Datasets:

SciKnowOrg
/

ALD-E-ImageMiner

{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_12_fig_5.jpg","caption":"Fig. 5. Normalized integrated absorbances of the W-F stretching vibration at  $\\sim 680~\\mathrm{cm^{-1}}$  and the  $\\mathrm{Si - H}$  stretching vibrations at 2115 and  $2275~\\mathrm{cm^{-1}}$  versus  $\\mathrm{WF}_6$  exposure during the  $\\mathrm{WF}_6$  half-reaction at  $425\\mathrm{K}$ .","id":"train/atomic-layer-deposition/experimental-usecase/12/fig_5","sample_id":"atomic-layer-deposition/experimental-usecase/12/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This mulitple line plot represents the normalized, integrated absorbance of key IR vibrations (W–F stretch and Si–H stretches) against the total WF₆ exposure (Pressure * Time) during the WF₆ half-reaction at 425 K. It shows the Si–H signal decaying and the W–F signal growing as exposure increases, both saturating at high exposure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| WF<sub>6</sub> Exposure (Torr·min) | Normalized Integrated Absorbance | Stretch       |\\n|-----------------------------------|----------------------------------|---------------|\\n| 0.0 | 0.0  | W–F Stretch |\\n| 0.0 | 0.0  | Si–H Stretch |\\n| 0.1 | 0.9  | W–F Stretch |\\n| 0.1 | 0.1  | Si–H Stretch |\\n| 0.2 | 0.95 | W–F Stretch |\\n| 0.2 | 0.08 | Si–H Stretch |\\n| 0.4 | 1.0  | W–F Stretch |\\n| 0.4 | 0.03 | Si–H Stretch |\\n| 0.6 | 1.0  | W–F Stretch |\\n| 0.6 | 0.03 | Si–H Stretch |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Normalized integrated absorbance is a measure of the total area under an FTIR absorption peak after scaling it relative to a reference value. It is obtained by integrating the absorbance intensity across the specific wavenumber range associated with a vibrational mode. The resulting value is then divided by a chosen baseline or maximum signal so that different spectra can be compared on the same scale.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The W–F curve rises exactly as the Si–H curve falls, proving each new tungsten-fluoride bond directly replaces a silicon-hydride bond. Both curves saturate at the same exposure, confirming the reaction is self-limiting and consumes a fixed number of surface sites. This perfect mirroring shows the half-reaction is complete, reproducible, and ideal for controlled atomic-layer deposition.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The surface structure transitions completely from Si–H termination to W–F termination as WF₆ exposure increases to the saturation point. The concurrent and complementary nature of the Si–H decay and W–F growth indicates a direct, one-to-one replacement reaction. The saturation of both signals at the same exposure value suggests the reaction consumes a fixed number of surface Si–H sites to produce a corresponding number of W–F sites, defining the stoichiometry of the surface-limited process.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This provides exceptional conformality on high-aspect-ratio 3D structures, a critical advantage. Because the reaction saturates once all surface sites are consumed, it proceeds equally on all exposed surfaces, regardless of geometry. The film property controlled with atomic-scale precision is thickness, as each ALD cycle adds a reproducible, sub-nanometer layer, enabling exact deposition in nanoscale device features.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":596,"height":576}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/J.W. Klaus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"12","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":600,"height":581,"image_format":"jpeg","image_sha256":"841faa92a4f9945fa27694499581b9817dd445b12852d5cc27def86e0f38d093","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_experimental-usecase_12_fig_7.jpg","caption":"Fig. 7. Normalized integrated absorbances of the W-F stretching vibration at  $\\sim 680 \\mathrm{cm}^{-1}$  and the  $\\mathrm{Si - H}$  stretching vibrations at 2115 and  $2275 \\mathrm{cm}^{-1}$  versus  $\\mathrm{Si}_2\\mathrm{H}_6$  exposure during the  $\\mathrm{Si}_2\\mathrm{H}_6$  half-reaction at  $425 \\mathrm{K}$ .","id":"train/atomic-layer-deposition/experimental-usecase/12/fig_7","sample_id":"atomic-layer-deposition/experimental-usecase/12/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This mulitple line plot represents the normalized, integrated absorbance of key IR vibrations (W–F stretch and Si–H stretches) against the total Si₂H₆ exposure during the WF₆ half-reaction at 425 K. It shows the W–F signal decaying and the Si–H signal growing as exposure increases, both saturating at high exposure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Si₂H₆ Exposure (Torr min) | Normalized Integrated Absorbance | Stretch |\\n|---|---|-----|\\n| 0.0 | 0.0 | W-F Stretch |\\n| 0.0 | 0.0 | Si-H Stretch |\\n| 0.1 | 0.04 | W-F Stretch |\\n| 0.1 | 0.8 | Si-H Stretch |\\n| 0.2 | 0.007 | W-F Stretch |\\n| 0.2 | 0.9 | Si-H Stretch |\\n| 0.3 | -0.01 | W-F Stretch |\\n| 0.3 | 1 | Si-H Stretch |\\n| 0.4 | -0.01 | W-F Stretch |\\n| 0.4 | 1 | Si-H Stretch |\\n| 0.5 | -0.02 | W-F Stretch |\\n| 0.5 | 1 | Si-H Stretch |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The plateau indicates a self-limiting control mechanism, a hallmark of ALD. Once the reactive Si-H surface sites are fully consumed (absorbance ~ 0) and replaced by W-F species, the reaction stops automatically, meaning extending the exposure time further yields no additional material growth.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"0.1 Torr·min (Speed): High throughput, but the Si-H signal is still ~0.3 (30% unreacted), and W-F coverage is only ~80%.\\n\\n0.6 Torr·min (Quality): Slower throughput, but Si-H is fully removed (~0.0) and W-F is saturated (1.0).\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At the crossover point, the surface is a heterogeneous mixture of approximately 50% unreacted Si-H bonds and 50% newly formed W-F species.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The saturation value greater than 1.0 indicates a high density of reactive Si–H sites, possibly including higher-order silane species (SiHₓ). Such a hydrogen-rich surface is advantageous because Si–H bonds serve as the reducing agent for WF₆. A larger population of these active sites promotes fast and uniform nucleation of tungsten in the following deposition step.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":1,"width":598,"height":580}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/12/J.W. Klaus et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"12","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":600,"height":586,"image_format":"jpeg","image_sha256":"7491f6e170ee23f7f950aefdf68964bd25d9ebbe7afc51529c5eb8f323bf8e50","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_13_fig_3.jpg","caption":"FIG. 3.  $1\\mathrm{kHz}C - V$  measurements for MOS capacitors,  $\\mathrm{Cu(100nm) / 10nm}$ $\\mathrm{Ti - Si - N}$  barrier layer (or without the barrier layer)  $\\mathrm{SiO_2}$ $100\\mathrm{nm}) / N$  -type Si. a and b are the  $C - V$  profiles for the MOS capacitor without the barrier layer before and after thermal treatment at  $600^{\\circ}\\mathrm{C}$  for  $60~\\mathrm{min}$  in  $\\mathrm{H}_{2}$ $(10\\%)/\\mathrm{Ar}(90\\%)$  respectively.c is the  $C - V$  profiles for the MOS capacitor with a  $10\\mathrm{nm}$  Ti-Si-N barrier layer after thermal treatment at  $800^{\\circ}\\mathrm{C}$  for  $60~\\mathrm{min}$  in  $\\mathrm{H}_{2}$ $(10\\%)/\\mathrm{Ar}(90\\%)$","id":"train/atomic-layer-deposition/experimental-usecase/13/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/13/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the normalized capacitance (C/Cox) against voltage for three different conditions: before thermal treatment, no barrier capacitor at 600°C, and Ti-Si-N (10 nm) capacitor at 800°C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Voltage [V] | Before thermal treatment | No barrier capacitor | Ti-Si-N capacitor |\\n| --- | --- | --- | --- |\\n| -15 | 0.21 | 1.00 | 0.19 |\\n| -10 | 0.22 | 1.00 | 0.20 |\\n| -5 | 0.35 | 0.60 | 0.25 |\\n| 0 | 1.00 | 0.90 | 1.00 |\\n| 5 | 1.00 | 1.00 | 1.00 |\\n| 10 | 1.00 | 1.00 | 1.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"For both curves, the MOS capacitor does not have the barrier layer.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"For all the curves, the normalized capacitance exhibits the same behavior at voltage values above 1V: it remains constant as the voltage increases.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately at -3V.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Cu atoms can enter the Si and form deep donor levels. These levels can act as the generation and recombination centers of electron-hole pairs.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":589,"height":563}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/13/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/13/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/13/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/13/Jae-Sik Min et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"13","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":595,"height":567,"image_format":"jpeg","image_sha256":"200bc2cfb4a66c2e0d2342296d85fbe6fc75aaa6455469102a3e99465c4c18c7","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_3.jpg","caption":"Figure 3. MPC for 400 ALD cycles of either  $\\mathrm{ZnO}$  (red diamonds) or  $\\mathrm{GeO}_y$  (blue squares). The deposition of  $\\mathrm{ZnO}$  was interrupted after 200 cycles by performing a single  $\\mathrm{GeO}_y$  cycle before depositing another 200 cycles of  $\\mathrm{ZnO}$ . Similarly, the deposition of  $\\mathrm{GeO}_y$  was interrupted after 200 cycles by performing a single  $\\mathrm{ZnO}$  cycle before depositing another 200 cycles of  $\\mathrm{GeO}_y$ .","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure plots the mass per cycle (MPC) during 400 ALD cycles for ZnO and GeOₓ films. Each deposition was interrupted after 200 cycles by inserting a single cycle of the opposite material, creating ZnO/GeOₓ/ZnO and GeOₓ/ZnO/GeOₓ sequences\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle Number | ZnO / 1 cycle GeOₓ / ZnO (ng/cm²) | GeOₓ / 1 cycle ZnO / GeOₓ (ng/cm²) |\\n|--------------|------------------------------------|--------------------------------------|\\n| 0            | 140                                | 20                                   |\\n| 50           | 145                                | 22                                   |\\n| 100          | 147                                | 23                                   |\\n| 150          | 148                                | 24                                   |\\n| 200 (Interruption) | 150            | 50                                   |\\n| 250          | 152                                | 24                                   |\\n| 300          | 153                                | 25                                   |\\n| 350          | 155                                | 26                                   |\\n| 400          | 158                                | 27                                   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The deposition of ZnO was interrupted after 200 cycles by performing a single GeOy cycle before depositing another 200 cycles of ZnO. Similarly, the deposition of GeOy was interrupted after 200 cycles by performing a single ZnO cycle before depositing another 200 cycles of GeOy. As one continuous process is interrupted, this is called an interruption experiment\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"An interruption experiment was performed to further understand the dynamics of the inhibition seen when mixing ZnO and GeOy subcycles into ZGO supercycles\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. Only a minor transient in MPC was seen after the ZnO process was interrupted with a single GeOy cycle.