46/figure_7","sample_id":"atomic-layer-etching/experimental-usecase/46/figure_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"QCM data showing TiO₂ etching for different thicknesses (50, 100, 200 ALD cycles) at 220 °C. Thin films etch immediately, while thick films show an incubation period (mass gain/plateau) before etching due to crystallization increasing chemical resistance.\"},{\"panel_id\":\"b\",\"text\":\"This mutliple line chart focusing on the mass loading of TiO₂ cycles and WF₆ doses over time, highlighting the effect of 50 doses at specific intervals.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Mass Loading (ng/cm²) | Condition |\\n|---------:|----------------------|-----------|\\n| 0     | 0       | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 0     | 0       | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 0     | 0       | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 2000  | 250     | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 2000  | 90     | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 2000  |355     | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 4000  | 552     | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 4000  | 405     | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 4000  | 700     | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 6000  | 234     | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 6000  | 675    | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 6000  | 1100    | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 8000  | 220    | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 8000  | 1555    | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 8000  | 1500    | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 10000 |  | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 10000 | 300     | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 10000 | 2000    | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 12000 |  | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 12000 | 0      | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 12000 | 2100    | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 16000 |  | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 16000 |  | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 16000 | 3400    | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 18000 |  | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 18000 | | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 18000 | 3800    | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 22000 | | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 22000 | | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 22000 | 3200    | 200 TiO₂ cycles + 50 WF₆ doses |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Mass Loading (ng/cm²) | Condition |\\n|---------:|----------------------|-----------|\\n| 200  | 500  | 50 TiO₂ cycles |\\n| 200  | 1700 | 100 TiO₂ cycles |\\n| 200  | 3300 | 200 TiO₂ cycles |\\n| 400  | 590  | 50 TiO₂ cycles |\\n| 400  | 1872 | 100 TiO₂ cycles |\\n| 400  | 3332 | 200 TiO₂ cycles |\\n| 600  | 680  | 50 TiO₂ cycles |\\n| 600  | 2096 | 100 TiO₂ cycles |\\n| 600  | 3500 | 200 TiO₂ cycles |\\n| 800  | 340  | 50 TiO₂ cycles + 50 WF₆ doses  |\\n| 800  | 1850 | 100 TiO₂ cycles + 50 WF₆ doses|\\n| 800  | 3600 | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 1000 | 325  | 50 TiO₂ cycles + 50 WF₆ doses|\\n| 1000 | 1700 | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 1000 | 3620 | 200 TiO₂ cycles + 50 WF₆ doses|\\n| 1200 | 280  | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 1200 | 1500 | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 1200 | 3600 | 200 TiO₂ cycles + 50 WF₆ doses |\\n| 1400 | 258  | 50 TiO₂ cycles + 50 WF₆ doses |\\n| 1400 | 1200 | 100 TiO₂ cycles + 50 WF₆ doses |\\n| 1400 | 3600 | 200 TiO₂ cycles + 50 WF₆ doses |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The WF₆ vapor etch process introduces a significant challenge. In an ideal ALE process, each cycle removes a fixed thickness regardless of initial film thickness. Here, the thickness-dependent incubation means etching would initiate at different times across the wafer if the starting thickness varies, leading to severe non-uniformity. The trade-off for using this simpler vapor etch is loss of uniformity control.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Surface fluorination. The WF₆ reacts with the surface to form solid fluorides (TiFₓ) or adsorbed tungsten species. Because the underlying film is resistant (incubation), these heavy species accumulate on the surface, adding mass before the volatilization reaction (TiF₄ ↑) initiates material removal.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Crystalline resistance. The 50-cycle film is likely amorphous (disordered structure), making it chemically vulnerable and allowing immediate etching. The 200-cycle film has likely crystallized into a denser phase (e.g., Anatase), which is thermodynamically more stable and resistant to chemical attack, requiring an incubation period to roughen the surface and create active sites before rapid etching can proceed.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It implies the process requires a robust endpoint detection method and likely an over-etch period. Since the etch start time is unpredictable based on thickness, one cannot rely on timed etching alone. The process must monitor for the end of the incubation/etch transition (e.g., via mass spectrometry or optical emission) and then continue for a controlled over-etch to ensure complete clearance across all thickness variations.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":548,"height":415},{"panel_id":"b","x":522,"y":2,"width":522,"height":408}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/46/Thermal Selective Vapor Etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"46","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1047,"height":417,"image_format":"jpeg","image_sha256":"7e6d8e84e7eaedd45ddbf1f4e5a218e49e4efb48704cd166603ce7bea28d0168","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_15_fig4.jpg","caption":"FIG.4.   t  t 3.  t 300 4.  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  a  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  h  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  s  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  u  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  e  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  o  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  c  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  b  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  n  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  w  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  f  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  i  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  y  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  r  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  d  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  m  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  g  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  1  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t  t","id":"train/atomic-layer-etching/simulation-usecase/15/fig4","sample_id":"atomic-layer-etching/simulation-usecase/15/fig4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multiple line chart"},{"panel_id":"f","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multiple line graph shows the density of Si and Cl atoms at different Z positions along the y-axis, indicating that after the first chlorination step Cl remains near the surface while Si dominates the bulk.\"},{\"panel_id\":\"b\",\"text\":\"The multiple line graph shows the density of Si and Cl atoms at different Z positions along the y-axis, indicating that after the first ion bombardment step the near-surface Si density decreases and the Cl profile becomes slightly redistributed.\"},{\"panel_id\":\"c\",\"text\":\"The multiple line graph shows the density of Si and Cl atoms at different Z positions along the y-axis, indicating that after the fourth chlorination step Cl penetration into the near-surface region increases while Si density near the surface is further reduced.\"},{\"panel_id\":\"d\",\"text\":\"The multiple line graph shows the density of Si and Cl atoms at different Z positions along the y-axis, indicating that after the fourth ion bombardment step Si is further sputtered from the surface and the Cl distribution becomes more diffuse.\"},{\"panel_id\":\"e\",\"text\":\"The multiple line graph shows the density of Si and Cl atoms at different Z positions along the y-axis, indicating that after the eighth chlorination step a pronounced Cl-rich surface layer forms with a corresponding depletion of Si near the interface.\"},{\"panel_id\":\"f\",\"text\":\"The multiple line graph shows the density of Si and Cl atoms at different Z positions along the y-axis, indicating that after the eighth ion bombardment step the surface Si density continues to decrease while the Cl profile is partially flattened due to repeated sputtering and mixing.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z Position (Å) |  Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.075 | 0.000 |\\n| 5  | 0.048 | 0.000 |\\n| 10 | 0.049 | 0.000 |\\n| 15 | 0.047 | 0.000 |\\n| 20 | 0.052 | 0.000 |\\n| 25 | 0.049 | 0.000 |\\n| 30 | 0.050 | 0.000 |\\n| 35 | 0.049 | 0.000 |\\n| 40 | 0.049 | 0.000 |\\n| 45 | 0.050 | 0.001 |\\n| 50 | 0.049 | 0.002 |\\n| 55 | 0.002 | 0.068 |\\n| 60 | 0.000 | 0.000 |\"},{\"panel_id\":\"b\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.070 | 0.000 |\\n| 5  | 0.048 | 0.000 |\\n| 10 | 0.050 | 0.000 |\\n| 15 | 0.049 | 0.000 |\\n| 20 | 0.052 | 0.000 |\\n| 25 | 0.050 | 0.000 |\\n| 30 | 0.050 | 0.000 |\\n| 35 | 0.052 | 0.001 |\\n| 40 | 0.050 | 0.002 |\\n| 45 | 0.048 | 0.003 |\\n| 50 | 0.038 | 0.005 |\\n| 55 | 0.012 | 0.005 |\\n| 60 | 0.000 | 0.000 |\"},{\"panel_id\":\"c\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.075 | 0.000 |\\n| 5  | 0.046 | 0.000 |\\n| 10 | 0.048 | 0.000 |\\n| 15 | 0.049 | 0.000 |\\n| 20 | 0.051 | 0.000 |\\n| 25 | 0.050 | 0.000 |\\n| 30 | 0.052 | 0.002 |\\n| 35 | 0.048 | 0.004 |\\n| 40 | 0.050 | 0.003 |\\n| 45 | 0.042 | 0.004 |\\n| 50 | 0.035 | 0.015 |\\n| 55 | 0.008 | 0.008 |\\n| 60 | 0.000 | 0.000 |\"},{\"panel_id\":\"d\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.075 | 0.000 |\\n| 5  | 0.045 | 0.000 |\\n| 10 | 0.049 | 0.000 |\\n| 15 | 0.048 | 0.000 |\\n| 20 | 0.051 | 0.000 |\\n| 25 | 0.050 | 0.000 |\\n| 30 | 0.050 | 0.001 |\\n| 35 | 0.051 | 0.003 |\\n| 40 | 0.052 | 0.004 |\\n| 45 | 0.050 | 0.003 |\\n| 50 | 0.042 | 0.003 |\\n| 55 | 0.010 | 0.003 |\\n| 60 | 0.000 | 0.000 |\"},{\"panel_id\":\"e\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.072 | 0.000 |\\n| 5  | 0.048 | 0.000 |\\n| 10 | 0.046 | 0.000 |\\n| 15 | 0.048 | 0.000 |\\n| 20 | 0.050 | 0.000 |\\n| 25 | 0.051 | 0.000 |\\n| 30 | 0.050 | 0.001 |\\n| 35 | 0.050 | 0.003 |\\n| 40 | 0.052 | 0.002 |\\n| 45 | 0.045 | 0.004 |\\n| 50 | 0.038 | 0.009 |\\n| 55 | 0.010 | 0.015 |\\n| 60 | 0.000 | 0.000 |\"},{\"panel_id\":\"f\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.073 | 0.000 |\\n| 5  | 0.047 | 0.000 |\\n| 10 | 0.049 | 0.000 |\\n| 15 | 0.048 | 0.000 |\\n| 20 | 0.050 | 0.000 |\\n| 25 | 0.051 | 0.000 |\\n| 30 | 0.049 | 0.001 |\\n| 35 | 0.048 | 0.004 |\\n| 40 | 0.051 | 0.003 |\\n| 45 | 0.046 | 0.004 |\\n| 50 | 0.040 | 0.006 |\\n| 55 | 0.009 | 0.005 |\\n| 60 | 0.000 | 0.000 |\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It implies a Steady-State Regime. Since the thickness of the damaged/mixed layer stops growing and remains constant from Cycle 4 to Cycle 8, the process has stabilized. This means the Etch Rate per Cycle (EPC) is now constant and predictable, which is the ideal operating condition for manufacturing precision devices.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Amorphization (or Lattice Damage). The vertical drop in (a) represents an atomically sharp, ordered crystal. The gradual slope in (f) indicates that the repeated ion bombardment has disordered the lattice, creating a region of low-density amorphous silicon and mixing it with the surface, rather than maintaining a perfect crystal termination.\"}]},{\"panel_id\":\"f\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It represents the Total Etch Depth (or material removed). The receding boundary shows that the cyclic ALE process is successfully removing silicon layer-by-layer, resulting in a 10 \\\\AA trench depth after 8 cycles.\"}]},{\"panel_id\":\"e\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Contamination / Defect Generation. The tail shows that Chlorine is trapped inside the silicon lattice, not just on top. If a new layer is grown on this surface, the trapped Chlorine acts as a contaminant. It can diffuse out, causing corrosion, void formation, or electrical defects that ruin the crystal quality of the subsequent epitaxial growth.