\\n 2. In contrast, when the GeOy process was interrupted by a single ZnO cycle, the MPC was lowered for more than 100 consecutive GeOy cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The implication of this for the ZGO supercycle processes would be that the more the GeOy cycles used in a row, the more the Ge−OH on the surface and the more the reaction sites end up being poisoned during the DEZ exposure. As a result, the GPC would decrease as the amount of GeOy cycles in the ZGO supercycle increases.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":501,"height":349}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_3.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 scatter plot","caption_source":"content.json:image_caption"},"width":503,"height":350,"image_format":"jpeg","image_sha256":"08936291d7c61e451c40bcffe53e87a2b96ae4076204cb6e560cd3ad3516c0ae","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_6.jpg","caption":"Figure 6. Optical absorption as a function of the cation ratio,  $x_{i}$  of a TGO and  $(\\hat{1})$  ZGO films deposited on fused silica.","id":"train/atomic-layer-deposition/experimental-usecase/25/figure_6","sample_id":"atomic-layer-deposition/experimental-usecase/25/figure_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows tauc plot for TGO at various GeOx concentrations.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows tauc plot for ZGO at various GeOx concentrations.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| E (eV) | x = 0% | x = 6.5% | x = 7.7% | x = 10% | x = 16% | x = 30% |\\n|--------|--------|----------|----------|---------|---------|---------|\\n| 3.0    | 1000   | 800      | 700      | 600     | 500     | 200     |\\n| 3.5    | 2500   | 2200     | 2100     | 1900    | 1700    | 800     |\\n| 4.0    | 4000   | 3700     | 3500     | 3300    | 3000    | 1500    |\\n| 4.5    | 5500   | 5200     | 5000     | 4700    | 4300    | 2200    |\\n| 5.0    | 6000   | 5800     | 5600     | 5300    | 4900    | 2800    |\"},{\"panel_id\":\"b\",\"text\":\"| E (eV) | x = 0% | x = 3.2% | x = 12% | x = 19% | x = 32% | x = 37% |\\n|--------|--------|----------|---------|---------|---------|---------|\\n| 3.0    | 800    | 600      | 400     | 300     | 200     | 100     |\\n| 3.5    | 2200   | 1900     | 1600    | 1400    | 1000    | 700     |\\n| 4.0    | 3700   | 3400     | 3000    | 2700    | 2200    | 1800    |\\n| 4.5    | 5200   | 4900     | 4500    | 4100    | 3500    | 3000    |\\n| 5.0    | 6000   | 5800     | 5400    | 5000    | 4400    | 3800    |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Eg increased as the ZGO cation ratio increased. The ZnO was found to be able to incorporate small amounts of Ge into the wurtzite lattice and thereby\\nincrease both conductivity and Eg, while larger amounts led to a high resistivity and worse crystal quality\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. For maximum doping of TGO 0.30, Eg = 3.5 eV\\n2. For ZGO, ZGO 0.32 gives 4.11 eV and \\n3. ZGO 0.32, Eg = 4.49 eV\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The bandgap Eg slowly increased as the cation ratio increased\"}]}]","bbox":[{"panel_id":"b","x":0,"y":301,"width":448,"height":348},{"panel_id":"a","x":0,"y":1,"width":446,"height":302}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/25/images/figure_6.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 scatter plot","caption_source":"content.json:image_caption"},"width":450,"height":653,"image_format":"jpeg","image_sha256":"87f4deccaaffc9687ffc6b7a61ca01ec4342d2049b9cc1c97604139ee6e614e7","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_2.jpg","caption":"Figure 2. Growth rate for ALD ITO versus percentage of  $\\mathrm{SnO_2}$  cycles determined using VASE  $(\\bullet)$  and XRF  $(\\nabla)$ . Solid line guides the eye, and dashed line shows expected growth rates calculated using a rule-of-mixtures formula.","id":"train/atomic-layer-deposition/experimental-usecase/26/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/26/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the variation in growth rate (in Å/cycle) as a function of SnO₂ cycle percentage. Measurements from ellipsometry and XRF are plotted, along with a dashed line representing the rule of mixtures. Both techniques show a decreasing trend in growth rate with increasing SnO₂ cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| % SnO₂ Cycles | Ellipsometry (Å/cycle) | XRF (Å/cycle) |\\n|---------------|-------------------------|----------------|\\n| 0             | 1.75                    | 1.60           |\\n| 1             | 1.45                    | 1.40           |\\n| 2             | 1.35                    | 1.30           |\\n| 3             | 1.25                    | 1.20           |\\n| 4             | 1.15                    | 1.10           |\\n| 5             | 1.05                    | 1.00           |\\n| 10            | 0.95                    | 0.90           |\\n| 15            | 0.85                    | 0.80           |\\n| 20            | 0.75                    | 0.70           |\\n| 25            | 0.65                    | 0.60           |\\n| 30            | 0.55                    | 0.50           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The growth rate decreases steadily as the % SnO₂ cycles increases, for both ellipsometry and XRF measurements, with ellipsometry consistently reading higher.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"No, the trend is nonlinear, especially in the low % SnO₂ region.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Ellipsometry, X-ray fluorescence (XRF),  Rule of Mixtures model\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It represents a theoretical growth rate expected from a weighted average of individual components.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":583,"height":586}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_2.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 scatter plot","caption_source":"content.json:image_caption"},"width":583,"height":586,"image_format":"jpeg","image_sha256":"f1b7ac677e57f609ec67bffcd23ddf8f8a17adcc301100bc1014c17faeaa2bcb","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_4.jpg","caption":"Figure 4. Growth rate versus ALD cycles measured using in situ QCM during ITO growth using  $10\\%$ $\\mathrm{SnO_2}$  cycles.","id":"train/atomic-layer-deposition/experimental-usecase/26/figure_4","sample_id":"atomic-layer-deposition/experimental-usecase/26/figure_4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the growth rate (Å/cycle) of SnO₂ and In₂O₃ as a function of ALD cycles. The SnO₂ growth rate exhibits periodic dips, while the In₂O₃ growth remains relatively steady with minor fluctuations.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD Cycles | SnO₂ | In₂O₃ |\\n|---|---|---|\\n| 0 | 1.15 | 1.15 |\\n| 5 | 1.20 | 1.20 |\\n| 10 | 1.10 | 1.10 |\\n| 15 | 1.25 | 1.25 |\\n| 20 | 1.30 | 1.30 |\\n| 25 | 1.20 | 1.20 |\\n| 30 | 1.10 | 1.10 |\\n| 35 | 1.25 | 1.25 |\\n| 40 | 1.30 | 1.30 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SnO₂ shows sharp dips at regular intervals, suggesting a periodic change, while In₂O₃ growth rate remains relatively stable.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SnO₂\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"SnO₂ growth may be affected by precursor pulsing or surface saturation ,In₂O₃ growth is more consistent, Useful for tuning multilayer structures.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":575,"height":583}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/26/images/figure_4.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 scatter plot","caption_source":"content.json:image_caption"},"width":575,"height":583,"image_format":"jpeg","image_sha256":"03652fa56847d52bdceb2812be97b195c7205678cb117db06e9fb3555b1905dd","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_27_fig_2.jpg","caption":"Fig. 2. AES depth profile (a) and impurity contents (b) of TiN films deposited by remote PEALD method using TDMAT precursor and nitrogen plasma.","id":"train/atomic-layer-deposition/experimental-usecase/27/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/27/fig_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple line chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The atomic concentration of various elements throughout a film, with carbon and oxygen decreasing deeper in the film.\"},{\"panel_id\":\"b\",\"text\":\"Atomic concentrations of carbon and oxygen as function of deposition temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputter Time (min) | C | O | O of SiO₂ | Ti | N | Si |\\n|---|---|---|---|---|---|---|\\n| 0 | 4 | 6 | 0 | 39 | 53 | 0 |\\n| 5 | 4 | 5 | 0 | 37 | 53 | 0 |\\n| 10 | 4 | 5 | 0 | 36 | 50 | 5 |\\n| 15 | 4 | 3 | 0 | 35 | 45 | 10 |\\n| 20 | 4 | 0 | 10 | 30 | 40 | 20 |\\n| 25 | 3 | 0 | 25 | 25 | 30 | 25 |\\n| 30 | 3 | 0 | 35 | 20 | 20 | 30 |\\n| 35 | 0 | 0 | 45 | 15 | 15 | 30 |\\n| 40 | 0 | 0 | 50 | 10 | 10 | 30 |\\n| 45 | 0 | 0 | 55 | 5 | 5 | 30 |\\n| 50 | 0 | 0 | 60 | 0 | 0 | 35 |\"},{\"panel_id\":\"b\",\"text\":\"| Deposition Temperature (°C) | Carbon | Oxygen |\\n|---|---|---|\\n| 150 | 11 | 10 |\\n| 200 | 7 | 6 |\\n| 250 | 3 | 3 |\\n| 300 | 4 | 3 |\\n| 350 | 5 | 3 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The lowest impurities are achieved at 250 degrees C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A common source of carbon impurities is the precursor ligands. Oxygen impurities can arise due to background water in the reactor which is incorporated by the plasma\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This is due to the penetration depth of the measurement method which is likely to be around 5-10 nm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The TiN was deposited on silicon oxide, this is shown by the Si and O from SiO2 signal.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":0,"width":538,"height":422},{"panel_id":"b","x":0,"y":498,"width":552,"height":436}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/27/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/27/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/27/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/27/Ju Youn KIM et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"27","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":553,"height":970,"image_format":"jpeg","image_sha256":"ed7dd6b7dabb9b2e15b5499dea2b3dd7e7f67e296a8206d3036e03d4f954cedd","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_2.jpg","caption":"Figure 2. Growth rates of thermal and PE-ALD  $\\mathrm{Ta_2O_5}$  as a function of water and oxygen plasma exposure time  $(t_{\\mathrm{r}})$  on  $\\mathrm{Si(001)}$  substrate at  $T_{\\mathrm{s}} = 250^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/32/figure_2","sample_id":"atomic-layer-deposition/experimental-usecase/32/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure compares growth rates of thermal ALD (using H₂O) and PE-ALD (using O₂ plasma) for Ta₂O₅ as a function of reactant exposure time at 250°C. Both methods exhibit saturation behavior for t_r ≥ 1 s, with PE-ALD achieving a higher saturated growth rate (~1.2 Å/cycle) compared to thermal ALD (~0.85 Å/cycle).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| t_r, s | Growth rate thermal ALD (Å/cycle) | Growth rate PE-ALD (Å/cycle) |\\n|---|---|---|\\n| 0 | 0.05 | 0.08 |\\n| 0.25 | 0.30 | 0.40 |\\n| 0.5 | 0.75 | 0.75 |\\n| 1 | 0.82 | 1.10 |\\n| 2 | 0.83 | 1.12 |\\n| 3 | 0.84 | 1.14 |\\n| 5 | 0.85 | 1.14 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"PE-ALD achieves approximately 1.2 Å/cycle while thermal ALD reaches approximately 0.85 Å/cycle, making PE-ALD about 40% faster.