\"}]}]","bbox":[{"panel_id":"a","x":22,"y":8,"width":409,"height":370},{"panel_id":"b","x":468,"y":8,"width":410,"height":370},{"panel_id":"c","x":931,"y":8,"width":408,"height":370},{"panel_id":"d","x":22,"y":389,"width":409,"height":367},{"panel_id":"e","x":469,"y":389,"width":408,"height":367},{"panel_id":"f","x":929,"y":389,"width":418,"height":367}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/Molecular dynamics study of silicon atomic layer etching by chlorine gas and argon ions.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"15","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1347,"height":756,"image_format":"jpeg","image_sha256":"22c8441ecb4cb5b943ce8770507896f95fc80e80502568f1c917b076c9e5e309","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_15_fig_9.jpg","caption":"FIG. 9. Si and Cl number density profiles at the beginning of  $\\mathsf{Ar}^+$  ion bombardment step [panel (a)] and after  $150\\mathrm{Ar}^+$  ion impacts [panel (b)]. The black dotted line corresponds to the density of Si at  $300\\mathrm{K}$  (Ref. 43). The cycle in this case corresponds to  $2255\\mathrm{Cl}_2$  molecule impacts and  $1000\\mathrm{Ar}^+$  ion impacts.","id":"train/atomic-layer-etching/simulation-usecase/15/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/15/fig_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure displays the depth profiles (density vs. Z-position) of Silicon (blue) and Chlorine (red) during an ALE cycle. Panel (a) shows the state immediately after chlorination but before ion bombardment, where a significant chlorine-rich layer exists at the surface (high Z).\"},{\"panel_id\":\"b\",\"text\":\"The label shows the state after 150 Argon ion impacts, revealing how the ion bombardment has begun to remove the surface material and push some chlorine deeper into the silicon lattice.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.060 | 0.000 |\\n| 5  | 0.047 | 0.000 |\\n| 10 | 0.048 | 0.000 |\\n| 15 | 0.048 | 0.000 |\\n| 20 | 0.049 | 0.000 |\\n| 25 | 0.050 | 0.000 |\\n| 30 | 0.053 | 0.000 |\\n| 35 | 0.045 | 0.005 |\\n| 40 | 0.051 | 0.002 |\\n| 45 | 0.048 | 0.002 |\\n| 50 | 0.038 | 0.009 |\\n| 55 | 0.005 | 0.012 |\\n| 60 | 0.000 | 0.000 |\"},{\"panel_id\":\"b\",\"text\":\"| Z Position (Å) | Si Density (atom/ų) | Cl Density (atom/ų) |\\n|----------------|----------------------|----------------------|\\n| 0  | 0.070 | 0.000 |\\n| 5  | 0.046 | 0.000 |\\n| 10 | 0.050 | 0.000 |\\n| 15 | 0.049 | 0.000 |\\n| 20 | 0.051 | 0.000 |\\n| 25 | 0.052 | 0.000 |\\n| 30 | 0.050 | 0.001 |\\n| 35 | 0.048 | 0.003 |\\n| 40 | 0.051 | 0.003 |\\n| 45 | 0.044 | 0.005 |\\n| 50 | 0.038 | 0.008 |\\n| 55 | 0.008 | 0.010 |\\n| 60 | 0.000 | 0.000 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Residue vs. Damage. Stopping at 150 impacts leaves a high concentration of residual chlorine (red line is still prominent), which could contaminate the next layer. However, continuing to 1000 impacts ensures a cleaner surface but risks deeper ion damage to the underlying silicon crystal structure.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This is caused by Knock-on Implantation or Mixing. The energetic Ar+ ions transfer momentum to the surface chlorine atoms, driving some of them deeper into the silicon lattice rather than just sputtering them off. This creates a mixed, amorphous Si-Cl sub-surface layer.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Increased Roughness. The gradual density transition indicates that the surface is not atomically flat but rather rough or intermixed, which can degrade carrier mobility in the final semiconductor device.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Physical Sputtering / Ion Bombardment. The 150 Ar^+ ion impacts have physically sputtered away the top-most layer of chlorinated silicon, reducing the total amount of chlorine and receding the surface (Z position decreases).\"}]}]","bbox":[{"panel_id":"b","x":686,"y":21,"width":652,"height":540},{"panel_id":"a","x":3,"y":26,"width":675,"height":541}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/Molecular dynamics study of silicon atomic layer etching by chlorine gas and argon ions.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"15","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1342,"height":570,"image_format":"jpeg","image_sha256":"cd3d8f7dd46179c4a88b064813ed8928fdf19a17f7d27a51a5b6cd6f1ffc331b","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_16_fig_6.jpg","caption":"Fig. 6. Line plots illustrating the effects of the gap distance on precursor separation, of which operating condition is (a) Case 1, (b) Case 2, (c) Case 3, and (d) Case 4 as described in Table 3. The shaded area indicates the precursor intermixing zone.","id":"train/atomic-layer-etching/simulation-usecase/16/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/16/fig_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the mole fraction of HF and TMA at different positions along the x-axis.\"},{\"panel_id\":\"b\",\"text\":\"The line chart displays the mole fraction of HF and TMA with sharp transitions at specific positions related to the tool geometry.\"},{\"panel_id\":\"c\",\"text\":\"The line chart shows the mole fraction of HF and TMA with sharp transitions at specific positions related to the tool geometry.\"},{\"panel_id\":\"d\",\"text\":\"The line chart illustrates the mole fraction of HF and TMA with sharp transitions at specific positions related to the tool geometry.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.2 |0.12|\\n|-50 | 0.21 |0.13|\\n|0 | 0.15 |0.18|\\n|50 | 0.13 |0.26|\\n|100 | 0.11 |0.25|\"},{\"panel_id\":\"b\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0 |0|\\n|-50 | 0.67 |0|\\n|0 | 0 |0|\\n|50 | 0 |0.55|\\n|100 | 0 |0|\"},{\"panel_id\":\"c\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0 |0|\\n|-50 | 0.67 |0|\\n|0 | 0 |0|\\n|50 | 0 |0.55|\\n|100 | 0 |0|\"},{\"panel_id\":\"d\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0 |0|\\n|-50 | 0.67 |0|\\n|0 | 0 |0|\\n|50 | 0 |0.55|\\n|100 | 0 |0|\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The width of the slits is equal to the width of the plateau, which is about 30 mm.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Part of the gases penetrates into the gas shield regions, which is why there is no precise cut-off.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No there will not be ALE, as TMA and HF already mix in the gas phase. This can result in continuous etching.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Low gap sizes usually mean that lower flows of precursor are required so there will be less gas consumption.\"}]}]","bbox":[{"panel_id":"c","x":3,"y":507,"width":572,"height":479},{"panel_id":"b","x":607,"y":4,"width":565,"height":483},{"panel_id":"a","x":6,"y":2,"width":562,"height":469},{"panel_id":"d","x":596,"y":507,"width":574,"height":477}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/Multiscale computational fluid dynamics modeling.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1175,"height":989,"image_format":"jpeg","image_sha256":"eaaf880d1caa39e77c462eddae077ad8ca95c74fbd6a5447e774347a1eee04fa","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_16_fig_7.jpg","caption":"Fig. 7. Line plots illustrating the effects of the  $\\mathbb{N}_2$  flow rate on precursor separation, of which operating condition is (a) Case 1, (b) Case 5, (c) Case 6, and (d) Case 7 as described in Table 3. The shaded area indicates the precursor intermixing zone.","id":"train/atomic-layer-etching/simulation-usecase/16/fig_7","sample_id":"atomic-layer-etching/simulation-usecase/16/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the mole fraction of HF and TMA at different positions along the x-axis. Where the blue area depicts gas mixing.\"},{\"panel_id\":\"b\",\"text\":\"The line chart shows the mole fraction of HF and TMA at different positions along the x-axis. Where the blue area depicts gas mixing.\"},{\"panel_id\":\"c\",\"text\":\"The line chart shows the mole fraction of HF and TMA at different positions along the x-axis. Where the blue area depicts gas mixing.\"},{\"panel_id\":\"d\",\"text\":\"The line chart shows the mole fraction of HF and TMA at different positions along the x-axis. Where the blue area depicts gas mixing.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.2 |0.12|\\n|-50 | 0.21 |0.13|\\n|0 | 0.15 |0.18|\\n|50 | 0.13 |0.26|\\n|100 | 0.11 |0.25|\"},{\"panel_id\":\"b\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.09 |0.03|\\n|-50 | 0.11 |0.04|\\n|0 | 0.05 |0.06|\\n|50 | 0.03 |0.12|\\n|100 | 0.03 |0.11|\"},{\"panel_id\":\"c\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.05 |0.01|\\n|-50 | 0.08 |0.01|\\n|0 | 0.03|0.03|\\n|50 | 0.02 |0.08|\\n|100 | 0.01 |0.06|\"},{\"panel_id\":\"d\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.02 |0.00|\\n|-50 | 0.05 |0.00|\\n|0 | 0.02|0.02|\\n|50 | 0.01 |0.05|\\n|100 | 0.00 |0.02|\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes this could be a problem, in that case the flows of reactants should also increase.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The TMA slit is located on the right side as there the TMA signal is the highest.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The gas mixing will result in continuous etching, which is detrimental to ALE.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes there is.\"}]}]","bbox":[{"panel_id":"c","x":4,"y":511,"width":568,"height":479},{"panel_id":"b","x":608,"y":0,"width":570,"height":487},{"panel_id":"a","x":8,"y":1,"width":564,"height":476},{"panel_id":"d","x":604,"y":511,"width":573,"height":477}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/Multiscale computational fluid dynamics modeling.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1183,"height":992,"image_format":"jpeg","image_sha256":"5f2d58bd9ff82735d591ed68d62df8c6bfe038b7ff2030104038dab63d8fa42c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_16_fig_8.jpg","caption":"Fig. 8. Line plots illustrating the effects of the precursor flow rate on precursor separation, of which operating condition is (a) Case 1, (b) Case 8, (c) Case 9, and (d) Case 10 as described in Table 3. The shaded area indicates the precursor intermixing zone.","id":"train/atomic-layer-etching/simulation-usecase/16/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/16/fig_8","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the mole fraction of HF and TMA at various positions along the x-axis, where the mixing of HF and TMA is shown blue.\"},{\"panel_id\":\"b\",\"text\":\"The line chart displays the mole fraction of HF and TMA at different positions, where the mixing of HF and TMA is shown blue.\"},{\"panel_id\":\"c\",\"text\":\"The line chart illustrates the mole fraction of HF and TMA at varying positions, where the mixing of HF and TMA is shown blue.\"},{\"panel_id\":\"d\",\"text\":\"The line chart depicts the mole fraction of HF and TMA at different positions, where the mixing of HF and TMA is shown blue.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.2 |0.12|\\n|-50 | 0.21 |0.13|\\n|0 | 0.15 |0.18|\\n|50 | 0.13 |0.26|\\n|100 | 0.11 |0.25|\"},{\"panel_id\":\"b\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.27 |0.14|\\n|-50 | 0.29 |0.16|\\n|0 | 0.19 |0.22|\\n|50 | 0.12 |0.31|\\n|100 | 0.11 |0.30|\"},{\"panel_id\":\"c\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.33 |0.13|\\n|-50 | 0.35 |0.14|\\n|0 | 0.2 |0.22|\\n|50 | 0.11 |0.4|\\n|100 | 0.11 |0.38|\"},{\"panel_id\":\"d\",\"text\":\"| Position (mm) | Mole Fraction HF |Mole Fraction TMA |\\n|---|---|---|\\n|-100 | 0.4 |0.11|\\n|-50 | 0.42 |0.12|\\n|0 | 0.25 |0.26|\\n|50 | 0.08 |0.43|\\n|100 | 0.08 |0.43|\"}]","vqa":"[{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The flow of nitrogen in the gas shield can be greatly increased.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It is likely that there will be continuous etching, even though the mole fractions of TMA and HF in their designated locations are way larger.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TMA has a higher mole fraction in the HF region as compared to HF in the TMA region. So TMA diffuses more into the other region.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.21\"}]}]","bbox":[{"panel_id":"c","x":5,"y":516,"width":558,"height":472},{"panel_id":"b","x":604,"y":1,"width":577,"height":489},{"panel_id":"a","x":5,"y":1,"width":572,"height":475},{"panel_id":"d","x":599,"y":518,"width":577,"height":470}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/Multiscale computational fluid dynamics modeling.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"16","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1183,"height":992,"image_format":"jpeg","image_sha256":"3316f48079e0f1a72eef219e27cc42f153a17413c0527c9ba4491ae86cffd577","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_10.jpg","caption":"FIG. 10. Fluxes to the wafer as a function of radius for the  $\\mathrm{Ar / c - C_4F_8}$  CCP. (a) Neutral and (b) ion fluxes.  $\\mathrm{CF}_3$ ,  $\\mathrm{C}_3\\mathrm{F}_5$ , and  $\\mathrm{F}$  are the dominant radicals.  $\\mathrm{Ar^{+}}$ ,  $\\mathrm{C}_2\\mathrm{F}_4^+$ , and  $\\mathrm{CF}_x^+$ $(x = 1 - 3)$  are the dominant ions.","id":"train/atomic-layer-etching/simulation-usecase/19/fig_10","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_10","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the neutral flux of various species as a function of radius of the wafer.