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Thermal ALD uses water vapor, while PE-ALD uses oxygen plasma.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The higher growth rate in PE-ALD is attributed to the greater reactivity of the oxygen plasma species, which include atomic oxygen and activated oxygen molecules. These reactive species can more efficiently oxidize the adsorbed PDMAT precursor and potentially create more reactive surface sites for subsequent adsorption cycles compared to water vapor in thermal ALD.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. Both methods show saturation of growth rate for reactant exposure times greater than approximately 1 second.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":670,"height":503}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/images/figure_2.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 scatter plot","caption_source":"content.json:image_caption"},"width":670,"height":503,"image_format":"jpeg","image_sha256":"e050b126404a37a5b13e1dec33ffe9118fd42889b19d03323f98e04fef926d10","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_3.jpg","caption":"Figure 3. Growth rates of thermal and PE-ALD  $\\mathrm{Ta_2O_5}$  as a function of growth temperature  $T_{\\mathrm{s}}$ .","id":"train/atomic-layer-deposition/experimental-usecase/32/figure_3","sample_id":"atomic-layer-deposition/experimental-usecase/32/figure_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the temperature dependence of Ta₂O₅ growth rate for thermal ALD and PE-ALD. Both methods exhibit three regimes: elevated growth rates at low temperatures, a stable ALD window in the middle, and increasing growth rates at high temperatures due to precursor decomposition. The ALD window spans 200 to 250°C for thermal ALD and 150 to 250°C for PE-ALD, demonstrating that PE-ALD enables ALD growth at approximately 50°C lower temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| T_s, °C | Growth rate thermal ALD (Å/cycle) | Growth rate PE-ALD (Å/cycle) |\\n|---|---|---|\\n| 100 | - | 1.40 |\\n| 125 | 1.32 | - |\\n| 150 | 1.20 | 1.20 |\\n| 175 | - | 1.18 |\\n| 200 | 0.90 | 1.17 |\\n| 250 | 0.85 | 1.15 |\\n| 275 | - | 1.25 |\\n| 300 | 1.00 | - |\\n| 500 | 1.58 | 1.70 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The ALD window is 200 to 250°C for thermal ALD and 150 to 250°C for PE-ALD.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At high temperatures, the PDMAT precursor undergoes thermal decomposition before the self-limiting surface reaction can complete. This leads to continuous CVD-like deposition rather than true ALD, resulting in higher apparent growth rates. The loss of self-limitation compromises the precise thickness control that is characteristic of ALD.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"PE-ALD uses oxygen plasma containing highly reactive species such as atomic oxygen and activated oxygen molecules. These radicals have sufficient reactivity to oxidize the adsorbed precursor even at lower temperatures where water vapor in thermal ALD would be insufficiently reactive. In fact, the authors note that Ta₂O₅ could not be deposited without plasma at temperatures below 250°C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 1.4 Å/cycle.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":684,"height":533}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/32/images/figure_3.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 scatter plot","caption_source":"content.json:image_caption"},"width":684,"height":533,"image_format":"jpeg","image_sha256":"7aad86506f50da1c24260d4860e54d192547c2a1d5f2e2cb91aa41faac97e275","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_35_figure_4.jpg","caption":"Figure 4. Composition of GST films as a function of (a)  $\\mathrm{GeCl}_2\\cdot \\mathrm{C}_4\\mathrm{H}_8\\mathrm{O}_2$  pulse length and (b)  $\\mathrm{GeTe / (GeTe + Sb_2Te_3)}$  cycling ratio.  $\\mathrm{SbCl}_3$  and  $(\\mathrm{Et}_3\\mathrm{Si})_2\\mathrm{Te}$  pulse lengths were always  $1.0~\\mathrm{s}$  and purges were  $2.0~\\mathrm{s}$ .","id":"train/atomic-layer-deposition/experimental-usecase/35/figure_4","sample_id":"atomic-layer-deposition/experimental-usecase/35/figure_4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the composition of a Ge, Sb, and Te film at different pulse lengths of GeCl₄-dioxane.\"},{\"panel_id\":\"b\",\"text\":\"The scatter plot displays the composition of a Ge, Sb, and Te film at various cycles of GeTe/(GeTe+Sb₂Te₃).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| GeCl₄-dioxane pulse length (s) | Composition (at.%)  Ge|Composition (at.%)  Sb|Composition (at.%)  Te|\\n|---|---|---|---|\\n| 1 | 15 |35|50|\\n| 2 | 25 |25|50|\\n| 3 | 30 |20|50|\\n| 4 | 30 |20|50|\\n| 6 | 30 |20|50|\"},{\"panel_id\":\"b\",\"text\":\"| GeTe/(GeTe+Sb₂Te₃) cycling | Composition (at.%)  Ge|Composition (at.%)  Sb|Composition (at.%)  Te|\\n|---|---|---|---|\\n| 0.25 | 20 |30|50|\\n| 0.33 | 22 |28|50|\\n|0.36|28|22|50|\\n| 0.5 | 30 |20|50|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"4 s\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Sb is replaced with Ge, as for increasing Ge exposure, the Ge percentage increases and the Sb percentage decreases. The Te percentage remains constant.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"three\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The method to show the cycle ratio can mean 1 to 3 but also 50 to 150, in both cases the composition will be different due to the probing depth of the measurement technique with which the composition is measured.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":5,"width":312,"height":265},{"panel_id":"b","x":332,"y":10,"width":303,"height":259}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/35/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/35/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/35/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/35/Viljami Pore et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"35","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":636,"height":270,"image_format":"jpeg","image_sha256":"79b3569ff9c435400c8cf0aa89660172dd90e2d8b66944edb9f5b2bb894a5bbd","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_fig9.jpg","caption":"Fig.9  $I_{DS} - V_{DS}$  characteristics of the conventional HEMT and the  $\\mathrm{Ga_2O_3}$  MOS-HEMT upon varying the value of  $V_{G}$  from  $-6$  to  $+2\\mathrm{V}$  at a step of  $+1\\mathrm{V}$","id":"train/atomic-layer-deposition/experimental-usecase/36/fig9","sample_id":"atomic-layer-deposition/experimental-usecase/36/fig9","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the drain current (I<sub>DSS</sub>) versus drain-source voltage (V<sub>DS</sub>) for Ni/Au-HEMT and Ga<sub>2</sub>O<sub>3</sub> MOS-HEMT devices.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| V<sub>DS</sub> (V) | Ni/Au-HEMT | Ga2O3 MOS-HEMT |\\n| --- | --- | --- |\\n| 0 | 0 | 0 |\\n| 5 | 800 | 1000 |\\n| 10 | 810 | 1010 |\\n| 15 | 805 | 1000 |\\n| 20 | 800 | 900 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Vg varies from -6 V to +2 V in steps of 1 V.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"In both cases, the current increases to 5 V and then remains almost constant.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen in the figure, Ga2O3 MOS-HEMT sample reaches higher current values.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"I_DSS of 608 mA/mm for Ni/Au-HEMT and I_DSS of 720 mA/mm for Ga2O3 MOS-HEMT.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":8,"width":609,"height":384}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig9.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 scatter plot","caption_source":"content.json:image_caption"},"width":611,"height":392,"image_format":"jpeg","image_sha256":"a9ea0f9e32da3be091e8c5fad45d866598310e53e569870b598350443bb4ef8c","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_11.jpg","caption":"Fig. 11 a Gate leakage current and b three-terminal off-state breakdown characteristics of the conventional HEMT and the  $\\mathrm{Ga}_{2}\\mathrm{O}_{3}$  MOS-HEMT","id":"train/atomic-layer-deposition/experimental-usecase/36/fig_11","sample_id":"atomic-layer-deposition/experimental-usecase/36/fig_11","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The graph shows the gate-drain current (I<sub>GD</sub>) versus gate-source voltage (V<sub>GD</sub>) for Ni/Au-HEMT and Ga<sub>2</sub>O<sub>3</sub> MOS-HEMT.\"},{\"panel_id\":\"b\",\"text\":\"The graph shows the drain-source current (I<sub>DS</sub>) versus drain-source voltage (V<sub>DS</sub>) for Ni/Au-HEMT and Ga<sub>2</sub>O<sub>3</sub> MOS-HEMT. Voltage values of 126 V and 170 V are highlighted.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| V<sub>GD</sub> (V) | Ni/Au-HEMT | Ga2O3 MOS-HEMT |\\n| --- | --- | --- |\\n| -10 | 10^-3 | 10^-5 |\\n| -8 | 10^-3 | 10^-5 |\\n| -6 |  10^-3  | 10^-5 |\\n| -4 |  10^-3  | 10^-5 |\\n| -2 |  10^-4  | 10^-6 |\\n| 0 |  10^-6  | 10^-9 |\\n| 2 |  10^0 | 10^0 |\"},{\"panel_id\":\"b\",\"text\":\"| V<sub>GD</sub> (V) | Ni/Au-HEMT | Ga2O3 MOS-HEMT |\\n| --- | --- | --- |\\n| 0 | 0 | 0 |\\n| 30 | 0 | 0 |\\n| 60 | 0 | 0 |\\n| 90 | 0 | 0 |\\n| 120 | 0.15 | 0.01 |\\n| 150 | - | 0.05 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Subfigure a shows the gate-drain current versus voltage gate-drain for both samples.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"For the Ni/Au-HEMT, the current remains constant but begins to increase after 90 V, with a significant increase at 126 V. Meanwhile, the Ga2O3 MOS-HEMT sample remains constant until approximately 150 V and then increases significantly at 170 V.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"170 V.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The ALD Ga2O3 film can suppressed the gate leakage current density and increased the drain-state voltage.\"}]}]","bbox":[{"panel_id":"a","x":14,"y":25,"width":462,"height":400},{"panel_id":"b","x":498,"y":18,"width":466,"height":409}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_11.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 scatter plot","caption_source":"content.json:image_caption"},"width":964,"height":427,"image_format":"jpeg","image_sha256":"555ed7cd0cf6e4f4dfbb6cfdc848e54605f047da36c47e120b7c4898d2e20e1f","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_8.jpg","caption":"Fig. 8 The capacitance-voltage (C-V) characteristics of the conventional HEMT and the  $\\mathrm{Ga}_{2}\\mathrm{O}_{3}$  MOS-HEMT","id":"train/atomic-layer-deposition/experimental-usecase/36/fig_8","sample_id":"atomic-layer-deposition/experimental-usecase/36/fig_8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the capacitance-voltage characteristics of the conventional HEMT and the Ga2O3 MOS-HEMT.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gate Voltage (V) | Ga2O3 MOS-HEMT | conventional HEMT |\\n|-----------------|-------|-------|\\n| -8             | 0   | 0   |\\n| -7             | 0   | 0   |\\n| -6             | 0   | 0   |\\n| -5             | 0   | 0   |\\n| -4             | 15  | 0   |\\n| -3             | 15  | 27   |\\n| -2             | 15  | 27   |\\n| -1             | 15  | 27   |\\n|  0             | 15  | 27   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The capacitance-voltage characteristics were measured at 1 MHz.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"For both the Ga2O3 MOS-HEMT and the conventional HEMT, the capacitance increases until it reaches plateaus of 15 and 7, respectively.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The capacitance value is o when the gate voltage varies from -8 to -4 V.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The Ga2O3 HEMT allows for the lowest possible capacitance values.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":617,"height":444}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/36/images/fig_8.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 scatter plot","caption_source":"content.json:image_caption"},"width":620,"height":450,"image_format":"jpeg","image_sha256":"32787a11603f9a0bacb2f76fd403b8770692af3b33c2e00c2bd2067bdc2226cb","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_40_fig_7.jpg","caption":"FIG. 7. Change of sheet resistance on  $\\mathrm{Cu / PEALD}$  TiN  $10 \\mathrm{nm} / \\mathrm{Si}$  and  $\\mathrm{Cu / PEALD}$  Ti-Si-N  $10 \\mathrm{nm} / \\mathrm{Si}$  as a function of annealing temperature.","id":"train/atomic-layer-deposition/experimental-usecase/40/fig_7","sample_id":"atomic-layer-deposition/experimental-usecase/40/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Multiple line chart showing sheet resistance evolution of Cu/PEALD TiN and Cu/PEALD Ti–Si–N (10 nm) films on Si as a function of annealing temperature. TiN remains conductive up to ~550 °C but exhibits a sharp resistance increase at 600 °C, indicating thermal degradation. In contrast, Ti–Si–N maintains near-zero resistance up to 600 °C and shows a delayed, gradual increase at higher temperatures, demonstrating enhanced thermal stability due to silicon incorporation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Annealing temperature (°C) | Sheet resistance: TiN  | Sheet resistance: Ti–Si–N @ Ar  |\\n|----------------------------|-----------------------------|--------------------------------------|\\n| 450                        | ~0                          | ~0                                   |\\n| 500                        | ~0                          | ~0                                   |\\n| 550                        | ~5                          | ~0                                   |\\n| 600                        | ~62                         | ~0                                   |\\n| 650                        | –                           | ~7                                   |\\n| 700                        | –                           | ~30                               |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"-Nitrogen out-diffusion from TiN\\n, -Grain growth and grain boundary formation\\n, -Copper diffusion into the barrier layer\\n, -Loss of film continuity during annealing\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TiN shows a sudden and dramatic rise in sheet resistance at 600 °C, indicating rapid loss of electrical continuity. In contrast, Ti–Si–N maintains low resistance through 600 °C and degrades only gradually at higher temperatures, reflecting superior thermal robustness.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Silicon incorporation disrupts long-range crystallinity and suppresses grain boundary formation, which are common diffusion pathways. This structural modification stabilizes the film against high-temperature degradation, delaying resistance increase compared to pure TiN.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Because Ti–Si–N maintains low sheet resistance at higher annealing temperatures, indicating improved thermal and diffusion-barrier stability.\"}]}]","bbox":[{"panel_id":"a","x":6,"y":2,"width":498,"height":396}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/40/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/40/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/40/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/40/Park et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":506,"height":400,"image_format":"jpeg","image_sha256":"5b52f5e9ac3fff195ed759c54063a29c615a481d668c6178c7b0925c8a9c77c7","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_41_fig_1.jpg","caption":"FIG. 1. Effective lifetime as a function of the excess carrier density for low resistivity  $p$ -type  $(300\\mu \\mathrm{m}$ ,  $(111)$ ,  $2.0\\Omega \\mathrm{cm}$  float zone  $c$ -Si substrate passivated with 30 and  $15\\mathrm{nm}$  thick  $\\mathrm{Al}_2\\mathrm{O}_3$  films.","id":"train/atomic-layer-deposition/experimental-usecase/41/fig_1","sample_id":"atomic-layer-deposition/experimental-usecase/41/fig_1","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between injection level and effective lifetime for two different thickness passivation layers (30 nm and 15 nm). The data points are clustered around a curve, indicating a non-linear relationship.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Injection level Δn (cm⁻³) | Effective lifetime (s), 30 nm|Effective lifetime (s), 15 nm|\\n|---|---|---|\\n| 10¹² | 2*10⁻⁴ |-|\\n| 10¹³ | 4*10⁻⁴ |10⁻³|\\n| 10¹⁴ | 6*10⁻⁴ |10⁻³ |\\n| 10¹⁵ | 9*10⁻⁴ |10⁻³ |\\n| 10¹⁶ | 10⁻³ |10⁻³ |\\n| 10¹⁷ | 10⁻⁴ |10⁻⁴ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The 15 nm thick layer, this thickness seems to have the best effect over a broader range of injection level.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.001 s.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Scattered data does not show a straight line. In addition, both axes are logarithmic.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.0001 s\"}]}]","bbox":[{"panel_id":"a","x":1,"y":2,"width":538,"height":411}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/images/fig_1.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/Hoex et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"41","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":539,"height":414,"image_format":"jpeg","image_sha256":"26b5af2e6f1c0da0701cc917e24f476767e25dda7919bca670465e674dfc4e71","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_41_fig_2.jpg","caption":"FIG. 2. Effective lifetime as a function of the excess carrier density for low resistivity  $n$ -type  $(275\\mu \\mathrm{m}$ ,  $(100)$ ,  $1.9\\Omega \\mathrm{cm}$  float zone  $c$ -Si substrate passivated with 30, 15, and  $7\\mathrm{nm}$  thick  $\\mathrm{Al}_2\\mathrm{O}_3$  films.","id":"train/atomic-layer-deposition/experimental-usecase/41/fig_2","sample_id":"atomic-layer-deposition/experimental-usecase/41/fig_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the effective lifetime of a system as a function of injection level Δn, with data points represented by different symbols corresponding to different thickness passivation layers (26, 13, and 7 nm).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Injection level Δn (cm⁻³) | Effective lifetime (s), 26 nm|Effective lifetime (s), 13 nm|Effective lifetime (s), 7 nm|\\n|---|---|---|---|\\n| 10¹² | 10⁻³ |-|-|\\n| 10¹³ | 2*10⁻³ |10⁻³|10⁻³|\\n| 10¹⁴ | 5*10⁻³ |2*10⁻³ |2*10⁻³ |\\n| 10¹⁵ | 7*10⁻³ |3*10⁻³ |3*10⁻³ |\\n| 10¹⁶ | 10⁻³ |8*10⁻⁴ |8*10⁻⁴|\\n| 10¹⁷ | - |-|10⁻⁴ |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The difference between the 7 and 13 nm thick passivating layers is negligable, whereas the 26 nm layer does show an increase in passivation.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The overal trends look similar, important is that the maximum effective lifetime seems to shift to higher injection levels with thicker passivation layers.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.008 s with a passivation layer of 26 nm.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10^15 per cubic centimeter.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":540,"height":426}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/41/Hoex et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"41","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":544,"height":433,"image_format":"jpeg","image_sha256":"6ed5a138a6cd736d10b12caa848929906f83974e8f96c58bd84e1f3870f82fda","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_fig4.jpg","caption":"Fig.4 Resistivity.  $\\rho$  a),carrier concentration,  $n$  b),mobility,  $\\mu$  c and doping efficiency  $\\eta$  (d) of the ZnO:X  $\\mathbb{X} = \\mathbb{A}\\mathbb{I}$  B) films as a function of the doping fraction (DF) measured by XPS. TIB was used as a B source while DMAI and TMA were used as A sources. All films had a thickness of  $45\\pm$ $5\\mathrm{nm}$  and were deposited on glass substrates at  $150^{\\circ}C$  The lines serve as guides to the eye. The error bars for  $\\rho$  are plotted in the graph (some of them lying within the data points) and the estimated relative statistical uncertainties for n,  $\\mu$  and  $\\eta$  are  $3\\%$ $5\\%$  and  $20\\%$  respectively.","id":"train/atomic-layer-deposition/experimental-usecase/50/fig4","sample_id":"atomic-layer-deposition/experimental-usecase/50/fig4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"multiple scatter plot"},{"panel_id":"d","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The resistivity of TIB, DMAI, and TMA decreases as the doping fraction increases and then slightly increases\"},{\"panel_id\":\"b\",\"text\":\"The carrier concentration of TIB, DMAI, and TMA initially increases and then decreases as the doping fraction increases.\"},{\"panel_id\":\"c\",\"text\":\"The Mobility also increases initially and then decreases\"},{\"panel_id\":\"d\",\"text\":\"Doping efficiency increases at the beginning and then decreases, but for TMA it remains constant.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Doping fraction, DF | TIB (mΩ·cm) | DMAI (mΩ·cm) | TMA (mΩ·cm) |\\n|---|---|---|---|\\n| 0.00 |  0          |   0           | 70          |\\n| 0.020   | 6           | 10           | 20          |\\n| 0.040   | 8           |  8           | 15          |\\n| 0.060   | 10         | 10          | 12          |\\n| 0.080   | 28          | 18           | -         |\\n| 0.100   | -             | 25           | 22          |\"},{\"panel_id\":\"b\",\"text\":\"| Doping fraction, DF | TIB (10^20 cm^-3) | DMAI (10^20 cm^-3) | TMA (10^20 cm^-3) |\\n|---|---|---|---|\\n| 0.00 | 0.0 | 0.0 | 0.0 |\\n| 0.01 | 1 | 1 | - |\\n| 0.02 | 1.5 | 1.5 | - |\\n| 0.03 |3.0 | 2.5 | 0.5 |\\n| 0.04 | -    | 2.5 | 1.0 |\\n| 0.05 | 2.75 | - | -   |\\n| 0.06 | 2.0 | 2.2 | 1.5 |\\n| 0.08 | 1.5 | - | - |\\n| 0.09 | 1.0 | 2.2 | - |\\n| 0.10 |- 000 | 2.2 | 1.5 |\"},{\"panel_id\":\"c\",\"text\":\"| Doping fraction, DF | TIB (cm²/V·s^-1) | DMAI (cm²/V·s^-1) | TMA (cm²/V·s^-1) |\\n|---|---|---|---|\\n| 0.00 | 7.0 | 7.0 | 7.0 |\\n| 0.01 | 8.0 | 8.0 | 8.0 |\\n| 0.02 | 9.0 | 9.0 | 9.0 |\\n| 0.03 | 10.0 | 10.0 | 10.0 |\\n| 0.04 | 11.0 | 11.0 | 11.0 |\\n| 0.05 | 12.0 | 12.0 | 12.0 |\\n| 0.06 | 10.0 | 10.0 | 10.0 |\\n| 0.07 | 8.0 | 8.0 | 8.0 |\\n| 0.08 | 6.0 | 6.0 | 6.0 |\\n| 0.09 | 4.0 | 4.0 | 4.0 |\\n| 0.10 | 2.0 | 2.0 | 2.0 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The addition of a small amount of B sharply reduced the resistivity of the films to a minimum of ~3.5 mOhm cm for a doping fraction (DF) of 0.016. The addition of Al also reduced the resistivity of the films to a minimum in both cases. Doping with DMAI resulted in ZnO:Al films with the same minimum resistivity than ZnO:B films at the same DF of 0.016, while doping with TMA resulted in ZnO:Al films with a higher minimum resistivity of 8 mO cm for a higher DF of 0.040. The resistivity of ZnO:Al films increased in both cases after achieving the minimum resistivity value, but this increase was more gradual than observed for the ZnO:B films.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The overall lower mobility values obtained for the three doping materials and the trend of mobility decreasing with increasing doping fraction over the full range studied can most likely be attributed to grain boundary scattering. Furthermore, it is rather striking that m increases when going\\nfrom undoped ZnO to low B-doping concentrations, as one would expect more ionized impurity scattering. Yet a similar  trend in mobility was observed for ZnO:B films prepared by the\\nsol–gel technique.\"}]},{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"TIB and DMAI, showed the highest doping efficiency at low doping levels, i.e., 30–40% for a range of DF of 0.016–0.018 that resulted in the lowest resistivities for the ZnO:B and ZnO:Al (doped with DMAI) films. Both alternative precursors outperformed TMA in terms of doping efficiency, as a maximum doping efficiency of 6% was achieved using this conventional precursor.