\"},{\"panel_id\":\"b\",\"text\":\"The figure displays the ion flux of ions as a function of radius of the wafer.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Species | Radius (cm) | Neutral Flux ($10^{17}$ cm$^{-2}$s$^{-1}$) | \\n|---|---|---|\\n|CF<sub>3</sub> | 0 | 5.82 | \\n|CF<sub>3</sub> | 4 | 5.82 | \\n|CF<sub>3</sub> | 10 | 5.88| \\n|CF<sub>2</sub> | 0 | 4.57 | \\n|CF<sub>2</sub> | 4 | 4.59 | \\n|CF<sub>2</sub> | 10 | 4.61| \\n|CF | 0 | 4.3 |\\n|CF | 4 | 4.26 |\\n|CF | 10 | 4.28 |\\n|C<sub>2</sub>F<sub>6</sub> | 0 | 2.66 |\\n|C<sub>2</sub>F<sub>6</sub> | 4 | 2.68 |\\n|C<sub>2</sub>F<sub>6</sub> | 10 | 2.74 |\\n|C<sub>3</sub>F<sub>5</sub> | 0 | 1.77 |\\n|C<sub>3</sub>F<sub>5</sub> | 4 | 1.75 |\\n|C<sub>3</sub>F<sub>5</sub> | 10 |1.77 |\\n|C<sub>4</sub>F<sub>7</sub> | 0 | 1.46 |\\n|C<sub>4</sub>F<sub>7</sub> | 4 | 1.46|\\n|C<sub>4</sub>F<sub>7</sub> | 10 | 1.46 |\\n|C<sub>2</sub>F<sub>4</sub> |  0 | 0.91 |\\n|C<sub>2</sub>F<sub>4</sub> |  4 | 0.89 |\\n|C<sub>2</sub>F<sub>4</sub> |  10 | 0.91 |\\n|CF<sub>4</sub> | 0 | 0.78 |\\n|CF<sub>4</sub> | 4 | 0.68|\\n|CF<sub>4</sub> | 10 | 0.48 |\\n|F | 0 |0.68 |\\n|F | 4 | 0.68 |\\n|F | 10 | 0.72 |\"},{\"panel_id\":\"b\",\"text\":\"| Species | Radius (cm) | Ion Flux ($10^{15}$ cm$^{-2}$s$^{-1}$) |\\n|---|---|---|\\n| Ar<sup>+</sup> | 0 | 12.9 | \\n| Ar<sup>+</sup> | 4 | 16.7 | \\n| Ar<sup>+</sup> | 10 | 18.2 | \\n|C<sub>2</sub>F<sub>4</sub><sup>+</sup> | 0 | 8.5 | \\n|C<sub>2</sub>F<sub>4</sub><sup>+</sup> | 4 | 10.1 | \\n|C<sub>2</sub>F<sub>4</sub><sup>+</sup> | 10 | 9.7 | \\n|C<sub>3</sub>F<sub>5</sub><sup>+</sup> | 0 | 4.9 | \\n|C<sub>3</sub>F<sub>5</sub><sup>+</sup> | 4 | 6.7 | \\n|C<sub>3</sub>F<sub>5</sub><sup>+</sup> | 10 | 6.6 | \\n|CF<sup>+</sup> + CF<sub>2</sub><sup>+</sup> + CF<sub>3</sub><sup>+</sup> | 0 | 3.8 |\\n|CF<sup>+</sup> + CF<sub>2</sub><sup>+</sup> + CF<sub>3</sub><sup>+</sup> | 4 | 5.6 |\\n|CF<sup>+</sup> + CF<sub>2</sub><sup>+</sup> + CF<sub>3</sub><sup>+</sup> | 10 | 5.4 |\\n|Ar<sup>+</sup> (Step 2) | 0 | 2.3 | \\n|Ar<sup>+</sup> (Step 2) | 4 | 2.2 | \\n|Ar<sup>+</sup> (Step 2) | 10 | 2.2 | \\n|F <sup>+</sup>  | 0 | 0.5 |\\n|F <sup>+</sup>  | 4 | 0.45 |\\n|F <sup>+</sup> | 10 | 0.39 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The argon ion is the primary species responsible for the non-uniformity observed near the edge of the wafer. The fluxes of other ions remain relatively flat across the radius. The argon ion flux exhibits a distinct rise from the center to the edge of the wafer.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"CF3.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At ~ 5 cm.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"CF4, C2F6 and C2F4.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":376,"height":376},{"panel_id":"b","x":1,"y":383,"width":377,"height":372}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":378,"height":756,"image_format":"jpeg","image_sha256":"d8be253135218ca1582aa9a36813b4ea8219a067fa7c597b496edbc24f813d07","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_16.jpg","caption":"FiG. 16. Fluxes to the wafer as a function of radius for the nonsinusoidal bias PALE process. (a) Neutral and (b) ion fluxes.  $\\mathrm{CF}_2$  CF,and  $\\mathrm{CF}_3$  are the dominant fluorine containing neutral radicals.  $\\mathrm{Ar^{+}}$ $\\mathrm{CF}_3^+$ $\\mathrm{CF}_2^+$  and  $\\mathrm{CF^{+}}$  are the dominant ion fluxes.","id":"train/atomic-layer-etching/simulation-usecase/19/fig_16","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_16","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the neutral flux of various species as a function of wafer radius (0 - 6 cm).The flux is dominated by CF3 radical.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the ion flux of various species as a function of of wafer radius (0 - 6 cm). The flux is dominated by Ar(+).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Radius (cm) | CF2 flux (10¹⁷ cm⁻² s⁻¹) | CF flux (10¹⁷ cm⁻² s⁻¹) | C flux (10¹⁷ cm⁻² s⁻¹) | CF3 (×2) flux (10¹⁷ cm⁻² s⁻¹) | C2F3 (×2) flux (10¹⁷ cm⁻² s⁻¹) | F (×2) flux (10¹⁷ cm⁻² s⁻¹) | C2F4 (×2) flux (10¹⁷ cm⁻² s⁻¹) | C2F5 (×2) flux (10¹⁷ cm⁻² s⁻¹) |\\n|---|---|---|---|---|---|---|---|---|\\n| 0 | 5.75 | 3.75 | 2.82 | 1.65 | 1.25 | 0.9 | 0.45 | 0.3 |\\n| 3 | 5.64 | 3.79 | 2.68 | 1.61 | 1.34 | 0.82 | 0.51 | 0.32 |\\n| 6| 5.79 | 4.12 | 2.53 | 1.54 | 1.75 | 0.63 | 0.63 | 0.33 |\"},{\"panel_id\":\"b\",\"text\":\"| Radius (cm) | Ar+ flux (10¹⁵ cm⁻² s⁻¹) | CF3+ flux (10¹⁵ cm⁻² s⁻¹)  | CF2+ flux (10¹⁵ cm⁻² s⁻¹)  | CF+ flux (10¹⁵ cm⁻² s⁻¹)  | C2F4+ flux (10¹⁵ cm⁻² s⁻¹)  | C2F5+ + C3F5+ flux (10¹⁵ cm⁻² s⁻¹)  | F+ flux (10¹⁵ cm⁻² s⁻¹) |\\n|---|---|---|---|---|---|---|---|\\n| 0 | 7 | 3.5 | 2.6 | 2.4 | 0.52 | 0.52 | 0.12 |\\n| 3 | 6 | 2.9 | 2.3 | 2.14 | 0.52 | 0.52 | 0.07 |\\n| 6 | 4.6 | 2.6 | 1.75 | 1.75 | 0.52 | 0.52 | 0.07 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ar(+).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"C2F5(+) + C3F5(+), C2F4(+), CF(+), CF2(+), CF3(+), Ar(+)\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At the center of the wafer (r = 0 cm).Since physical sputtering is driven by the bombardment of ions, the region receiving the highest flux of heavy Argon ions will experience the most intense sputtering. The Ar(+) flux shows its peak at the center of the wafer and decreases towards the edge.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"CF2, CF, C, C2F3, CF3, F, C2F4, C2F5\"}]}]","bbox":[{"panel_id":"a","x":5,"y":0,"width":371,"height":377},{"panel_id":"b","x":0,"y":392,"width":377,"height":365}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_16.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_16.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":378,"height":759,"image_format":"jpeg","image_sha256":"0e042de76bb3458cb08db49cafae86276ebac58708df0b8d16fcf0b308c644e6","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_19_fig_3.jpg","caption":"FIG. 3. (Color online) Ion properties for the PALE cycle. (a) Ion fluxes to the wafer as a function of radius.  $\\mathrm{Cl}^+$  is the major ion during step 1 (passivation) due to dissociation of the  $\\mathrm{Cl}_2$  feedstock. The  $\\mathrm{Ar^{+}}$  flux is higher during step 2 (etching) step due to lack of any competing processes. (b) Total IEDs averaged over the wafer. Low ion energies are necessary to minimize etching during step 1 while higher ion energies in step 2 enable etching of the  $\\mathrm{SiCl}_x$  passivation. (IEADs are plotted using a log scale over 2 decades.)","id":"train/atomic-layer-etching/simulation-usecase/19/fig_3","sample_id":"atomic-layer-etching/simulation-usecase/19/fig_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"contour heatmap"},{"panel_id":"c","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The  chart shows the ion flux of different ions (Ar<sup>+</sup>, Cl<sup>+</sup>, and Cl<sub>2</sub><sup>+</sup>) for step 1 (passivation) and step 2(etching) at various radii from the center of the wafer.\"},{\"panel_id\":\"b\",\"text\":\"The heatmap displays the ion energy and angular distribution (IEAD) for step 1 (passivation).\"},{\"panel_id\":\"c\",\"text\":\"The heatmap displays the ion energy and angular distribution (IEAD) for step 2 (etching).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Radius (cm) | Ion Flux (10<sup>16</sup> cm<sup>-2</sup>s<sup>-1</sup>) | Species|\\n|---|---|---|\\n| 0 | 7.93 | Ar (+) Step 2|\\n| 3 | 7.27 | Ar (+) Step 2|\\n| 6 | 5.18 | Ar (+) Step 2|\\n| 0 | 1.35 | Ar (+) Step 1|\\n| 3 | 1.2 | Ar (+) Step 1|\\n| 6 | 0.7 | Ar (+) Step 1|\\n| 0 | 3.6 | Cl (+) Step 1|\\n| 3 | 3.1 | Cl (+) Step 1|\\n| 6 | 1.89 | Cl (+) Step 1|\\n| 0 | 0.5 | Cl2(+) Step 1|\\n| 3 | 0.5 | Cl2(+) Step 1|\\n| 6 | 0.5 | Cl2(+) Step 1|\"},{\"panel_id\":\"b\",\"text\":\"| Angle (deg) | Energy (eV) | Intensity |\\n|---|---|---|\\n| -15 | 14 | 0.01 |\\n| 0 | 14 | 0.01 |\\n| 15 | 14 | 0.01 |\\n| -15 | 16 | 0.69|\\n| 0 | 16 | 0.69 |\\n| 15 | 16 | 0.36 |\\n| -15 | 17 | 0.69 |\\n| 0 | 17 | 1 |\\n| 15 | 17 | 0.36 |\\n| -15 | 21 | 0.01 |\\n| 0 | 21 | 0.01 |\\n| 15 | 21 | 0.01 |\"},{\"panel_id\":\"c\",\"text\":\"| Angle (deg) | Energy (eV) | Intensity |\\n|---|---|---|\\n| 0 | 42 | 0.01 |\\n| 0 | 51 | 0.85 |\\n| -9 | 51 | 0.01 |\\n| 9 | 51 | 0.01|\\n| 0 | 57 | 1 |\\n| 9 | 57 | 0.01 |\\n| -9 | 57 | 0.01 |\\n| 0 | 61 | 0.01 |\\n| 9 | 17 | 0.01 |\\n| -9 | 21 | 0.01 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Cl (+).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Cl2 (+).\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Perpendicular to the wafer surface (0 degrees).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ ± 10 degrees.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":368,"height":364},{"panel_id":"b","x":8,"y":372,"width":201,"height":442},{"panel_id":"c","x":202,"y":370,"width":170,"height":437}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/19/Plasma atomic layer etching using conventional plasma equipment.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":375,"height":808,"image_format":"jpeg","image_sha256":"60b54ffaf3f0daa8efa070f8c2f4d6ec47a3d6ab6f9bb6d9ff4d9166fa13c555","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_20_figure_3.jpg","caption":"Figure 3. Various contributions to the reaction FEP for the CVE1 reaction of  $\\mathrm{TiO}_2 + \\mathrm{HF}\\rightarrow \\mathrm{TiF}_4 + \\mathrm{H}_2\\mathrm{O}$  at an HF reactant pressure of 0.2 Torr.","id":"train/atomic-layer-etching/simulation-usecase/20/figure_3","sample_id":"atomic-layer-etching/simulation-usecase/20/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure plots the thermodynamic contributions to the reaction free energy profile (FEP) for the etching reaction TiO₂(bulk) + 4HF(g) → TiF₄(g) + 2H₂O(g) as a function of temperature (0–800 K) at a constant HF pressure of 0.2 Torr. It shows how the electronic energy (ΔE), zero-point correction (ΔE+ΔZPE), Gibbs free energy (ΔG), and entropic term (TΔS) evolve with temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature [K] | E [eV / TiO<sub>2</sub>] |  |\\n|---|---|----|\\n| 0    | -1.4 | ΔE |\\n| 0    | -1.2 | ΔE + ΔZPE |\\n| 0    | -1.2 | ΔG |\\n| 0    | 0.0  | TΔS |\\n| 0    | 0.0  | ΔW |\\n| 0    | 0.0  | RT ln(Q) |\\n| 200  | -1.4 | ΔE |\\n| 200  | -1.2 | ΔE + ΔZPE |\\n| 200  | -1.2 | ΔG |\\n| 200  | -0.1 | TΔS |\\n| 200  | -0.03  | ΔW |\\n| 200  | -0.01  | RT ln(Q) |\\n| 400  | -1.4 | ΔE |\\n| 400  | -1.2 | ΔE + ΔZPE |\\n| 400  | -1.1 | ΔG |\\n| 400  | -0.2 | TΔS |\\n| 400  | -0.1 | ΔW |\\n| 400  | -0.02  | RT ln(Q) |\\n| 600  | -1.4 | ΔE |\\n| 600  | -1.2 | ΔE + ΔZPE |\\n| 600  | -1.0 | ΔG |\\n| 600  | -0.4 | TΔS |\\n| 600  | -0.1 | ΔW |\\n| 600  | -0.04  | RT ln(Q) |\\n| 800  | -1.4 | ΔE |\\n| 800  | -1.2 | ΔE + ΔZPE |\\n| 800  | -0.8 | ΔG |\\n| 800  | -0.6 | TΔS |\\n| 800  | -0.1 | ΔW |\\n| 800  | -0.1 | RT ln(Q) |\\n| 1000 | -1.4 | ΔE |\\n| 1000 | -1.2 | ΔE + ΔZPE |\\n| 1000 | -0.7 | ΔG |\\n| 1000 | -0.7 | TΔS |\\n| 1000 | -0.2 | ΔW |\\n| 1000 | -0.1 | RT ln(Q) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes, it is spontaneous. At 300 K, the Gibbs free energy (ΔG) is approximately −1.2 eV, and a negative ΔG indicates a spontaneous reaction.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"700 K is better for ensuring spontaneity (more negative ΔG). The trade-off is the higher thermal budget, which may damage temperature-sensitive materials or increase energy costs, whereas 400 K offers lower thermal stress but may result in a non-spontaneous or slower etch rate.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ΔG becomes more negative (more spontaneous) as temperature increases. This is primarily driven by the large, positive TΔS term (entropy gain) due to the production of gas-phase molecules (TiF₄ and H₂O) from a solid and gas, which outweighs the endothermic enthalpy contribution at higher temperatures.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies that HF gas etching of TiO₂ is not feasible at room temperature (ΔG > 0). To enable etching, the process must be heated above ~500 K, or an alternative, more reactive chemistry (e.g., plasma activation) must be used to lower the kinetic barrier and drive the reaction at lower temperatures.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":539,"height":498}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/Prediction and Validation of the Process Window for Atomic Layer Etching HF Exposure on TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":544,"height":502,"image_format":"jpeg","image_sha256":"66f7456d5ca0d56dc4ddcfd2ac98b1ca36080a79c49c87506e0b7b2843dd73a4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_20_figure_4.jpg","caption":"Figure 4. FEPs for the CVE1, CVE2, CVE3, SL1 SL2, and SL3 reactions for HF interacting with  $\\mathrm{TiO_2}$  versus temperature. The labels X and Y correspond to the values listed in Table 1.","id":"train/atomic-layer-etching/simulation-usecase/20/figure_4","sample_id":"atomic-layer-etching/simulation-usecase/20/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This multiple line chart image compares the free energy profiles (FEPs) of three Continuous Vapor Etching (CVE) reactions (red, red-dashed, blue) with their corresponding self-limiting (SL) surface modification reactions (green, green-dashed, yellow). For ideal Atomic Layer Etching (ALE), the SL reaction must be spontaneous (ΔG < 0) while the CVE reaction remains non-spontaneous (ΔG > 0), defining a “Goldilocks” window where the surface is modified but bulk etching is suppressed.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature [K] | ΔG [ev / TiO<sub>2</sub>] | Materials |\\n|--------|---------|---------|\\n| 0    | -1.2 | CVE1(Y1) |\\n| 0    | 2.1  | CVE2(Y2) |\\n| 0    | 2.2  | CVE3(Y3) |\\n| 0    | -2.2 | SL1(X1) |\\n| 0    | -1.1 | SL2(X2) |\\n| 0    | -0.1 | SL3(X3) |\\n| 200  | -1.2 | CVE1(Y1) |\\n| 200  | 1.7  | CVE2(Y2) |\\n| 200  | 1.9  | CVE3(Y3) |\\n| 200  | -1.6 | SL1(X1) |\\n| 200  | -0.8 | SL2(X2) |\\n| 200  | 0.2  | SL3(X3) |\\n| 400  | -1.1 | CVE1(Y1) |\\n| 400  | 1.4  | CVE2(Y2) |\\n| 400  | 1.7  | CVE3(Y3) |\\n| 400  | -0.9 | SL1(X1) |\\n| 400  | -0.5 | SL2(X2) |\\n| 400  | 0.6  | SL3(X3) |\\n| 600  | -1.0 | CVE1(Y1) |\\n| 600  | 1.1  | CVE2(Y2) |\\n| 600  | 1.4  | CVE3(Y3) |\\n| 600  | -0.3 | SL1(X1) |\\n| 600  | -0.2 | SL2(X2) |\\n| 600  | 1.0  | SL3(X3) |\\n| 800  | -0.8 | CVE1(Y1) |\\n| 800  | 0.9  | CVE2(Y2) |\\n| 800  | 1.2  | CVE3(Y3) |\\n| 800  | 0.3  | SL1(X1) |\\n| 800  | 0.1  | SL2(X2) |\\n| 800  | 1.3  | SL3(X3) |\\n| 1000 | -0.7 | CVE1(Y1) |\\n| 1000 | 0.6  | CVE2(Y2) |\\n| 1000 | 1.0  | CVE3(Y3) |\\n| 1000 | 0.9  | SL1(X1) |\\n| 1000 | 0.4  | SL2(X2) |\\n| 1000 | 1.6  | SL3(X3) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"CVE1: 4 HF molecules.\\n\\nCVE2: 2 HF molecules.\\n\\nCVE3: 3 HF molecules.\\n\\nConclusion: CVE1 consumes the most reactant (4 HF).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The CVE2 pathway is not viable for a spontaneous, continuous etch. Its ΔG remains positive (non-spontaneous) across the entire temperature range shown, unlike CVE1 which becomes favorable above ~700 K. To achieve continuous etching, one must use conditions that favor CVE1, accepting a higher thermal budget.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A CVE pathway produces volatile gas-phase products (e.g., TiF₄(g)), leading to complete material removal. An SL pathway produces a modified, non-volatile surface layer (e.g., TiF₄(surf)). The SL reactions are always more exothermic (more negative ΔE) than their CVE counterparts because forming a surface layer is energetically cheaper than completely breaking bonds to release gas molecules.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies that isotropic, continuous etching of TiO₂ with HF gas requires a high-temperature process (>700K). At lower temperatures, the reaction will self-limit, only modifying the surface. Therefore, for patterning, one must either operate at high temperature or use an alternative activation method (like plasma) to drive the CVE pathway at lower temperatures.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":556,"height":501}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/Prediction and Validation of the Process Window for Atomic Layer Etching HF Exposure on TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":563,"height":505,"image_format":"jpeg","image_sha256":"bcfcfc8bca6256ccdac3e9f91a0ae8069bab603525b3643e46503284c1dfdec4","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_20_figure_5.jpg","caption":"Figure 5. FEPs of CVE1 and SL1 showing three regions: preferred self-limiting, preferred etching, and purely etching. SL1 and CVE1 line crossing occurs at  $360~\\mathrm{K}$ $(87^{\\circ}\\mathrm{C})$","id":"train/atomic-layer-etching/simulation-usecase/20/figure_5","sample_id":"atomic-layer-etching/simulation-usecase/20/figure_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure plots the Free Energy Profiles (FEPs) for the CVE1 (continuous vapor etch) and SL1 (self-limiting surface reaction) pathways for HF interacting with TiO₂. The lines cross at 360 K, defining three regimes: below 360 K, SL1 is favored (self-limiting); between 360 K and ~700 K, CVE1 becomes progressively more favored (preferred etching); above 700 K, CVE1 is strongly favored (purely etching).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature [K] | ΔG [eV / TiO<sub>2</sub>] | Label    | Region                |\\n|-----------------|---------------------------|----------|-----------------------|\\n| 0               | -2.2                      | SL1(X1)  | Preferred self-limiting |\\n| 0               | -1.1                      | CVE1(Y2) | Preferred self-limiting |\\n| 200             | -1.5                      | SL1(X1)  | Preferred self-limiting |\\n| 200             | -1.1                      | CVE1(Y2) | Preferred self-limiting |\\n| 400             | -0.9                      | SL1(X1)  | Preferred etching       |\\n| 400             | -1.0                      | CVE1(Y2) | Preferred etching       |\\n| 600             | -0.3                      | SL1(X1)  | Preferred etching       |\\n| 600             | -0.9                      | CVE1(Y2) | Preferred etching       |\\n| 800             | 0.2                       | SL1(X1)  | Purely etching          |\\n| 800             | -0.8                      | CVE1(Y2) | Purely etching          |\\n| 1000            | 0.8                       | SL1(X1)  | Purely etching          |\\n| 1000            | -0.7                      | CVE1(Y2) | Purely etching          |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Max Temperature: 360 K (87°C).Above 360 K, the preference switches; the bulk etching reaction (CVE1) becomes more energetically favorable than the surface modification (SL1).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"300 K (Self-limiting regime): Pro: Excellent depth control, atomic-layer precision. Con: Very slow or negligible material removal rate.\\n800 K (Purely etching regime): Pro: High etch rate, fast throughput. Con: Poor depth control, isotropic etching.\\nTrade-off: The choice is between controllability (low temp) and throughput (high temp). For precise patterning, a low-temperature, self-limiting process is necessary, sacrificing speed.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The shift is driven by entropy. The CVE1 pathway produces more gas molecules, granting it a larger, favorable entropy term (+TΔS). As temperature rises, this term dominates the free energy (ΔG = ΔH - TΔS), making the complete volatilization reaction (CVE1) increasingly favorable over the surface passivation reaction (SL1).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, it should be operated below 360 K. The SL1 pathway is thermodynamically favored over CVE1. This makes the reaction inherently self-limiting, as it preferentially forms a stable surface layer rather than volatile products, ensuring precise, layer-by-layer removal essential for controlled etching.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":555,"height":488}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/20/Prediction and Validation of the Process Window for Atomic Layer Etching HF Exposure on TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"20","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":561,"height":495,"image_format":"jpeg","image_sha256":"a142268e0d7643f523a6e21bc1104902ae357e96950f12123db3467997d4af4f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_26_fig_5.jpg","caption":"FIG. 5. EPCs of Si PA-ALE processes obtained from MD simulations as functions of  $\\mathrm{Ar^{+}}$  ion incident energy. The solid and dashed lines represent those for the thin and thick Cl layer deposition cases. Other ALE conditions are listed in Table I. The value of EPC given here is the average of etched depths per cycle from C3 to C6.","id":"train/atomic-layer-etching/simulation-usecase/26/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/26/fig_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the relationship between Ar+ ion incident energy (in eV) and etch per cycleEPC (in Å) for thin and thick Cl layer deposition methods.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Energy (eV) | Thin Cl layer deposition (Å) | Thick Cl layer deposition (Å) |\\n|-------------|-------------------------------|--------------------------------|\\n| 0           | 0                             | 0                              |\\n| 10          | 1                             | 2                              |\\n| 20          | 2                             | 4                              |\\n| 30          | 4                             | 6                              |\\n| 40          | 7                             | 10                             |\\n| 50          | 11                            | 14                             |\\n| 60          | 16                            | 19                             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The value of EPC given here is the average of etched depths per cycle from C3 to C6.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Because from the image, it can be seen that the EPC is higher in the thick Cl layer deposition case than in the thin Cl layer deposition case\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 20 eV Ar+ ion energy, the EPC correspond to Si monolayer as follows\\n1. The simulated EPCs  for thin Cl layer deposition is  1.6  Å/cycle = 1.2 Si Monolayer/cycle\\n2. The simulated EPC for thick Cl layer deposition is 4.2 Å/cycle = 3 Si Monolayer/cycle.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":639,"height":439}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/Surface damage formation during atomic layer etching of silicon with chlorine adsorption.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":639,"height":439,"image_format":"jpeg","image_sha256":"7ee32e3c7d7a9deb9201add27ae08e0076fdb0ff6a158685cd4319ad1e00acb7","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_26_fig_7.jpg","caption":"FIG. 7. Thicknesses of damaged-layers formed during the Si PA-ALE processes, obtained from MD simulations, as functions of the  $\\mathrm{Ar^{+}}$  ion incident energy. The solid and dashed lines represent those for thin and thick Cl layer deposition cases. Other ALE conditions are listed in Table I. The value of the thickness here is the average of those over all six cycles, i.e., from C1 to C6.","id":"train/atomic-layer-etching/simulation-usecase/26/fig_7","sample_id":"atomic-layer-etching/simulation-usecase/26/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between energy (in eV) and thickness (in nm) for thin and thick Cl layer depositions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Energy (eV) | Thin Cl layer deposition (nm) | Thick Cl layer deposition (nm) |\\n|-------------|--------------------------------|---------------------------------|\\n| 0           | 0.0                            | 0.0                             |\\n| 10          | 0.8                            | 0.9                             |\\n| 20          | 1.5                            | 1.7                             |\\n| 30          | 1.8                            | 2.0                             |\\n| 40          | 2.0                            | 2.0                             |\\n| 50          | 1.8                            | 2.1                             |\\n| 60          | 1.8                            | 2.3                             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The damaged-layer thickness observed here is significantly higher than the corresponding EPC at 20 eV ion energy and comparable at 60 eV\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It can be concluded that, for both thin and thick Cl layer deposition cases, the damaged-layer thicknesses weakly and more-or-less linearly increase with the increasing incident ion energy in the energy range studied here. With an increasing ion incident energy, the penetration depths of Arþ ions and recoiled Cl atoms increase but so do the etch rates. In other words, a large part of the damaged-layer formed during the desorption step is etched away at high ion incident energy. It is\\nalso seen that the damaged-layer is slightly thicker in the thick Cl layer deposition case.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The damaged-layer thickness was averaged over the first six cycles (i.e., from C1 to C6)\\nunder each ALE condition\"}]}]","bbox":[{"panel_id":"a","x":8,"y":8,"width":631,"height":458}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/Surface damage formation during atomic layer etching of silicon with chlorine adsorption.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":639,"height":466,"image_format":"jpeg","image_sha256":"d550cae19c19658303033ad1468fcbda333a3d9a9f8b4776d3189004dd178a4b","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_26_fig_9.jpg","caption":"FIG. 9. Numbers of desorbed species as functions of the  $\\mathsf{Ar^{+}}$  ion dose during the desorption step of the 6th cycle (C6) of Si PA-ALE processes. The ALE conditions are the same as those in Table I. The vertical axis denotes the accumulated number of species desorbed from a unit area of the substrate surface during the desorption step of this cycle. The adsorption steps for (a) and (b) and (c) are the thin Cl layer deposition and those for (c) and (d) are the thick Cl layer deposition. The incident energy of  $\\mathsf{Ar^{+}}$  ions is  $20\\mathrm{eV}$  for (a) and (c) and  $50\\mathrm{eV}$  for (b) and (d).","id":"train/atomic-layer-etching/simulation-usecase/26/fig_9","sample_id":"atomic-layer-etching/simulation-usecase/26/fig_9","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the number of species per cm² as a function of Ar⁺ ion dose for thin Cl layer deposition at 20 eV.