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The increase of the carrier concentration for low doping fractions in the three cases can be assigned to the effective doping of B and Al, respectively (i.e., substitution of Zn2+ by B3+ or Al3+). The sharp decrease of the carrier concentration for TIB and the local saturation for DMAI and TMA might occur as a result of dopant occupying interstitial sites, dopant clustering, formation of oxides, or formation of metastable phases, making dopant inactive, which is possibly related to reaching the solubility limit of the B and Al into the ZnO lattice.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":579,"height":317},{"panel_id":"b","x":0,"y":319,"width":577,"height":341},{"panel_id":"c","x":4,"y":662,"width":574,"height":320},{"panel_id":"d","x":4,"y":991,"width":576,"height":370}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/50/images/fig4.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 scatter plot","caption_source":"content.json:image_caption"},"width":581,"height":1370,"image_format":"jpeg","image_sha256":"a724d3409c392cec0b7d8bf5b4c4734ca01c0651d6f41521e5cf15c27aa00496","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_64_fig_5.jpg","caption":"FIG. 5. Growth rate of the ruthenium films grown from  $\\mathrm{RuCp}_2$  on iridium as a function of the deposition temperature, the film thicknesses being measured by EDX and XRR. The  $\\mathrm{RuCp}_2$  and oxygen pulse times were 0.5 and  $0.2\\mathrm{s}$ , respectively, and the purge period between the pulses was  $0.5\\mathrm{s}$ .","id":"train/atomic-layer-deposition/experimental-usecase/64/fig_5","sample_id":"atomic-layer-deposition/experimental-usecase/64/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the growth rate of ruthenium films deposited from RuCp2 on an iridium substrate as a function of deposition temperature. Film thicknesses were measured by EDX and XRR, and the growth rate increases with temperature, ranging from about 0.12 Å/cycle at 225 °C to 0.35 Å/cycle at 275 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition temperature (°C) | Growth rate (Å cycle⁻¹) |\\n|---|---|\\n| 225 | 0.12 |\\n| 230 | 0.15 |\\n| 250 | 0.25 |\\n| 275 | 0.35 |\\n| 280 | 0.33 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increasing the deposition temperature generally improves the growth rate and can enhance film uniformity by promoting consistent precursor adsorption and reaction across the surface. However, excessively high temperatures might risk precursor decomposition or unwanted side reactions, so optimization is required to balance growth rate and film quality.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, the growth rate on iridium is higher than on Al2O3 at similar deposition temperatures. For example, at 275 °C, the growth rate on iridium is 0.35 Å/cycle, while on Al2O3 it is 0.23 Å/cycle.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the deposition temperature increases from 225 °C to 275 °C, the growth rate of ruthenium films increases almost linearly. At lower temperatures, the growth rate is slower (around 0.12 Å/cycle) and reaches approximately 0.35 Å/cycle at higher temperatures. This indicates that higher temperatures enhance the reactivity of the RuCp2 precursor and improve film deposition efficiency.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Iridium provides a higher nucleation density, promoting faster film growth, Al₂O₃ shows a longer incubation period at the beginning, reducing the overall growth rate, Substrate type influences the minimum temperature required for film growth due to precursor reactivity differences\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":528,"height":359}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/64/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/64/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/64/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/64/Aaltonen et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"64","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":528,"height":359,"image_format":"jpeg","image_sha256":"ad14f55fb71e6d291daa4ef9c93dba4dfe7ce9607c3de5c9bd65d3c5be1840d0","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_8.jpg","caption":"Figure 8.  $P - V$  hysteresis loops for (a) Fe-; (b) Co-; and (c) Ni-doped  $\\mathrm{ZnO}$  films.","id":"train/atomic-layer-deposition/experimental-usecase/66/figure_8","sample_id":"atomic-layer-deposition/experimental-usecase/66/figure_8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"scatter plot"},{"panel_id":"c","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the polarization (P) as a function of applied voltage (V) for ZnO:Fe\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the polarization (P) as a function of applied voltage (V) for ZnO:Co\"},{\"panel_id\":\"c\",\"text\":\"The figure shows the polarization (P) as a function of applied voltage (V) for ZnO:Ni\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Applied Voltage (V) | P (μC/cm²) at 1.5 kHz | P (μC/cm²) at 3 kHz | P (μC/cm²) at 5 kHz |\\n|---|---|---|---|\\n| -3.0 | -40 | -38 | -36 |\\n| -2.0 | -30 | -28 | -26 |\\n| -1.0 | -15 | -14 | -13 |\\n| 0.0  | 0   | 0   | 0   |\\n| 1.0  | 15  | 14  | 13  |\\n| 2.0  | 30  | 28  | 26  |\"},{\"panel_id\":\"b\",\"text\":\"| Applied Voltage (V) | P (μC/cm²) |\\n|---|---|\\n| -3.0 | -40 |\\n| -2.0 | -30|\\n| -2.0 | -30 |\\n| -1.0 | -15 |\\n| 0.0  | 0   |\\n| 1.0  | 15  |\\n| 2.0  | 30  |\"},{\"panel_id\":\"c\",\"text\":\"| Applied Voltage (V) | P (μC/cm²) |\\n|---|---|\\n| -3.0 | -40 |\\n| -2.0 | -30 |\\n| -1.0 | -15 |\\n| 0.0  | 0   |\\n| 1.0  | 15  |\\n| 2.0  | 30  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The ferroelectric P−V hysteresis loops for TM doped ZnO films are measured using a Sawyer−Tower circuit\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Fe-doped layers (Figure 8a) reveal a ferroelectric like behavior with a well-resolved P−V loop and the polarization increasing with the decreasing frequency\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The results for Ni- and Co-doped layers, though showing some P− V hysteresis, are not that convincing for the existence of ferroelectricity\"}]}]","bbox":[{"panel_id":"a","x":2,"y":3,"width":379,"height":316},{"panel_id":"b","x":380,"y":4,"width":377,"height":311},{"panel_id":"c","x":759,"y":6,"width":363,"height":309}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/66/images/figure_8.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 scatter plot","caption_source":"content.json:image_caption"},"width":1125,"height":317,"image_format":"jpeg","image_sha256":"9af67368603bedf2f92010824937427ea3b105b1936f11492060c30907c3efa3","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_69_FIG10_a.jpg","caption":"","id":"train/atomic-layer-deposition/experimental-usecase/69/FIG10_a","sample_id":"atomic-layer-deposition/experimental-usecase/69/FIG10_a","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":4,"y":4,"width":659,"height":483}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/69/images/FIG10_a.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/69/images/FIG10_a.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/69/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/69/Ruben Blomme et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"experimental-usecase","paper_id":"69","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"not_found"},"width":664,"height":491,"image_format":"jpeg","image_sha256":"40c99107efd48e5720d438ce900eb5c418e6a4d439a1521576c4d47f2d8105d5","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_95837a0f09f651369047f60c135b85eaee479b486b3df8a1df6a96de654e6120.jpg","caption":"","id":"train/atomic-layer-deposition/experimental-usecase/7/95837a0f09f651369047f60c135b85eaee479b486b3df8a1df6a96de654e6120","sample_id":"atomic-layer-deposition/experimental-usecase/7/95837a0f09f651369047f60c135b85eaee479b486b3df8a1df6a96de654e6120","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Δm₁ (open markers) is consistently higher than Δm₀ (filled markers) across 100–400 °C for all three cycle-time sequences. Δm₁ peaks around ~250–300 °C (depending on marker set) and then trends downward toward 400 °C, while Δm₀ stays comparatively flat through ~350 °C. The 5–2–2–5 s condition shows a notable Δm₀ increase at ~400 °C relative to the other sequences. Overall, varying the first-step pulse time mainly impacts Δm₁ magnitude and its temperature dependence, more than Δm₀.\"},{\"panel_id\":\"b\",\"text\":\"Two cycle-time sequences are compared, and the separation between Δm₁ (open) and Δm₀ (filled) is maintained across the full temperature range. Δm₁ rises toward a maximum near ~300 °C and then decreases at higher temperature, whereas Δm₀ remains near ~35–40 with only modest changes. Differences between the two purge-time conditions are clearer in Δm₁ than in Δm₀. The data suggest temperature sensitivity is concentrated in the Δm₁ response rather than the baseline Δm₀ increment.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Growth temperature (°C) | Δm1 (0.5-2-2-5 s, open ○) | Δm1 (2-2-2-5 s, open □) | Δm1 (5-2-2-5 s, open △) | Δm0 (0.5-2-2-5 s, filled ●) | Δm0 (2-2-2-5 s, filled ■) | Δm0 (5-2-2-5 s, filled ▲) |\\n|---|---:|---:|---:|---:|---:|---:|\\n| 100 | ~86 | ~92 | ~88 | ~44 | ~45 | ~47 |\\n| 150 | ~75 | ~84 | ~85 | ~35 | ~39 | ~40 |\\n| 200 | ~88 | ~90 | ~95 | ~33 | ~37 | ~37 |\\n| 250 | ~88 | ~105 | ~102 | ~36 | ~37 | ~37 |\\n| 300 | ~101 | ~100 | ~112 | ~37 | ~38 | ~44 |\\n| 350 | ~96 | ~101 | ~100 | ~38 | ~39 | ~39 |\\n| 400 | ~90 | ~85 | ~98 | ~45 | ~45 | ~64 |\"},{\"panel_id\":\"b\",\"text\":\"| Growth temperature (°C) | Δm1 (2-2-0.5-5 s, open ○) | Δm1 (2-2-5-5 s, open □) | Δm0 (2-2-0.5-5 s, filled ●) | Δm0 (2-2-5-5 s, filled ■) |\\n|---|---:|---:|---:|---:|\\n| 100 | ~80 | ~99 | ~41 | ~52 |\\n| 150 | ~75 | ~90 | ~36 | ~39 |\\n| 200 | ~87 | ~98 | ~34 | ~38 |\\n| 250 | ~89 | ~101 | ~35 | ~37 |\\n| 300 | ~104 | ~110 | ~36 | ~37 |\\n| 350 | ~98 | ~99 | ~38 | ~35 |\\n| 400 | ~83 | ~87 | ~40 | ~40 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The third step (purge time) was varied between 0.5 s and 5 s.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. 0.5 - 2 - 2 - 5 s\\n, 2. 2 - 2 - 2 - 5 s\\n, 3. 5 - 2 - 2 - 5 s\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It varies only weakly, staying clustered near ~35–45 arbitrary units across most of 100–350 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Δm₁ is systematically larger than Δm₀ across the entire temperature range, typically ~80–110 versus ~30–45 in arbitrary units. Δm₁ also shows a clearer temperature response, with a broad maximum around ~250–300 °C before decreasing at higher temperature. By contrast, Δm₀ is comparatively flat through ~350 °C, with only small offsets between marker sets. The main temperature sensitivity therefore sits in Δm₁ rather than Δm₀.\"}]}]","bbox":[{"panel_id":"b","x":0,"y":581,"width":470,"height":556},{"panel_id":"a","x":4,"y":1,"width":467,"height":554}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/95837a0f09f651369047f60c135b85eaee479b486b3df8a1df6a96de654e6120.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/95837a0f09f651369047f60c135b85eaee479b486b3df8a1df6a96de654e6120.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 scatter plot","caption_source":"not_found"},"width":472,"height":1139,"image_format":"jpeg","image_sha256":"501a721d35369db0fbed069278c6718ac94a8588ca1d4d07d992f80860bf5796","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_fig2.jpg","caption":"Fig.2.  $\\Delta m_0$  and  $\\Delta m_1$  as functions of growth temperature. Buffer layers were deposited at 100 and  $300^{\\circ}\\mathrm{C}$ . The  $\\mathrm{TiCl_4}$  pulse time, first purge time,  $\\mathrm{H}_2\\mathrm{O}$  pulse time, and second purge time were 2, 2, 2, and  $5\\mathrm{s}$ , respectively.","id":"train/atomic-layer-deposition/experimental-usecase/7/fig2","sample_id":"atomic-layer-deposition/experimental-usecase/7/fig2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Scatter plot comparing mass increments Δm₀ and Δm₁ vs. growth temperature for films grown on buffer layers deposited at 100 °C (circles) and 300 °C (squares). It shows that the 300 °C buffer layer generally results in higher or increasing Δm₁ values, particularly above 200 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Growth Temp (°C) | Buffer 100°C: Δm0 (●) | Buffer 100°C: Δm1 (○) | Buffer 300°C: Δm0 (■) | Buffer 300°C: Δm1 (□) |\\n|---|---|---|---|---|\\n| 100 | ~50 | ~95 | ~45 | ~92 |\\n| 150 | ~35 | ~75 | ~40 | ~85 |\\n| 200 | ~30 | ~68 | ~38 | ~92 |\\n| 250 | ~25 | ~60 | ~38 | ~105 |\\n| 350 | ~28 | ~62 | ~40 | ~100 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1.  2 s (TiCl₄)\\n, 2.  2 s (First purge)\\n, 3.  2 s (H₂O pulse)\\n, 4.  5 s (Second purge)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, the open symbols are consistently plotted higher on the y-axis than the corresponding filled symbols.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The buffer layer deposited at 300 °C  (open squares) generally results in significantly higher Δm₁ values compared to the buffer deposited at 100 °C (open circles). This difference is most notable at higher growth temperatures (above 200 °C), where the 300 °C buffer sample shows an increasing trend, reaching values over 100 units, while the 100 °C buffer sample decreases to around 60 units.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The temperature at which the buffer layer was grown (300 °C for squares vs 100 °C for circles).\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":458,"height":547}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig2.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 scatter plot","caption_source":"content.json:img_caption"},"width":461,"height":553,"image_format":"jpeg","image_sha256":"9f2d1979ee9d5f9d98c358cf3ad1817ecc2df349f6e11f4b8162d6ee0e5dcf0a","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_fig6.jpg","caption":"Fig.6.  $\\Delta m_0$ ,  $\\Delta m_1$  and  $\\Delta m_1 - \\Delta m_2^{\\prime}$  recorded at  $350^{\\circ}\\mathrm{C}$  as functions of  $\\mathrm{TiCl_4}$  pulse time. The first purge time,  $\\mathrm{H}_2\\mathrm{O}$  pulse time, and second purge time were 10, 2 and  $5\\mathrm{s}$ , respectively.","id":"train/atomic-layer-deposition/experimental-usecase/7/fig6","sample_id":"atomic-layer-deposition/experimental-usecase/7/fig6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure shows Δm₀, Δm₁, and Δm₁ − Δm₂′ as functions of TiCl₄ pulse time at 350 °C. Δm₁ and Δm₁ − Δm₂′ increase rapidly at short pulse times and saturate beyond ~2 s, consistent with self-limiting surface adsorption. In contrast, Δm₀ continues to increase with pulse time, indicating an additional non-saturating mass-gain contribution. The divergence between Δm₁ saturation and Δm₀ growth suggests the onset of parasitic or CVD-like processes at longer TiCl₄ exposures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| TiCl4 Pulse Time (s) | Δm1 (Triangles) | Δm1 - Δm2' (Squares) | Δm0 (Circles) |\\n|---|---|---|---|\\n| 0.5 | ~82 | ~69 | ~37 |\\n| 1.0 | ~84 | ~73 | ~40 |\\n| 2.0 | ~88 | ~74 | ~41 |\\n| 5.0 | ~88 | ~75 | ~43 |\\n| 10.0 | ~88 | ~80 | ~50 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"350 °C.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Yes—closer to Δm₁.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Δm₁ rises rapidly at short pulse times and then plateaus, indicating saturation of surface-limited reactions. Δm₀, however, continues to increase with pulse time, suggesting an additional mass-gain mechanism that does not self-limit. This contrast implies that extended precursor exposure contributes to non-ideal growth behavior.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Since Δm₁ saturates by ~2 s, longer TiCl₄ pulses do not improve self-limiting growth but instead promote parasitic mass gain. Optimizing pulse time near the saturation threshold minimizes cycle time while avoiding unwanted CVD-like contributions.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":4,"width":464,"height":441}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig6.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 scatter plot","caption_source":"content.json:img_caption"},"width":467,"height":447,"image_format":"jpeg","image_sha256":"3e5994ac832b2a5b794cb6b10e666cb760e9ec8a75afde12b2709073ec50d748","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_fig8.jpg","caption":"Fig.8.  $\\Delta m_0 / (\\Delta m_1 - \\Delta m_2')$  as a function of growth temperature determined at different  $\\mathrm{TiCl_4}$  pulse times. Buffer layers were grown at temperatures of 100 and  $300^{\\circ}\\mathrm{C}$ . Notation of cycle times is the same as in Fig. 3.","id":"train/atomic-layer-deposition/experimental-usecase/7/fig8","sample_id":"atomic-layer-deposition/experimental-usecase/7/fig8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Scatter plot showing the ratio Δm₀/(Δm₁−Δm₂) as a function of growth temperature (~100–400 °C) for three different cycle sequences where the first pulse time varies (0.5 s, 2 s, 5 s). The ratio generally decreases with temperature up to ~250–300 °C for the longer pulse-time conditions before increasing again at higher temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Growth Temp (°C) | 5s Pulse (Squares) | 0.5s Pulse (Open Circ) | 2s Pulse (Filled Circ) | 2s Pulse (Triangles, TB=100) |\\n|---|---|---|---|---|\\n| 100 | ~0.56 | ~0.52 | ~0.50 | - |\\n| 150 | ~0.50 | ~0.50 | ~0.50 | ~0.53 |\\n| 200 | ~0.41 | ~0.40 | ~0.42 | ~0.48 |\\n| 250 | ~0.39 | ~0.41 | ~0.39 | ~0.46 |\\n| 300 | ~0.44 | ~0.40 | ~0.43 | - |\\n| 350 | ~0.45 | ~0.44 | ~0.45 | ~0.47 |\\n| 400 | ~0.75 | ~0.58 | ~0.65 | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Δm₀/(Δm₁−Δm₂)\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, the ratio increases significantly from approximately 0.44 to 0.75 in that temperature range.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The mass ratio for the 5 s pulse time sample starts at approximately 0.56 at 100 °C and decreases steadily as the temperature rises, reaching a minimum value around 250 °C. After this point, the trend reverses. Above 300 °C, the ratio increases sharply, reaching its maximum value of ~0.75 at 400 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"100 °C and 300 °C.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":5,"width":459,"height":473}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig8.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 scatter plot","caption_source":"content.json:img_caption"},"width":464,"height":480,"image_format":"jpeg","image_sha256":"3e41555eb1be74ee3ef2faa94de88e5240d8663b7993054cc0f37807fe567fda","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_3.jpg","caption":"Fig. 3.  $\\Delta m_0$  and  $\\Delta m_1$  as functions of substrate temperature measured at (a) different  $\\mathrm{TiCl_4}$  pulse times and (b) different  $\\mathrm{H}_2\\mathrm{O}$  pulse times. Buffer layer was grown at  $300^{\\circ}\\mathrm{C}$ . Cycle times shown in the inset denote the  $\\mathrm{TiCl_4}$  pulse time, first purge time,  $\\mathrm{H}_2\\mathrm{O}$  pulse time, and second purge time, respectively. Fig. 5.  $\\Delta m_0$  as a function of (a) the first purge time and (b) the second purge time. Notation of cycle times is the same as in Fig. 3. Buffer layer was deposited at  $300^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/experimental-usecase/7/fig_3","sample_id":"atomic-layer-deposition/experimental-usecase/7/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Δm₀ decreases sharply with increasing first purge time up to ~1–2 s and then plateaus for both growth temperatures. At all purge times, Δm₀ remains higher at 150 °C than at 350 °C.\"},{\"panel_id\":\"b\",\"text\":\"A similar behavior is observed when varying the second purge time, with Δm₀ dropping rapidly at short purge times and stabilizing beyond ~2 s. The persistence of a temperature offset indicates that growth temperature influences the net mass increment even after sufficient purging.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| First purge time x (s) | Δm₀ at 150 °C (circles) | Δm₀ at 350 °C (squares) |\\n|---|---|---|\\n| 0.1 | ~56 | ~41 |\\n| 0.5 | ~44 | ~35 |\\n| 1.0 | ~41 | ~36 |\\n| 2.0 | ~41 | ~34 |\\n| 5.0 | ~39 | ~30 |\\n| 10.0 | ~40 | ~30 |\"},{\"panel_id\":\"b\",\"text\":\"| Second purge time x (s) | Δm₀ at 150 °C (circles) | Δm₀ at 350 °C (squares) |\\n|---|---|---|\\n| 0.1 | ~54 | ~43 |\\n| 0.5 | ~45 | ~38 |\\n| 1.0 | ~45 | ~34 |\\n| 2.0 | ~42 | ~34 |\\n| 5.0 | ~40 | ~35 |\\n| 10.0 | ~40 | ~35 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The similar purge-time dependence indicates that both purge steps play comparable roles in removing residual species. Once a critical purge duration is reached, surface reactions become self-limiting regardless of purge order.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The first purge step following the TiCl₄ pulse.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"b","x":1,"y":347,"width":471,"height":326},{"panel_id":"a","x":1,"y":4,"width":471,"height":323}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig_3.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 scatter plot","caption_source":"content.json:img_caption"},"width":475,"height":677,"image_format":"jpeg","image_sha256":"1872bdd68afa790590f4f4ddf2f8fc5755072f9980c346050048fd9f0e6211a5","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_4.jpg","caption":"Fig. 4.  $\\Delta m_2^{\\prime}$  as a function of growth temperature recorded at different  $\\mathrm{TiCl_4}$  pulse times. Buffer layers were deposited at  $T_{\\mathrm{B}} = 100$  and  $300^{\\circ}\\mathrm{C}$ . Notation of cycle times is the same as in Fig. 3.","id":"train/atomic-layer-deposition/experimental-usecase/7/fig_4","sample_id":"atomic-layer-deposition/experimental-usecase/7/fig_4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure shows Δm₂′ as a function of growth temperature (≈100–400 °C) for different TiCl₄ pulse-time sequences, with buffer layers grown at TB = 100 °C and TB = 300 °C. Δm₂′ increases with temperature for all conditions, with a pronounced rise at ≳300 °C. At lower growth temperatures (≈100–250 °C), samples with TB = 100 °C consistently exhibit lower Δm₂′ values than those with TB = 300 °C. At higher temperatures, Δm₂′ converges to larger values across pulse times, indicating reduced sensitivity to buffer-layer temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Growth Temp (°C) | 5 s pulse (■) | 2 s pulse (●, TB = 300 °C) | 0.5 s pulse (○) | 2 s pulse (▲, TB = 100 °C) |\\n|---|---|---|---|---|\\n| 100 | ~4.0 | ~3.0 | — | ~2.0 |\\n| 150 | ~3.0 | ~5.0 | — | ~4.0 |\\n| 200 | ~6.0 | ~7.0 | ~6.5 | ~4.5 |\\n| 250 | ~7.0 | ~9.0 | — | ~5.0 |\\n| 300 | ~13.0 | ~13.0 | ~11.0 | — |\\n| 350 | ~13.0 | ~13.0 | ~11.0 | ~10.0 |\\n| 400 | ~13.0 | ~15.0 | ~12.0 | — |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Δm₂ (arbitrary units).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, Δm₂  generally increases with increasing growth temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At lower growth temperatures (≈100–250 °C), the TB = 100 °C triangle series sits below the TB = 300 °C circle/square series, meaning Δm2′ is smaller when the buffer is grown at 100 °C. For example, near 200 °C the triangle point is around ~4–5, while TB=300 °C points cluster closer to ~6–7. At higher temperatures (≥300 °C), Δm2′ rises for all conditions and the separation becomes less consistent, with the largest values appearing near 400 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1.     0.5 - 2 - 2 - 5 s\\n, 2.     2 - 2 - 2 - 5 s\\n, 3.     5 - 2 - 2 - 5 s\"}]}]","bbox":[{"panel_id":"a","x":1,"y":1,"width":459,"height":391}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/experimental-usecase/7/images/fig_4.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 scatter plot","caption_source":"content.json:img_caption"},"width":461,"height":391,"image_format":"jpeg","image_sha256":"59835e988ca650bbf7851e901b18be91b4bf897664e52dce14efa0bb28984a26","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_10_figure_4.jpg","caption":"Figure 4. Film thickness measured by in situ SE as a function of the  $\\mathrm{In}_2\\mathrm{O}_3\\cdot \\mathrm{H}$  ALD cycles on  $\\mathrm{SiO}_2$  open triangles) and on a-Si:H (open circles). Ex situ SE measurements were taken only for 600 and 780 ALD cycles on a-Si:H. Above 600 ALD cycles the selectivity of the process appears to degrade.","id":"train/atomic-layer-deposition/simulation-usecase/10/figure_4","sample_id":"atomic-layer-deposition/simulation-usecase/10/figure_4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between ALD cycles and the thickness of In_{2}O_{3}:H, with SiO_{2} and a-Si:H marked by different symbols\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | SiO₂ (nm) | a-Si:H (nm) |\\n|---|---|---|\\n| 0 | 0 | 0 |\\n| 50 | 0 | 0 |\\n| 100 | 2 | 0 |\\n| 150 | 7 | 0 |\\n| 200 | 12 | 0 |\\n| 250 | 17 | 0 |\\n| 300 | 22 | 0 |\\n| 350 | 27 | 0 |\\n| 400 | 32 | 0 |\\n| 450 | 38 | 0 |\\n| 500 | 43 | 0 |\\n| 550 | 48 | 0 |\\n| 600 | 53 | 0 |\\n| 650 | 58 | 0 |\\n| 700 | 63 | 0 |\\n| 750 | 68 | 0 |\\n| 800 | 75 | 10 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"x-axis: ALD cycles\\n\\n, y-axis: In₂O₃:H thickness (nm)\\n\\n, Triangles: SiO₂ substrate\\n\\n, Circles: a-Si:H substrate\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"On SiO₂, the film thickness increases steadily with cycle number after a small initial delay, indicating normal nucleation and a roughly constant growth per cycle. On a-Si:H, the thickness stays close to zero across most of the cycle range, indicating a long nucleation delay. Only at very high cycle count does measurable growth begin to appear on a-Si:H.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SiO₂ typically presents oxygen-containing / hydroxylated sites that can readily react with many ALD precursors, enabling fast nucleation. In contrast, a-Si:H is H-terminated and comparatively less reactive toward oxide ALD chemistry until it is modified (e.g., oxidized or activated). The plot is consistent with that difference: rapid nucleation on SiO₂ and strongly delayed nucleation on a-Si:H.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Area-selective ALD relies on having one surface where growth proceeds and another where growth is delayed or suppressed over the desired number of cycles. The strong inhibition on a-Si:H relative to SiO₂ provides that selectivity window. This enables pattern formation by locally converting only selected regions into the “fast nucleation” surface while keeping the rest in the “slow nucleation” state.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":1,"width":520,"height":387}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/10/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/10/images/figure_4.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/10/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/10/Alfredo Mameli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"10","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":525,"height":389,"image_format":"jpeg","image_sha256":"8ec2c668760cfc60c12ec924c25d269245c6f1f2f46bdb724e75405942fa248f","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_12.jpg","caption":"FIG. 12. (a) Average wafer coverage and (b)  $3\\sigma$  wafer thickness variability for an ideal and a soft-saturating ALD process. Simulation conditions are the same as for Fig. 10(c).","id":"train/atomic-layer-deposition/simulation-usecase/11/fig_12","sample_id":"atomic-layer-deposition/simulation-usecase/11/fig_12","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the average coverage over time for two conditions: 'Ideal' and 'f2 = 0.1'. The 'Ideal' condition reaches a plateau at approximately 0.9, while 'f2 = 0.1' reaches a plateau slightly below.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates the 3σ variability percentage over time for two conditions: 'Ideal' and 'f2 = 0.1'. Both conditions show a decreasing trend, with 'Ideal' reaching near-zero variability around 1.0 seconds and 'f2 = 0.1' stabilizing around 1.5%.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time, s | Average coverage f_2 = 0.1 | Average coverage Ideal |\\n|---|---|---|\\n| 0.0 | 0.0 | 0.0 |\\n| 0.5 | 0.7 | 0.75 |\\n| 1.0 | 0.9 | 1.0 |\\n| 1.5 | 0.9 | 1.0 |\\n| 2.0 | 0.9 | 1.0 |\\n| 2.5 | 0.9 | 1.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Time, s | 3σ variability (%) f_2 = 0.1 | 3σ variability (%) Ideal |\\n|---|---|---|\\n| 1.0 | 3.0 | 0 |\\n| 1.5 | 2.0 | 0 |\\n| 2.0 | 1.5 | 0 |\\n| 2.5 | 1.5 | 0 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is 0% for the ideal case and about 1.5% for f_2 = 0.1.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Near 0 s, the reaction is not saturated and the surface thickness is highly dependent on dose time, which also means the surface thickness varies a lot in this region. When the surface starts to become saturated, the surface thickness no longer depends on the dose time (ideal case) and only very little for the non-ideal case. In this last case, the surface thickness becomes much less variable, which is reflected in the figure as the variability plateaus.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It is 1.0 for the ideal case and about 0.95 for f_2 = 0.1.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The same as in figure 10c.\"}]}]","bbox":[{"panel_id":"a","x":14,"y":27,"width":657,"height":394},{"panel_id":"b","x":32,"y":452,"width":639,"height":397}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_12.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 scatter plot","caption_source":"content.json:img_caption"},"width":700,"height":853,"image_format":"jpeg","image_sha256":"d4979ee43c00b05e84ef645decb53d84b07fb2160704abbc68d74140ac98e7d8","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_15.jpg","caption":"FIG. 15. Evolution of the average coverage (a) and  $3\\sigma$  variability (b) in  $300\\mathrm{mm}$  wafers for two values of by-product reaction probability:  $\\beta_{bp} = 10^{-4}$  and  $\\beta_{bp} = 10^{-3}$ . The precursor reactivity is  $\\beta_{10} = 10^{-2}$ .","id":"train/atomic-layer-deposition/simulation-usecase/11/fig_15","sample_id":"atomic-layer-deposition/simulation-usecase/11/fig_15","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the average coverage over time for two different β<sub>bp</sub> values, where β<sub>bp</sub> = 10<sup>-4</sup> and β<sub>bp</sub> = 10<sup>-3</sup>.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the 3σ variability over time for two different β<sub>bp</sub> values, where β<sub>bp</sub> = 10<sup>-4</sup> and β<sub>bp</sub> = 10<sup>-3</sup>.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time, s | Average coverage β<sub>bp</sub> = 10<sup>-4</sup> | Average coverage β<sub>bp</sub> = 10<sup>-3</sup> |\\n|---|---|---|\\n| 0.0 | 0.0 | 0.0 |\\n| 0.5 | 0.4 | 0.4 |\\n| 1.0 | 0.85 | 0.8 |\\n| 1.5 | 1.0 | 0.85 |\\n| 2.0 | 1.0 | 0.85 |\\n| 2.5 | 1.0 | 0.85 |\\n| 3.0 | 1.0 | 0.85 |\"},{\"panel_id\":\"b\",\"text\":\"| Time, s | 3σ variability (%) β<sub>bp</sub> = 10<sup>-4</sup> | 3σ variability (%) β<sub>bp</sub> = 10<sup>-3</sup> |\\n|---|---|---|\\n| 1.0 | 95 | 80 |\\n| 1.5 | 10 | 20 |\\n| 2.0 | 2 | 20 |\\n| 2.5 | 2 | 20 |\\n| 3.0 | 2 | 20 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 80 for β<sub>bp</sub> = 10<sup>-3</sup> and 90 for β<sub>bp</sub> = 10<sup>-4</sup>.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Around 20 for β<sub>bp</sub> = 10<sup>-3</sup> and 2 for β<sub>bp</sub> = 10<sup>-4</sup>.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 0.4.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 0.85 for β<sub>bp</sub> = 10<sup>-3</sup> and 1.00 for β<sub>bp</sub> = 10<sup>-4</sup>.\"}]}]","bbox":[{"panel_id":"a","x":35,"y":24,"width":638,"height":399},{"panel_id":"b","x":20,"y":441,"width":652,"height":406}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_15.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/11/images/fig_15.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 scatter plot","caption_source":"content.json:img_caption"},"width":677,"height":850,"image_format":"jpeg","image_sha256":"87c049f1194e0bf91c4bdd66f976cb934a0cd2b14bfd7a4e9ab5a157f1575d1c","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_19_figure_2.jpg","caption":"Figure 2. Comparison of electrical characteristics of the PEALD-IGZO TFTs deposited using $\\mathrm { O } _ { 2 }$ and $\\mathrm { N } _ { 2 } \\mathrm { O }$ plasma reactants. (a) Transfer curves and (b) PBTS results of the IGZO TFTs. XPS (c) O 1s and (d) N 1s peaks of the IGZO films.","id":"train/atomic-layer-deposition/simulation-usecase/19/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/19/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"multi spectra chart"},{"panel_id":"d","label":"multi spectra chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the drain current–gate voltage characteristics of IGZO TFTs fabricated under different ambient conditions. The shift in threshold voltage reflects changes in carrier concentration and device switching behavior.\"},{\"panel_id\":\"b\",\"text\":\"The figure illustrates the positive bias temperature stress (PBTS) stability of IGZO TFTs measured at 95 °C and an electric field of 2 MV/cm. Differences in threshold voltage shift indicate variations in stress reliability under different ambients.\"},{\"panel_id\":\"c\",\"text\":\"The figure presents the O 1s XPS spectra of IGZO films processed under different ambient conditions. The relative peak areas reveal changes in lattice oxygen and oxygen-defect concentrations.\"},{\"panel_id\":\"d\",\"text\":\"The figure shows the N 1s XPS spectra of IGZO films grown under different ambient conditions. The presence and intensity of the N 1s peak confirm nitrogen incorporation into the IGZO matrix.