\"},{\"panel_id\":\"b\",\"text\":\"The chart shows the number of species per cm² as a function of Ar⁺ ion dose for thin Cl layer deposition at 50 eV.\"},{\"panel_id\":\"c\",\"text\":\"The chart shows the number of species per cm² as a function of Ar⁺ ion dose for thick Cl layer deposition at 20 eV.\"},{\"panel_id\":\"d\",\"text\":\"The chart shows the number of species per cm² as a function of Ar⁺ ion dose for thick Cl layer deposition at 50 eV.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ar⁺ Ion Dose (×10¹⁶/cm²) | Si   | Cl   | SiCl | SiCl₂ | SiCl₃ |\\n|---------------------------|------|------|------|-------|-------|\\n| 0                        | 0    | 0    | 0    | 0     | 0     |\\n| 5                        | 1    | 2    | 3    | 0.5   | 0.2   |\\n| 10                       | 2    | 3    | 4    | 0.8   | 0.3   |\\n| 15                       | 3    | 4    | 4.5  | 1.0   | 0.4   |\\n| 18                       | 3.5  | 4.5  | 5    | 1.2   | 0.5   |\"},{\"panel_id\":\"b\",\"text\":\"| Ar⁺ Ion Dose (×10¹⁶/cm²) | Si   | Cl   | SiCl | SiCl₂ | SiCl₃ |\\n|---------------------------|------|------|------|-------|-------|\\n| 0                        | 0    | 0    | 0    | 0     | 0     |\\n| 5                        | 10   | 1    | 2    | 0.5   | 0.2   |\\n| 10                       | 20   | 1.5  | 2.5  | 0.8   | 0.3   |\\n| 15                       | 30   | 2    | 3    | 1.0   | 0.4   |\\n| 18                       | 35   | 2.2  | 3.2  | 1.2 | -   |\"},{\"panel_id\":\"c\",\"text\":\"| Ar⁺ Ion Dose (×10¹⁶/cm²) | Si   | Cl   | SiCl | SiCl₂ | SiCl₃ |\\n|---------------------------|------|------|------|-------|-------|\\n| 0                        | 0    | 0    | 0    | 0     | 0     |\\n| 5                        | 0.5  | 25   | 3    | 1.0   | 0.5   |\\n| 10                       | 1    | 28   | 4    | 1.5   | 0.8   |\\n| 15                       | 1.2  | 30   | 4.5  | 1.8   | 1.0   |\\n| 18                       | 1.5  | 32   | 5    | 2.0   | 1.2   |\"},{\"panel_id\":\"d\",\"text\":\"| Ar⁺ Ion Dose (×10¹⁶/cm²) | Si   | Cl   | SiCl | SiCl₂ | SiCl₃ |\\n|---------------------------|------|------|------|-------|-------|\\n| 0                        | 0    | 0    | 0    | 0     | 0     |\\n| 5                        | 10   | 20   | 3    | 1.0   | 0.5   |\\n| 10                       | 20   | 25   | 4    | 1.5   | 0.8   |\\n| 15                       | 30   | 28   | 4.5  | 1.8   | 1.0   |\\n| 18                       | 35   | 30   | 5    | 2.0   | 1.2   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"These images focus on the accumulated numbers of desorbed species in the desorption step of C6 as functions of the Ar+ ion dose under various conditions\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The accumulated numbers of desorbed monoatomic Si atoms and SiCl radicals, which are the next highest after monoatomic Cl, are almost the same initially. However, as the amount of Cl on the\\nsurface depletes, the rates of desorption for SiCl and SiCl2 decrease while monoatomic Si atoms continue to desorb. Furthermore, the desorption rate of monoatomic Si decreases as that of monoatomic Cl also decreases. It should be noted that the desorption of monoatomic Si does not mean the etching is purely due to physical sputtering. The physical sputtering yield of Si by 20 eV Ar+ ion bombardment is null essentially. The enhanced desorption of monoatomic Si atoms is likely caused by the bond breaking of Si atoms by Cl atoms remaining in the substrate\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. For thin Cl deposited films at high Ar+ ion, Si atoms mostly desorb as Cl bonded species in an\\nearly stage of the desorption step\\n2. For thick Cl deposited films, the desorption of SiClm is observed only in the very early stage of the Ar+ ion injection\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":10,"y":44,"width":632,"height":458},{"panel_id":"b","x":709,"y":44,"width":636,"height":459},{"panel_id":"c","x":10,"y":572,"width":632,"height":498},{"panel_id":"d","x":709,"y":572,"width":636,"height":498}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/26/Surface damage formation during atomic layer etching of silicon with chlorine adsorption.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":1345,"height":1070,"image_format":"jpeg","image_sha256":"8664e3e4cd1075704ed893e7c58909d0e8e5f89e47d5da61a5dffe634714df27","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_27_fig_3.jpg","caption":"FIG. 3. (a) Contributions of Ga-O and Ga-N to the Ga3d peak area plotted as a function of sputter depth. The vertical dashed line indicates the spectra shown in (b). (b) The Ga3d signal after  $20\\mathrm{s}$  of sputtering, highlighting the positions of the Ga-O and the Ga-N peaks at  $20.5$  and  $19.3\\mathrm{eV}$ , respectively.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_3","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"stacked spectra chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"XPS depth profile showing the Ga-N and Ga-O contributions to the Ga3d peak area as a function of sputter depth. The Ga-O signal is highest at the surface and decreases with depth, indicating an oxidized surface layer extending approximately 10 nm into the film. The Ga-N signal increases with depth, reaching a maximum in the bulk region.\"},{\"panel_id\":\"b\",\"text\":\"Fitted Ga3d XPS spectrum after 20 s of sputtering, showing deconvolution into Ga-N (19.3 eV) and Ga-O (20.5 eV) components. At this depth, both bonding environments contribute to the signal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputter depth (nm) | Ga-N Area (CPS eV) | Ga-O Area (CPS eV) |\\n|---|---|---|\\n| 0 | 2200 | 4100 |\\n| 1 | 3500 | 2800 |\\n| 3 | 4200 | 1900 |\\n| 5 | 4500 | 1500 |\\n| 10 | 4900 | 1000 |\\n| 20 | 5400 | 600 |\\n| 30 | 5800 | 100 |\\n| 35 | 5600 | 0 |\\n| 40 | 3500 | 0 |\\n| 45 | 100 | 0 |\"},{\"panel_id\":\"b\",\"text\":\"| Binding Energy (eV) | Envelope (Norm.) | Ga-N Component (Norm.) | Ga-O Component (Norm.) |\\n|---|---|---|---|\\n| 23.0 | 0.00 | 0.00 | 0.00 |\\n| 21.5 | 0.15 | 0.00 | 0.15 |\\n| 20.6 | 0.90 | 0.45 | 0.45 |\\n| 19.8 | 1.00 | 0.80 | 0.20 |\\n| 19.4 | 0.90 | 0.85 | 0.05 |\\n| 18.0 | 0.20 | 0.20 | 0.00 |\\n| 16.0 | 0.00 | 0.00 | 0.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It marks the sputter depth (after 20 s of sputtering) corresponding to the spectrum shown in subfigure (b).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The oxidized layer represents surface contamination that the ALE process aims to remove, and it affects the initial etch rate before reaching stoichiometric GaN.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Ga-O signal is highest at the surface (~4000 CPS·eV) and decreases rapidly within the first 10 nm, indicating a heavily oxidized surface layer. Beyond ~25 nm sputter depth, the Ga-O signal becomes negligible while Ga-N dominates, representing the transition to stoichiometric GaN in the bulk. This oxidized layer formed during storage in ambient conditions.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Ga-N is at 19.3 eV and Ga-O is at 20.5 eV.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":12,"width":346,"height":424},{"panel_id":"b","x":384,"y":12,"width":282,"height":424}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/Isotropic atomic layer etching of GaN using SF6 plasma and Al(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"27","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":666,"height":436,"image_format":"jpeg","image_sha256":"e403971f719da76da3c452f544fffb5c3598eadac368d3b38fd8f2a0ed0cb996","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_27_fig_4.jpg","caption":"FIG. 4. Gibbs free energy change  $(\\Delta G)$  as a function of temperature for the fluorination half-cycle considering both the conversion (blue) and chemical vapor etching (red) reactions. Solid lines indicate plasma fluorination reactions (F radicals), while the thermal fluorination reactions (HF) are indicated by dashed lines. A vertical dashed line at  $450^{\\circ}C$  highlights the change for the plasma fluorination reactions from the preferred conversion regime to the preferred etching regime. The ligand-exchange half-cycle is also included (dashed-dotted line, green) and shows a negative  $\\Delta G$  over the entire evaluated temperature range.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_4","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"DFT simulation results showing Gibbs free energy change (ΔG) vs. temperature for GaN ALE reactions. Plasma fluorination (F radicals, solid lines) shows both conversion and CVE reactions are thermodynamically favorable (ΔG < 0), with conversion preferred below 450°C. Thermal fluorination (HF, dashed lines) shows conversion is only favorable below ~250°C and CVE is unfavorable at all temperatures. The TMA ligand-exchange reaction (dash-dotted green) is favorable across the entire temperature range.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | F radical etching (eV/GaN) | F radical conversion (eV/GaN) | TMA Ligand-exchange (eV/GaN) | HF etching (eV/GaN) | HF conversion (eV/GaN) |\\n|---|---|---|---|---|---|\\n| -200 | -18.0 | -20.8 | -0.6 | 0.2 | -2.3 |\\n| 0 | -16.6 | -18.6 | -1.5 | 0.4 | -1.4 |\\n| 200 | -15.2 | -16.4 | -2.3 | 0.7 | -0.6 |\\n| 300 | -14.5 | -15.2 | -2.6 | 0.8 | -0.1 |\\n| 400 | -13.8 | -14.1 | -3.0 | 0.9 | 0.5 |\\n| 450 | -13.5 | -13.5 | -3.2 | 1.0 | 0.9 |\\n| 500 | -13.1 | -13.0 | -3.4 | 1.1 | 1.2 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The DFT simulations show that F radical reactions have much more negative ΔG values than HF reactions, indicating stronger thermodynamic driving force. For HF, the conversion reaction becomes unfavorable above ~250°C and CVE is never favorable. In contrast, F radical conversion remains favorable up to 450°C with ΔG values around -15 to -20 eV/GaN, explaining why GaN is resistant to thermal fluorination but can be effectively fluorinated using plasma.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The process transitions from the preferred conversion regime to the preferred etching regime, making it unsuitable for self-limiting ALE.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The reaction is thermodynamically favorable and can proceed spontaneously.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. The ligand-exchange reaction has negative ΔG across the entire evaluated temperature range, so it does not thermodynamically constrain the process window.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":653,"height":532}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/Isotropic atomic layer etching of GaN using SF6 plasma and Al(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"27","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":655,"height":536,"image_format":"jpeg","image_sha256":"7d6b9f50ee937491497f376e2869e192b48ae880553407ddd7efe6e64fda2846","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_3_fig_13.jpg","caption":"FIG. 13. (Color online) Trends in the ALE synergy  $(S_{y})$  as a function of pulse times for an AR of 4. (a)  $S_{y}$  as a function of ion bombardment time  $(T_{I})$  for three values of passivation time  $(T_{P})$ . (b)  $S_{y}$  as a function of  $T_{I}$  and  $T_{P}$ . The solid black line represents  $\\mathrm{EPC} = 1$  and the dotted lines bound the range  $0.9 < \\mathrm{EPC} < 1.1$ .","id":"train/atomic-layer-etching/simulation-usecase/3/fig_13","sample_id":"atomic-layer-etching/simulation-usecase/3/fig_13","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"contour heatmap"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the relationship between ALE synergy and ion bombardment time, showing a peak at around 0.5 seconds.\"},{\"panel_id\":\"b\",\"text\":\"The heatmap visualizes the ALE synergy across different passivation times and ion bombardment times, with color intensity indicating the level of synergy.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Bombardment Time (Ti, s) | ALE Synergy (%) Tp = 44 ms | Tp = 66 ms | Tp = 88 ms |\\n|------------------------------|----------------------------|------------|------------|\\n| 0.00                         | 90                         | 85         | 82         |\\n| 0.25                         | 93                         | 91         | 86         |\\n| 0.50                         | 92                         | 89         | 87         |\\n| 0.75                         | 89                         | 87         | 86         |\\n| 1.00                         | 87                         | 86         | 85         |\\n| 1.25                         | 85                         | 84         | 82         |\\n| 1.50                         | 84                         | 83         | 81         |\"},{\"panel_id\":\"b\",\"text\":\"| Passivation Time (Tp, ms) | Ion Bombardment Time (Ti, s) | ALE Synergy (%) |\\n|---------------------------|------------------------------|------------------|\\n| 5                         | 0.30                         | ~75             |\\n| 25                        | 0.30                         | ~85             |\\n| 45                        | 0.30                         | ~90             |\\n| 64                        | 0.30                         | ~88             |\\n| 84                        | 0.30                         | ~85             |\\n| 104                       | 0.30                         | ~82             |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For short TI (<0.15 s), a large fraction of the passivated Si which covers the surface of the feature is not removed during the ion bombardment step. The continuous etching produced by ions in the passivation phase then makes a significant contribution to the total rate of etching, which then lowers Sy.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For long TI (>1 s), the passivated Si covering the feature is completely removed and the now bare silicon is exposed to ion bombardment for an extended time. The continuous etching produced during the ion bombardment step by the presence of chlorine radicals and physical sputtering then accounts for a significant portion of the EPC, which then lowers Sy\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In general, to minimize nonidealities during ALE, the  optimized process should use the shortest pulse times which result in ~1ML material removal per cycle for the largest AR feature being etched. These optimized process times will be given by the point closest to the origin on the solid line in image (b). These conditions also correspond to the smallest contribution of continuous chemical sputtering, and will result in the smoothest surface and widest processing window in terms of AR.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"b","x":0,"y":433,"width":572,"height":549},{"panel_id":"a","x":3,"y":2,"width":568,"height":423}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/images/fig_13.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/3/Atomic layer etching of 3D structures in silicon Self-limiting.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"3","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":575,"height":983,"image_format":"jpeg","image_sha256":"faeb14d46918ec4feedfa8aed401ba7941fa4bf630df57d3b36943717e0b7ef8","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_31_figure_3.jpg","caption":"Figure 3. Electronic potential energies  $(\\Delta E)$  and Gibbs free energies  $(\\Delta G^{80^{\\circ}C})$  of nickel formate  $\\left(\\mathrm{Ni}(\\mathrm{FA})_{2}\\right)$  and formamidinate (Ni $\\left(\\mathrm{FAmd}\\right)_2)$  oligomer (olig.) complexes relative to the monomer (mono.) complex. Energy differences are normalized by nuclearity (nucl.) for comparison on a per-metal atom basis.","id":"train/atomic-layer-etching/simulation-usecase/31/figure_3","sample_id":"atomic-layer-etching/simulation-usecase/31/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Multiple line chart comparing electronic potential energies (ΔE) and Gibbs free energies (ΔG at 80 °C) per metal atom for nickel formate (FA) and formamidinate (FAmd) oligomers. The FA system shows its lowest ΔG at the dimer stage, while the FAmd system reaches its minimum ΔG at the trimer. These trends indicate ligand-dependent stabilization as oligomer size increases.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Oligomer State | ΔE_FA (eV) | ΔG_FA (80°C)(eV) | ΔE_FAmd(eV) | ΔG_FAmd (80°C)(eV) |\\n|----------------|-------------|------------------|--------------|------------------|\\n| Monomer | 0.0    | 0.0    | 0.0     | 0.0     |\\n| Dimer   | ~ -1.0 | ~ -0.5 | ~ -1.50 | ~ -0.75 |\\n| Trimer  | ~ -1.0 | ~ -0.4 | ~ -1.90 | ~ -1.25 |\\n| Tetramer| ~ -0.6 | ~ 0.00 | ~ -1.10 | ~ -0.25  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Trimer.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The FA complexes reach their lowest Gibbs free energy at the dimer stage, after which stability decreases for larger oligomers. In contrast, the FAmd complexes continue to stabilize beyond the dimer, reaching a minimum at the trimer before becoming less favorable at the tetramer. This indicates that ligand chemistry strongly influences the preferred oligomer size.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Ligand systems that favor specific oligomer sizes may influence volatility, transport, or surface reactivity in vapor-phase processes. The observed preference for FA dimers and FAmd trimers suggests that ligand choice can tune aggregation behavior, which may be relevant when designing precursors for controlled deposition or etching.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":7,"width":610,"height":425}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/Thermodynamics of Atomic Layer Etching Chemistry on Copper and Nickel Surfaces from First Principles.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"31","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":613,"height":436,"image_format":"jpeg","image_sha256":"b5330113e43df77c12411a6862da4e08159ed6d37fe5295828f2e04eb21122c9","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_31_figure_4.jpg","caption":"Figure 4. Adsorption energies of different terminations of the  $\\mathrm{Ni / N}$ ,  $\\mathrm{Cu / N}$ ,  $\\mathrm{Ni / O}$ , and  $\\mathrm{Cu / O}$  systems as a function of coverage. The labels \"PR\" and \"MP\" represent the \"pairing row\" and \"missing row\" reconstructions of the (110) surfaces, respectively.","id":"train/atomic-layer-etching/simulation-usecase/31/figure_4","sample_id":"atomic-layer-etching/simulation-usecase/31/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Multiple line chart showing Gibbs free energy of adsorption (ΔG_ads^80 °C) versus nitrogen coverage for the Ni/N system across several crystallographic facets. Adsorption becomes progressively less favorable with increasing coverage for all facets.\"},{\"panel_id\":\"b\",\"text\":\"Multiple line chart showing ΔG_ads^80 °C versus nitrogen coverage for the Cu/N system. Adsorption energies are consistently more positive than in Ni/N, indicating weaker nitrogen binding on Cu surfaces.\"},{\"panel_id\":\"c\",\"text\":\"Multiple line chart showing ΔG_ads^80 °C versus oxygen coverage for the Ni/O system, including labeled surface reconstructions such as (100) p(2×2), (110) (2×1)-MR, and (3×1)-MR. Oxygen adsorption is strongly favorable at low coverage and weakens with increasing coverage.\"},{\"panel_id\":\"d\",\"text\":\"Multiple line chart showing ΔG_ads^80 °C versus oxygen coverage for the Cu/O system, including pairing-row reconstructions. Adsorption becomes less favorable with increasing oxygen coverage and approaches thermoneutral behavior at high coverage.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Coverage θ_N (Å⁻²) | Blue Solid | Green Dash | Red Dash-Dot | Blue Dash | Purple Dash Dot | Yellow Solid |Teal Dash |\\n|---|---|---|---|---|---|---|---|\\n| 0.025 | ~ -0.6 | ~ +0.1 | ~ +0.3 | ~ 0.0  | ~ -0.25 | ~ +0.0 |~ +0.2 |\\n| 0.05  | ~ -0.4 | ~ +0.1 | ~ +0.3 | ~ +0.1 | ~ +0.00 | ~ +0.1 |~ +0.2 |\\n| 0.10  | ~ +0.25| ~ +0.4 | ~ +0.9 | ~ +0.4 | ~ +0.20 | ~ +0.5 |~ +0.5 |\\n| 0.15  | ~ +1.0 | ~ +1.1 | ~ +1.5 |   -    | ~ +0.90 | ~ +1.0 |~ +1.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Coverage θ_N (Å⁻²) | Blue Solid | Green Dash | Red Dash-Dot | Blue Dash | Purple Dash Dot | Yellow Solid |Teal Dash |\\n|---|---|---|---|---|---|---|---|\\n| 0.025 | ~ +0.8 | ~ +1.5 | ~ +1.7 | ~ +1.4 | ~ +1.00 | ~ +1.2 |~ +1.2 | \\n| 0.05  | ~ +0.8 | ~ +1.5 | ~ +1.6 | ~ +1.5 | ~ +1.00 | ~ +1.2 |~ +1.2 |\\n| 0.10  | ~ +1.5 | ~ +1.8 | ~ +2.1 |   -    |   -     | ~ +1.8 |~ +1.5 |\\n| 0.125 | ~ +1.8 | ~ +0.0 | ~ +2.9 |   -    |   -     | ~ +2.1 |~ +2.3 |\"},{\"panel_id\":\"c\",\"text\":\"| Coverage θ_O (Å⁻²) | Blue Solid | Green Dash | Red Dash-Dot | Blue Dash | Purple Dash Dot | Yellow Solid |Teal Dash |\\n|---|---|---|---|---|---|---|---|\\n| 0.05 | ~ -2.1 | ~ -1.8 | ~ -1.9 | ~ -1.9 | ~ -2.1  | ~ -2.0 |~ -2.0 |\\n| 0.10 | ~ -1.6 | ~ -1.5 | ~ -1.5 | ~ -1.6 | ~ -1.75 | ~ -1.6 |~ -1.5 |\\n| 0.15 | ~ -1.25| ~ -1.3 | ~ -0.9 | ~ -1.4 | ~ -1.25 | ~ -1.2 |~ -1.3 |\\n| 0.20 |   -    |   -    | ~ -0.5 | ~ -1.0 | ~ -1.00 | ~ -0.9 |~ -1.0 |\"},{\"panel_id\":\"d\",\"text\":\"| Coverage θ_O (Å⁻²) | Blue Solid | Green Dash | Red Dash-Dot | Blue Dash | Purple Dash Dot | Yellow Solid |Teal Dash |\\n|---|---|---|---|---|---|---|---| \\n| 0.05 | ~ -1.6 | ~ -1.4 | ~ -1.3  | ~ -1.4 | ~ -1.60 | ~ -1.4 |~ -1.5 | \\n| 0.10 | ~ -1.3 | ~ -1.2 | ~ -0.75 | ~ -1.2 | ~ -1.20 | ~ -1.0 |~ -1.2 |\\n| 0.15 | ~ -0.3 | ~ -0.6 | ~ -0.00 | ~ -0.9 | ~ -0.90 | ~ -0.6 |~ -0.9 |\\n| 0.20 |    -   |    -   |   -     | ~ -0.4 | ~ -0.50 |    -   |~  0.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"80 °C (indicated by the superscript in ΔG_ads^80 °C).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, the energy generally increases (becomes more positive/less stable) as coverage increases.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Subfigure (c) explicitly labels several surface reconstructions for the Ni/O system. For the (100) facet, it identifies $p(2 \\\\times 2)$ and $c(2 \\\\times 2)$ structures, while for the (110) facet, it highlights $(2 \\\\times 1)-MR$ and $(3 \\\\times 1)-MR$ \\\"missing row\\\" reconstructions. At the lowest coverage shown, the $(110) (2 \\\\times 1)-MR$ reconstruction (marked with a green cross) appears to be the most stable, having the lowest $\\\\Delta G$ value below -2.0 eV.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Ni/N system (a).\"}]}]","bbox":[{"panel_id":"a","x":11,"y":0,"width":481,"height":365},{"panel_id":"b","x":580,"y":0,"width":482,"height":363},{"panel_id":"c","x":10,"y":415,"width":482,"height":361},{"panel_id":"d","x":581,"y":415,"width":481,"height":357}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/31/Thermodynamics of Atomic Layer Etching Chemistry on Copper and Nickel Surfaces from First Principles.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"31","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":1094,"height":861,"image_format":"jpeg","image_sha256":"6c62d2ff52c7e64e6adadda23dc0559fb2fbb69ed3f910694687769be5152b35","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_4f80ad6bb517a6a7dc64ff0faf1cdcbf0b92399b65f69269988d6c0387e6b73f.jpg","caption":"(a)","id":"train/atomic-layer-etching/simulation-usecase/33/4f80ad6bb517a6a7dc64ff0faf1cdcbf0b92399b65f69269988d6c0387e6b73f","sample_id":"atomic-layer-etching/simulation-usecase/33/4f80ad6bb517a6a7dc64ff0faf1cdcbf0b92399b65f69269988d6c0387e6b73f","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart compares the total energy per atom over time between two computational methods: Density Functional Theory (DFT) and a Neural Network Potential (NNP). The x-axis represents time in picoseconds (ps), while the y-axis shows energy in electron volts per atom (eV/atom). Both methods exhibit similar dynamic energy profiles with overlapping fluctuations, suggesting that the NNP method reliably captures the same energy behavior as the more computationally expensive DFT. This highlights NNP’s potential for efficient and accurate simulations in extended time or larger systems.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (ps) | DFT (eV/atom) | NNP (eV/atom) |\\n|-----------|---------------|---------------|\\n| 0         | -7.20         | -7.22         |\\n| 10        | -7.05         | -7.07         |\\n| 20        | -6.95         | -6.98         |\\n| 30        | -7.10         | -7.13         |\\n| 40        | -6.90         | -6.92         |\\n| 50        | -7.00         | -7.03         |\\n| 60        | -6.85         | -6.88         |\\n| 70        | -6.95         | -6.97         |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The NNP model closely mirrors the energy profile predicted by DFT over the full simulation time, maintaining similar fluctuation amplitudes and trends. While minor deviations are visible at certain time points, the overall agreement between the two methods supports the accuracy of the NNP in reproducing DFT-level energy predictions, making it suitable for large-scale simulations.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"DFT\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Reduced computational cost, Faster simulation times, Comparable accuracy for energy predictions, Scalability to larger systems or longer timescales\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":414,"height":200}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/4f80ad6bb517a6a7dc64ff0faf1cdcbf0b92399b65f69269988d6c0387e6b73f.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/4f80ad6bb517a6a7dc64ff0faf1cdcbf0b92399b65f69269988d6c0387e6b73f.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/Atomistic Simulation of HF Etching Process of Amorphous.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":414,"height":200,"image_format":"jpeg","image_sha256":"8ecadbbd44329ab71781a6f76519f32aa67b97ccff9800953807cbf2b665c3bb","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_33_figure_6.jpg","caption":"(a) Figure 6. (a) The number of Si, N, H, and F atoms in the simulation cell during the etching process at  $E_{\\mathrm{in}} = 50 \\mathrm{eV}$  and  $\\theta_{\\mathrm{in}} = 0^{\\circ}$ . (b) Etching yield as a function of incident energy. The black dashed line represents the fitted function of the universal etching behavior.[52] The error bars indicate the standard deviations among three independent runs.","id":"train/atomic-layer-etching/simulation-usecase/33/figure_6","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_6","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents the number of Si, N, H, and F atoms in a system as a function of HF dose. As HF exposure increases from 0 to 200 HF/nm², both silicon (Si) and nitrogen (N) atoms show a steady decline, indicating progressive etching or transformation. Fluorine (F) atoms increase significantly at lower doses and then level off, suggesting that surface fluorination saturates over time. Hydrogen (H) levels remain relatively constant, suggesting limited involvement or a dynamic equilibrium. Together, these trends reveal the elemental evolution occurring during HF exposure and suggest that nitrogen and silicon are preferentially removed or transformed while fluorine accumulates on the surface.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Dose (HF/nm²) | Si | N | H | F |\\n|---------------|----|----|----|----|\\n| 0             | 400 | 500 | 100 | 0   |\\n| 50            | 370 | 460 | 110 | 120 |\\n| 100           | 340 | 420 | 110 | 140 |\\n| 150           | 310 | 390 | 110 | 150 |\\n| 200           | 280 | 360 | 110 | 150 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the HF dose increases, the number of nitrogen and silicon atoms steadily decreases, indicating progressive material removal or chemical transformation. In contrast, the number of fluorine atoms increases sharply at lower doses and then plateaus, suggesting surface saturation or limited incorporation beyond a certain dose. The hydrogen atom count remains nearly constant, implying either low reactivity or steady-state behaviour. These trends highlight the distinct chemical roles each element plays during HF exposure.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Fluorine (F)\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Si and N are progressively removed ,Fluorine accumulates , Hydrogen remains constant\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":281,"height":264}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/Atomistic Simulation of HF Etching Process of Amorphous.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"33","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:image_caption"},"width":281,"height":264,"image_format":"jpeg","image_sha256":"9c3fcdadec4a42bd6cdfdb4139b4146fbdc63a77a9798a8e6deedefc650ba84a","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_4_figure_2.jpg","caption":"Figure 2. Etch per cycle of TiN films with varying  $\\mathbb{C}\\mathbb{C}\\mathbb{I}_4$  pulse time at  $460^{\\circ}C$  6 s long  $\\mathbb{N}_2$  purges were used. The figure compares the EPC for the binary process (red) to the CVE process (blue).","id":"train/atomic-layer-etching/simulation-usecase/4/figure_2","sample_id":"atomic-layer-etching/simulation-usecase/4/figure_2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Figure 2 illustrates the etch per cycle (EPC) of TiN films as a function of CCl₄ pulse time at 460 °C, comparing two etching processes: the NbF₅–CCl₄ binary process and the CCl₄-only CVE process. The EPC increases with CCl₄ pulse time for both processes, with the binary process showing higher etch rates (up to ~0.8 Å at 3 s) compared to the CVE process (~0.5 Å at 3 s). No clear saturation of EPC is observed for either process, and no etching occurs when only NbF₅ is applied, consistent with computational predictions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| CCl₄ pulse time [s] | Etch per cycle NbF₅–CCl₄ [Å] | Etch per cycle CCl₄-only CVE [Å] |\\n|---|---|---|\\n| 0.0 | 0.0 | 0.0 |\\n| 0.5 | 0.3 | 0.3 |\\n| 1.0 | 0.4 | 0.4 |\\n| 3.0 | 0.8 | 0.5 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For both processes, the etch per cycle increases with increasing CCl₄ pulse time. The NbF₅–CCl₄ binary process consistently shows higher EPC values than the CCl₄-only CVE process. At 3 s pulse time, the EPC reaches approximately 0.8 Å for the binary process and 0.5 Å for the CVE process. No clear saturation is observed, indicating a nearly linear increase across the measured pulse times.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Enhanced reactivity of CCl₄ with fluorinated TiN surfaces, Formation of more volatile etch products such as TiCl₄ instead of less volatile species (TiCl₃, TiOCl, TiF₃), Possible three-dimensional etching due to fluorine diffusion through TiN grain boundaries, Self-limiting reactions of NbF₅ pulses followed by catalyzed CCl₄ adsorption at Nb sites\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.5 Å.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":4,"width":624,"height":595}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/Combining Experimental and DFT Investigation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":633,"height":598,"image_format":"jpeg","image_sha256":"c2a162c89584ecf10b088bb8ad3343b7d5a39c552229ba0526dc010bddeda724","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_4_figure_3.jpg","caption":"Figure 3. Effect of etch temperature on etch per cycle for the binary process (blue curve) as well as  $\\mathrm{CCl_4}$  alone (blue curve) process. Both the precursor pulses and the purge times were fixed to 1 and 6 s, respectively.","id":"train/atomic-layer-etching/simulation-usecase/4/figure_3","sample_id":"atomic-layer-etching/simulation-usecase/4/figure_3","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Figure 3 shows the effect of etch temperature on the etch per cycle of TiN films for the NbF₅–CCl₄ binary process and the CCl₄-only process. The etch per cycle increases with temperature for both processes, with the binary process consistently exhibiting higher etch rates. No significant etching occurs below 370 °C, and the trend is approximately linear up to 460 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Etch temperature [°C] | Etch per cycle NbF₅–CCl₄ [Å] | Etch per cycle CCl₄-only [Å] |\\n|---|---|---|\\n| 370 | 0.03 | 0.07 |\\n| 380 | 0.10 | 0.12 |\\n| 390 | 0.20 | 0.18 |\\n| 400 | 0.40 | 0.22 |\\n| 460 | 0.55 | 0.30 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch per cycle increases with temperature for both processes. The NbF₅–CCl₄ binary process consistently shows higher EPC values than the CCl₄-only process. At 370 °C, the EPC is 0.03 Å for the binary process and 0.07 Å for CCl₄-only. By 460 °C, the EPC rises to 0.55 Å for the binary process and 0.3 Å for the CCl₄-only process, demonstrating a nearly linear increase with temperature.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Enhanced reactivity of CCl₄ with fluorinated TiN surface, Formation of more volatile etch products such as TiCl₄ instead of less volatile species, Self-limiting reactions of NbF₅ followed by catalyzed CCl₄ adsorption, Higher activation of surface reactions at elevated temperatures\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 0.4 Å.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":4,"width":633,"height":589}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/Combining Experimental and DFT Investigation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":642,"height":595,"image_format":"jpeg","image_sha256":"c59b09f92224112076fab01d65532c1691bf4a04c3d7e988beec6c61ba4d045c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_4_figure_4.jpg","caption":"Figure 4. A change in the thickness values with total number of etch cycles at  $460^{\\circ}\\mathrm{C}$  for TiN,  $\\mathrm{SiO_2}$ ,  $\\mathrm{Al}_2\\mathrm{O}_3$ , and  $\\mathrm{Si}_3\\mathrm{N}_4$  films. A TiN is selectively etched away by  $\\mathrm{CCl_4}$  alone over other materials.","id":"train/atomic-layer-etching/simulation-usecase/4/figure_4","sample_id":"atomic-layer-etching/simulation-usecase/4/figure_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Figure 4 shows the change in film thickness with the total number of etch cycles at 460 °C for TiN, SiO₂, Al₂O₃, and Si₃N₄ films. The TiN film is selectively etched by the CCl₄-only CVE process, with thickness decreasing linearly at approximately 0.3 Å per cycle, reaching about 13 nm after 500 cycles. In contrast, the thickness of Al₂O₃, SiO₂, and Si₃N₄ remains unchanged even after 1000 etch cycles, demonstrating the high selectivity of TiN etching over other materials.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Etch cycles | TiN (ALD) | Al₂O₃ (ALD) | SiO₂ (TOx) | Si₃N₄ (LPCVD) |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 100 | 3 | 0 | 0 | 0 |\\n| 300 | 9 | 0 | 0 | 0 |\\n| 500 | 13 | 0 | 0 | 0 |\\n| 1000 | 30 | 0 | 0 | 0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"no\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The TiN film thickness decreases linearly with the number of etch cycles, reflecting an etch rate of approximately 0.3 Å per cycle. In contrast, the thicknesses of Al₂O₃, SiO₂, and Si₃N₄ remain essentially unchanged even after 1000 cycles. This demonstrates the high selectivity of TiN etching over the other materials under the same process conditions.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Enables precise patterning and removal of TiN without damaging other substrate materials, Maintains integrity of dielectric layers (Al₂O₃, SiO₂) and nitride layers (Si₃N₄), Reduces the risk of contamination or undesired etching in multi-layer devices, Facilitates reliable manufacturing in microelectronics and nanofabrication processes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 13 nm.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":618,"height":567}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/Combining Experimental and DFT Investigation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":625,"height":575,"image_format":"jpeg","image_sha256":"3d50cc1867ca8e1cf4c6b42bf228f7c2de0551989d53b1b0b5a1a2774495f158","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_4_figure_7.jpg","caption":"Figure 7. An XPS depth profiling through remaining TiN film after partial etching at  $460^{\\circ} \\text{C}$ . Total 300 cycles were performed with  $0.5 \\text{s}$  of both  $\\text{NbF}_5$  and  $\\text{CCl}_4$  pulse lengths with  $6 \\text{s}$  of  $\\text{N}_2$  purges.","id":"train/atomic-layer-etching/simulation-usecase/4/figure_7","sample_id":"atomic-layer-etching/simulation-usecase/4/figure_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Figure 7 presents an XPS depth profile of a partially etched TiN film after 300 cycles of the NbF₅–CCl₄ etch process at 460 °C, with 0.5 s pulses and 6 s N₂ purges. The surface initially contains high carbon (~26%) and oxygen (~31%), along with nitrogen (~20%), titanium (~15%), niobium (~5%), and small amounts of fluorine and silicon (<2%). After 15 s of argon sputtering, carbon decreases to ~4% and further to ~3% after 60 s, while Ti, N, and O stabilize at ~24–32%. Nb and F remain at 5–6% and 0.9–1.7%, respectively, indicating partial incorporation or diffusion within the TiN surface layer. Si originates from the underlying SiO₂. The profile confirms the elemental distribution and partial etch depth of the TiN film.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputter time [s] | C1s (%) | N1s (%) | O1s (%) | F1s (%) | Si2p (%) | Ti2p (%) | Nb3d (%) |\\n|---|---|---|---|---|---|---|---|\\n| 0 | 26 | 20 | 31 | 1.5 | 1.5 | 15 | 5 |\\n| 15 | 4 | 32 | 28 | 1.0 | 1.5 | 24 | 5.5 |\\n| 60 | 3 | 32 | 28 | 0.9 | 1.5 | 24 | 6 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Initially, the surface contains a high carbon concentration (~26%) and lower Ti (~15%) and N (~20%). After 15 s of argon sputtering, carbon decreases sharply to ~4%, while Ti, N, and O increase slightly and stabilize at ~24–32%. After 60 s of sputtering, carbon drops further to ~3%, and Ti, N, and O remain stable, reflecting the bulk composition of the remaining TiN film. This trend shows that surface contamination diminishes with depth while the elemental composition of the film core is revealed.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Determines the elemental distribution within the remaining TiN film, Reveals surface contamination and its removal with sputtering, Confirms partial incorporation of Nb and F into the TiN surface layer, Helps assess the effectiveness and uniformity of the etch process, Provides information about possible diffusion or adsorption mechanisms during etching\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 24%.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":595,"height":544}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/4/Combining Experimental and DFT Investigation.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"4","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":600,"height":550,"image_format":"jpeg","image_sha256":"f874063fa37bbbf369670d108326bc4a8f3437bbfa1fbba410a2c2d73006a500","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_5_fig_5.jpg","caption":"Fig. 5 (a) Density profile of Ge and Cl element change before and after 200 times Ar bombardment with different bombardment energy on  $25\\mathrm{eV}$  chlorinated surface. (b) Total etched number change (blue: Ge, red: Cl) with Ar bombardment energy for each chlorination condition, averaged from three samples with error bars calculated.","id":"train/atomic-layer-etching/simulation-usecase/5/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/5/fig_5","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"},{"panel_id":"b","label":"multiple line chart"},{"panel_id":"c","label":"multiple line chart"},{"panel_id":"d","label":"multiple line chart"},{"panel_id":"e","label":"multiple line chart"},{"panel_id":"f","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the number density of Ge and Cl atoms at different z distances for initial and final states on 25 eV chlorinated surface.\"},{\"panel_id\":\"b\",\"text\":\"The line chart shows the number density of Ge and Cl atoms at different z distances for initial and final states on 50 eV chlorinated surface.\"},{\"panel_id\":\"c\",\"text\":\"The line chart shows the number density of Ge and Cl atoms at different z distances for initial and final states on 75 eV chlorinated surface.\"},{\"panel_id\":\"d\",\"text\":\"The line chart shows the number density of Ge and Cl atoms at different z distances for initial and final states on 100 eV chlorinated surface.\"},{\"panel_id\":\"e\",\"text\":\"The line chart illustrates the total etched numbers of Ge atoms at different Ar energies.\"},{\"panel_id\":\"f\",\"text\":\"The line chart illustrates the total etched numbers of Cl atoms at different Ar energies.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| z distance (Å) | Density Ge_initial | Density Cl_initial | Density Ge_final | Density Cl_final |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 50 | 75 | 75 | 65 | 50 |\"},{\"panel_id\":\"b\",\"text\":\"| z distance (Å) | Density Ge_initial | Density Cl_initial | Density Ge_final | Density Cl_final |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 50 | 75 | 75 | 50 | 25 |\"},{\"panel_id\":\"c\",\"text\":\"| z distance (Å) | Density Ge_initial | Density Cl_initial | Density Ge_final | Density Cl_final |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 50 | 75 | 75 | 40 | 10 |\"},{\"panel_id\":\"d\",\"text\":\"| z distance (Å) | Density Ge_initial | Density Cl_initial | Density Ge_final | Density Cl_final |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 50 | 75 | 75 | 15 | 5 |\"},{\"panel_id\":\"e\",\"text\":\"| Ar energy (eV) | Ge_5eV | Ge_10eV | Ge_25eV | Ge_50eV |\\n|---|---|---|---|---|\\n| 25 | 100 | 150 | 250 | 400 |\\n| 50 | 150 | 250 | 350 | 550 |\\n| 75 | 500 | 500 | 600 | 700 |\\n| 100 | 550 | 600 | 700 | 800 |\"},{\"panel_id\":\"f\",\"text\":\"| Ar energy (eV) | Ge_5eV | Ge_10eV | Ge_25eV | Ge_50eV |\\n|---|---|---|---|---|\\n| 25 | 50 | 100 | 200 | 300 |\\n| 50 | 100 | 200 | 250 | 400 |\\n| 75 | 250 | 250 | 400 | 500 |\\n| 100 | 250 | 300 | 450 | 600 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The threshold energy decreases as the chlorination energy increases.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Figure 5 demonstrates that both the chlorination energy and the Ar bombardment energy have a strong and interrelated influence on the onset and efficiency of etching. As the chlorination energy increases, the surface becomes more disordered and vulnerable to atom detachment, which lowers the threshold Ar energy required to initiate etching. At the same time, higher bombardment energies lead to increased etching of both Ge and Cl atoms, indicating higher etching rates. However, for lower chlorination energies, the etched atom numbers tend to reach a plateau as the available chlorine becomes limited, reflecting a self-limiting chemical etching process. In contrast, higher chlorination energies delay this plateau because more chlorine is initially present, allowing etching to continue over a larger range of bombardment energies before transitioning into physical sputtering.\"}]},{\"panel_id\":\"e\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Cl density peak starts to decrease and the Ge surface begins to recede already at an Ar bombardment energy of 25 eV, with the etching effect becoming more obvious at 50 eV.\"}]},{\"panel_id\":\"f\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Complete etching occurs between 60–75 eV for the 5 eV chlorinated surface, 75-100 eV for the 10 and 25 eV chlorinated surfaces, and above 100 eV for the 50 eV chlorinated surface.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":301,"height":230},{"panel_id":"b","x":307,"y":5,"width":301,"height":232},{"panel_id":"c","x":1,"y":238,"width":300,"height":229},{"panel_id":"d","x":311,"y":233,"width":294,"height":226},{"panel_id":"e","x":4,"y":473,"width":308,"height":220},{"panel_id":"f","x":315,"y":471,"width":288,"height":220}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/5/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/5/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/5/Computational modelling of atomic layer etching of chlorinated germanium surfaces by argon.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":608,"height":694,"image_format":"jpeg","image_sha256":"55f34b8678815727fcc8ceb6ad2cefba4257aca61f956a9d8acba3d06b1cb07c","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_6_fig2.jpg","caption":"FIG.2.  $\\mathbb{C}\\mathbb{I}_2$  mass flow, pressure, and DC bias during one 7 s ALE cycle. The applied DC bias is  $-65\\mathrm{V}$  for this case. The Ar flow (80 SCCM) is on continuously throughout the cycle. ICP power is 250 W.","id":"train/atomic-layer-etching/simulation-usecase/6/fig2","sample_id":"atomic-layer-etching/simulation-usecase/6/fig2","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multi-axis chart shows the evolution of DC self-bias voltage, chlorine gas flow, and chamber pressure over time. The process is divided into four clearly marked steps, each showing how plasma conditions and gas flow influence system parameters. The other panels are not separate figures but sub-intervals within (a). Their summaries have been removed to avoid redundancy and reflect the unified timeline.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | DC Self-Bias (V) | Cl₂ Flow (seccm) | P (mTorr) |\\n|----------|------------------|------------------|-----------|\\n| 0        | 0                | 0                | 22        |\\n| 1        | 0                | 20               | 22        |\\n| 2        | -60              | 20               | 21        |\\n| 3        | -60              | 0                | 20        |\\n| 4        | -60              | 0                | 20        |\\n| 5        | 0                | 20               | 21        |\\n| 6        | 0                | 0                | 22        |\\n| 7        | 0                | 0                | 22        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Step 1: Cl₂ On, ICP Off, Bias Off; Step 2: ICP On, Bias On; Step 3: Cl₂ Off, Bias Off; Step 4: All Off\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Pressure decreases slightly with ICP on, then returns to baseline; Cl₂ flow toggles between 20 seccm and 0 based on step.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"DC Self-Bias (V), Cl₂ Flow (seccm), Pressure (mTorr)\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":678,"height":475}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/Dynamics of plasma atomic layer etching Molecular.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":678,"height":475,"image_format":"jpeg","image_sha256":"0d4aa542f6fdebdad861e3355b99790a7e5c932dce2847f8daeface47db3a613","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_8_fig_4.jpg","caption":"FIG. 4. Gibbs free energy change  $(\\Delta \\mathbb{G})$  as a function of temperature for the fluorination half-cycle considering both the conversion (blue) and chemical vapor etching (red) reactions. Solid lines indicate plasma fluorination reactions (F radicals), while the thermal fluorination reactions (HF) are indicated by dashed lines. A vertical dashed line at  $450^{\\circ}C$  highlights the change for the plasma fluorination reactions from the preferred conversion regime to the preferred etching regime. The ligand-exchange half-cycle is also included (dashed-dotted line, green) and shows a negative  $\\Delta \\mathbb{G}$  over the entire evaluated temperature range.","id":"train/atomic-layer-etching/simulation-usecase/8/fig_4","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_4","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart illustrates the change in Gibbs free energy (ΔG) per unit GaN with respect to temperature for various etching methods.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | ΔG (eV/GaN) HF CVE |ΔG (eV/GaN) HF conversion |ΔG (eV/GaN) TMA ligand-exchange |ΔG (eV/GaN) F radical CVE |ΔG (eV/GaN) F radical conversion |\\n|---|---|---|---|---|---|\\n|-200 | 0 |-3|-0.5|-18.5|-21|\\n|-100 | 0 |-2|-1|-18|-20|\\n|0 | 0.5 |-2|-1.5|-16|-19|\\n|100 | 1 |-1|-2|-16|-18|\\n|200 | 1 |0|-2.5|-15|-17|\\n|300 | 1|0|-3|-14|-16|\\n|400 | 1.5 |1|-3.5|-14|-14|\\n|500 | 1.5 |1.5|-4|-14.5|-13|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The use of F radicals in a plasma as this lowers the gibbs energy the most.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"F radical conversion, as this has lower delta gibbs energies. Only from 450 degrees celcius and up will there be etching.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The chance decreases.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"+1 eV/GaN\"}]}]","bbox":[{"panel_id":"a","x":5,"y":3,"width":650,"height":532}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/Isotropic atomic layer etching of GaN using SF6 plasma and AI(CH3)3.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"8","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":655,"height":536,"image_format":"jpeg","image_sha256":"7d6b9f50ee937491497f376e2869e192b48ae880553407ddd7efe6e64fda2846","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_9_fig_7.jpg","caption":"Fig. 7. Schematic of the energy conversion process and snapshots of the density distribution with time evolution, the distributions on excited-state and ground-state potential energy curves are represented by solid and dashed lines, respectively. The vertical line  $Z = Z_{d}$  is the defined critical value, the  $\\mathrm{SiCl_2}$  molecule crossing this line can be identified as to be completely desorbed. Since the distribution for  $Z > Z_{d}$  is small, it was multiplied by a factor of  $4\\times 10^{5}$  if  $Z > Z_{d}$  to make it visible.","id":"train/atomic-layer-etching/simulation-usecase/9/fig_7","sample_id":"atomic-layer-etching/simulation-usecase/9/fig_7","subset":"multiple-line-chart","split":"train","classification":[{"panel_id":"a","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure illustrates the potential energy of a system as a function of surface coordinate Z at various time points (t=0 fs, t=5 fs, t=10 fs, t=400 fs). It shows the transition between the ground state and excited state, with terms like 'Excitation', 'Quenching', and 'Evolving' indicating the dynamics of the system.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Surface coordinate Z (Å) | Potential energy (eV) Ground state | Potential energy (eV) Excited state | Density Ground state |Density Excited state |\\n|---|---|---|---|---|\\n|-0.7 | 3 |4.2|-|-|\\n| 0 | 0 |3.1|Peak|Peak|\\n| 1 | 1.6 |2.6|-|-|\\n| 2 | 2.4 |2.5|-|-|\\n| 3 | 2.5 |2.5|-|-|\\n| 4 | 2.5 |2.5|-|-|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The critical distance is 3 angstroms, this is the distance from which a SiCl2 unit is defined to be fully desorbed.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Quenching is the transition from the excited state to the ground state by energy dissipation in the system, so from the red line to the blue line.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The excited state is induced by the absorption of photons with the correct energy. The amount of excited species reduces over time in the system, explaining the trend in the densities for the excited state and the ground state.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The desorption energy is 2.5 eV\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":665,"height":465}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/9/Mechanism of photoassisted atomic layer etching of chlorinated Si(111) surfaces Insight.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"9","first_classification_panel_id":"a","first_classification_label":"multiple line chart","caption_source":"content.json:img_caption"},"width":666,"height":470,"image_format":"jpeg","image_sha256":"3588df2fe5c35bd9f911f66bc1602725f1a2716519139c855aec31c9fbf7a99f","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"}