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gate Voltage (V) | Drain Current (A) | Condition          |\\n|-----------------:|-----------------:|------------------|\\n| -6  | 10⁻¹² | O₂⁺ 1 s |\\n| -6  | 10⁻¹² | N₂O⁺ 1 s |\\n| -6  | 10⁻¹³ | N₂O⁺ 5 s |\\n| -5  | 10⁻¹¹ | O₂⁺ 1 s |\\n| -5  | 10⁻¹¹ | N₂O⁺ 1 s |\\n| -5  | 10⁻¹² | N₂O⁺ 5 s |\\n| -4  | 10⁻¹⁰ | O₂⁺ 1 s |\\n| -4  | 10⁻¹⁰ | N₂O⁺ 1 s |\\n| -4  | 10⁻¹¹ | N₂O⁺ 5 s |\\n| -3  | 10⁻⁹  | O₂⁺ 1 s |\\n| -3  | 10⁻⁹  | N₂O⁺ 1 s |\\n| -3  | 10⁻¹⁰ | N₂O⁺ 5 s |\\n| -2  | 10⁻⁷  | O₂⁺ 1 s |\\n| -2  | 10⁻⁷  | N₂O⁺ 1 s |\\n| -2  | 10⁻⁸  | N₂O⁺ 5 s |\\n| -1  | 10⁻⁵  | O₂⁺ 1 s |\\n| -1  | 10⁻⁵  | N₂O⁺ 1 s |\\n| -1  | 10⁻⁶  | N₂O⁺ 5 s |\\n| 0   | 10⁻⁴  | O₂⁺ 1 s |\\n| 0   | 10⁻⁴  | N₂O⁺ 1 s |\\n| 0   | 10⁻⁵  | N₂O⁺ 5 s |\\n| 1   | 10⁻³ | O₂⁺ 1 s |\\n| 1   | 10⁻³ | N₂O⁺ 1 s |\\n| 1   | 10⁻⁴| N₂O⁺ 5 s |\\n| 2   | 10⁻³| O₂⁺ 1 s |\\n| 2   | 10⁻³ | N₂O⁺ 1 s |\\n| 2   | 10⁻³| N₂O⁺ 5 s |\\n| 3   | 10⁻³ | O₂⁺ 1 s |\\n| 3   | 10⁻³ | N₂O⁺ 1 s |\\n| 3   | 10⁻³ | N₂O⁺ 5 s |\\n| 4   | 10⁻³ | O₂⁺ 1 s |\\n| 4   | 10⁻³ | N₂O⁺ 1 s |\\n| 4   | 10⁻³ | N₂O⁺ 5 s |\"},{\"panel_id\":\"b\",\"text\":\"| Gate Voltage (V) | Drain Current (A) | Condition      |\\n|-----------------:|------------------:|----------------|\\n| -6 | 10⁻¹² | O₂⁺ 1 s |\\n| -6 | 10⁻¹² | N₂O⁺ 1 s |\\n| -6 | 10⁻¹² | N₂O⁺ 5 s |\\n| -5 | 10⁻¹² | O₂⁺ 1 s |\\n| -5 | 10⁻¹² | N₂O⁺ 1 s |\\n| -5 | 10⁻¹² | N₂O⁺ 5 s |\\n| -4 | 10⁻¹¹ | O₂⁺ 1 s |\\n| -4 | 10⁻¹¹ | N₂O⁺ 1 s |\\n| -4 | 10⁻¹¹ | N₂O⁺ 5 s |\\n| -3 | 10⁻¹⁰ | O₂⁺ 1 s |\\n| -3 | 10⁻¹⁰ | N₂O⁺ 1 s |\\n| -3 | 10⁻¹⁰ | N₂O⁺ 5 s |\\n| -2 | 10⁻⁸  | O₂⁺ 1 s |\\n| -2 | 10⁻⁸  | N₂O⁺ 1 s |\\n| -2 | 10⁻⁸  | N₂O⁺ 5 s |\\n| -1 | 10⁻⁶  | O₂⁺ 1 s |\\n| -1 | 10⁻⁶  | N₂O⁺ 1 s |\\n| -1 | 10⁻⁶  | N₂O⁺ 5 s |\\n| 0  | 10⁻⁵  | O₂⁺ 1 s |\\n| 0  | 10⁻⁵  | N₂O⁺ 1 s |\\n| 0  | 10⁻⁵  | N₂O⁺ 5 s |\\n| 1  | 10⁻⁴  | O₂⁺ 1 s |\\n| 1  | 10⁻⁴  | N₂O⁺ 1 s |\\n| 1  | 10⁻⁴  | N₂O⁺ 5 s |\\n| 2  | 10⁻⁴  | O₂⁺ 1 s |\\n| 2  | 10⁻⁴  | N₂O⁺ 1 s |\\n| 2  | 10⁻⁴  | N₂O⁺ 5 s |\\n| 3  | 10⁻⁴  | O₂⁺ 1 s |\\n| 3  | 10⁻⁴  | N₂O⁺ 1 s |\\n| 3  | 10⁻⁴  | N₂O⁺ 5 s |\\n| 4  | 10⁻⁴  | O₂⁺ 1 s |\\n| 4  | 10⁻⁴  | N₂O⁺ 1 s |\\n| 4  | 10⁻³  | N₂O⁺ 5 s |\"},{\"panel_id\":\"c\",\"text\":\"| Binding energy (eV) | Intensity (a.u.) |\\n|---|---|\\n|534 | 70.5% |\\n|532 | 22.5% |\\n|530 | 7.0% |\"},{\"panel_id\":\"d\",\"text\":\"| Binding Energy (eV) | Normalized Intensity | Condition        | Remark |\\n|-------------------:|---------------------:|------------------|--------|\\n| 408 | 0.0 | All O₂⁺ 1 s | — |\\n| 408 | 0.0 | All N₂O⁺ 1 s | — |\\n| 408 | 0.1 | All N₂O⁺ 5 s | — |\\n| 404 | 0.0 | All O₂⁺ 1 s | — |\\n| 404 | 0.1 | All N₂O⁺ 1 s | — |\\n| 404 | 0.1 | All N₂O⁺ 5 s | — |\\n| 400 | 0.2 | All O₂⁺ 1 s | — |\\n| 400 | 0.2 | All N₂O⁺ 1 s | — |\\n| 400 | 0.2 | All N₂O⁺ 5 s | — |\\n| 396 | 1.0 | All O₂⁺ 1 s | Peak |\\n| 396 | 1.0 | All N₂O⁺ 1 s | Peak |\\n| 396 | 1.0 | All N₂O⁺ 5 s | Peak |\\n| 392 | 0.2 | All O₂⁺ 1 s | — |\\n| 392 | 0.2 | All N₂O⁺ 1 s | — |\\n| 392 | 0.2 | All N₂O⁺ 5 s | — |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Metal–nitrogen (Ga–N or M–N) bonds. This indicates successful incorporation of nitrogen into the oxide lattice via anion substitution, which helps passivate oxygen vacancies and improve device stability.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Decrease. The relative area of the O𝑑𝑒𝑓 peak decreases from 22.5% in the O₂ sample to 19.9% in the N₂O (5 s) sample, indicating a reduction in oxygen vacancy concentration.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The curves shift to the right, indicating that N2O plasma treatment increases the turn-on voltage compared to the O2 plasma reference, likely due to a reduction in carrier concentration or modification of interface traps.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lower ΔV𝑇𝐻. The N₂O-treated devices (Red/Blue) exhibit significantly smaller threshold voltage shifts under stress compared to the O₂ reference (Black), which is critical for maintaining stable current driving in OLED pixels.\"}]}]","bbox":[{"panel_id":"a","x":45,"y":44,"width":372,"height":400},{"panel_id":"b","x":477,"y":0,"width":903,"height":449},{"panel_id":"c","x":76,"y":501,"width":864,"height":470},{"panel_id":"d","x":1025,"y":562,"width":363,"height":407}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/Dong-Gyu Kim et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1389,"height":972,"image_format":"jpeg","image_sha256":"a054701d795fbac77bda27ed9618d42eef0d1b9c74b1fb43a4cb9b1abf98cfa4","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_19_figure_7.jpg","caption":"Figure 7. PBTS stability of the selective $\\mathrm { N } _ { 2 } \\mathrm { O }$ applied IGZO TFTs. (a) Transfer curves as a function of PBTS time of the reference and selectivel $\\mathrm { N } _ { 2 } \\mathrm { O }$ plasma applied IGZO TFTs. (b) Degree of $\\mathrm { V } _ { \\mathrm { T H } }$ shift after 10000 s of PBTS time for IGZO TFTs.","id":"train/atomic-layer-deposition/simulation-usecase/19/figure_7","sample_id":"atomic-layer-deposition/simulation-usecase/19/figure_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"grouped bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Transfer characteristics (drain current versus gate voltage) of IGZO TFTs under positive bias temperature stress (PBTS) for devices where individual cation layers (In, Ga, or Zn) are selectively treated with N₂O plasma, showing different degrees of threshold voltage shift.\"},{\"panel_id\":\"b\",\"text\":\"Summary of the corresponding threshold voltage shifts (ΔV_TH), demonstrating that N₂O treatment during the gallium oxide sub-cycle significantly improves stability (~50% reduction), while treatment during the indium oxide sub-cycle degrades stability (≈68% increase in ΔV_TH).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Gate voltage (V) | Pristine (A) | 100 s (A) | 500 s (A) | 1500 s (A) | 2500 s (A) | 3500 s (A) | 3600 s (A) | 7200 s (A) | 10000 s (A) |\\n|------------------|--------------|-----------|-----------|------------|------------|------------|------------|------------|-------------|\\n| -6               | 1e-12        | 1e-12     | 1e-12     | 1e-12      | 1e-12      | 1e-12      | 1e-12      | 1e-12      | 1e-12       |\\n| -4               | 1e-11        | 1e-11     | 1e-11     | 1e-11      | 1e-11      | 1e-11      | 1e-11      | 1e-11      | 1e-11       |\\n| -2               | 1e-10        | 1e-10     | 1e-10     | 1e-10      | 1e-10      | 1e-10      | 1e-10      | 1e-10      | 1e-10       |\\n| 0                | 1e-9         | 1e-9      | 1e-9      | 1e-9       | 1e-9       | 1e-9       | 1e-9       | 1e-9       | 1e-9        |\\n| 2                | 1e-7         | 1e-7      | 1e-7      | 1e-7       | 1e-7       | 1e-7       | 1e-7       | 1e-7       | 1e-7        |\\n| 4                | 1e-5         | 1e-5      | 1e-5      | 1e-5       | 1e-5       | 1e-5       | 1e-5       | 1e-5       | 1e-5        |\"},{\"panel_id\":\"b\",\"text\":\"| Controlled cation elements | O₂ 1s (Ref) ΔV_TH (V) | N₂O 1s ΔV_TH (V) | N₂O 5s ΔV_TH (V) |\\n|----------------------------|-----------------------|------------------|------------------|\\n| Indium oxide               | ~2.4                  | ~1.6 (33% decrease) | ~1.2 (50% decrease) |\\n| Gallium oxide              | ~2.0                  | ~1.0 (50% decrease) | ~0.8 (60% decrease) |\\n| Zinc oxide                      | ~1.8                  | ~1.7 (8% decrease)  | ~1.6 (11% decrease) |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Gallium oxide. The Gallium oxide modification results in a ~50% decrease in ΔV𝑇𝐻 (improved stability), whereas the Indium oxide modification causes a ~68% increase (worsened stability) relative to the reference.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Suppression of oxygen vacancies. Gallium is a strong oxygen binder. Selective N₂O treatment during the Gallium cycle effectively reduces oxygen vacancies and defect states (which act as electron traps) without severely disrupting channel conduction, leading to the most stable device operation.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. For the Zinc Oxide–controlled cation, increasing the N₂O exposure to 5 s significantly increases the instability, resulting in a ΔV𝑇𝐻 of +0.83 V compared to the more stable Gallium condition (+0.45 V).\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Selective N₂O on Gallium Oxide. This condition yields the lowest ΔV𝑇𝐻 (+0.45 V) compared to Indium (+1.51 V) or Zinc (+0.83 V), ensuring the most stable switching performance under thermal and bias stress.\"}]}]","bbox":[{"panel_id":"a","x":17,"y":11,"width":905,"height":349},{"panel_id":"b","x":978,"y":38,"width":416,"height":316}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/19/Dong-Gyu Kim et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"19","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1394,"height":364,"image_format":"jpeg","image_sha256":"b44085f6376ab0c47f9d6cdaafd407b19bb616149283cc6b21f19a1b44343b24","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_5.jpg","caption":"Figure 5. Growth properties as a function of surface hydroxyl density. (a) Al atoms added vs OH density. (b) Ligand/Al ratio vs surface OH concentration. Present results compared to data from Puurunen.","id":"train/atomic-layer-deposition/simulation-usecase/25/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/25/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows Al atoms added per unit area (Al/nm²) against OH surface density (OH/nm²) for the present study (blue) and Puurunen (black). Both datasets show that generally higher OH surface density corresponds to greater Al uptake.\"},{\"panel_id\":\"b\",\"text\":\"The scatter plot displays the ligand-to-aluminum ratio versus OH surface density  for the present study (blue) and Puurunen (black). Generally, the ligand:Al ratio decreases as OH density increases.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"|Study| Al atoms added, Al/nm² | OH Surface Density, OH/nm² | \\n|---|---|---|\\n|Puurunen|2.23| 1.48|\\n|Puurunen|2.94| 3.51|\\n|Puurunen|3.71| 5.49|\\n|Puurunen|4.43| 7.48|\\n|Present study|2.78| 2.17|\\n|Present study|3.32| 4.11|\\n|Present study|3.44| 4.08|\\n|Present study|3.62| 4.22|\\n|Present study|3.86| 4.33|\\n|Present study|3.69| 4.49|\\n|Present study|3.12| 4.61|\\n|Present study|4.06| 4.59|\\n|Present study|3.85| 4.72|\\n|Present study|3.86| 5.16|\"},{\"panel_id\":\"b\",\"text\":\"|Study| OH Surface Density, OH/nm² |  Ligand : Aluminum Ratio |\\n|---|---|---|\\n|Puurunen|1.18| 2.39|\\n|Puurunen|3.40| 1.84|\\n|Puurunen|6.01| 1.49|\\n|Puurunen|8.78| 1.2|\\n|Present study|2.24| 1.88|\\n|Present study|4.06| 1.68|\\n|Present study|4.13| 1.58|\\n|Present study|4.29| 1.65|\\n|Present study|4.26| 1.59|\\n|Present study|4.47| 1.37|\\n|Present study|4.54| 1.55|\\n|Present study|4.64| 1.48|\\n|Present study|4.69| 1.42|\\n|Present study|5.14| 1.52|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Puurunen's results show a positive correlation. This conclusion is supported by the upward trend of the data points, which rises from left to right. This indicates that as the OH surface density increases (on the x-axis), the number of Al atoms added also increases (on the y-axis).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Puurunen's results  show a negative correlation. This is supported by the downward trend  of the data points, which descends from left to right. This indicates that as the OH surface density increases, the Ligand : Aluminum Ratio also decreases.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Pruununen’s  study covers a wider range of OH surface densities.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":6,"width":561,"height":452},{"panel_id":"b","x":569,"y":3,"width":553,"height":449}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/25/images/figure_5.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 scatter plot","caption_source":"content.json:img_caption"},"width":1125,"height":459,"image_format":"jpeg","image_sha256":"a17496a518bb6a2408d50a52b34263c2089f37bc13b02fb8a6fb41dfa4bd90e3","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_26_figure_15.jpg","caption":"Figure 15. Growth rate of  $\\mathsf{HfO_2}$  in different circumstances versus time for the 16th-20th cycles. The thickness of a monolayer of  $\\mathsf{HfO_2}$  is  $2.8\\dot{A}$ . Cross shows growth rate considering all observed ALD reactions. Circle shows growth rate when direct ligand exchange during oxygen pulse is considered rather than cooperative effect. Triangle shows growth rate when migration process is switched off. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]","id":"train/atomic-layer-deposition/simulation-usecase/26/figure_15","sample_id":"atomic-layer-deposition/simulation-usecase/26/figure_15","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between simulation time and growth rate, with three different conditions represented by different markers: considering all ALD reactions (red +), cooperative effect is excluded (green circles), and when the migration process is switched off (blue triangle). Excl. cooperation gives the highest growth (~1.4–1.6 Å/cycle), incl. all is intermediate (~0.5–0.75 Å/cycle), and excl. migration is lowest (~0.12–0.22 Å/cycle).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Simulation time (s) | Growth rate – incl. all (Å/cycle) | Growth rate – excl. cooperation (Å/cycle) | Growth rate – excl. migration (Å/cycle) |\\n|---|---|---|---|\\n| 0.0064 | 0.75 | 1.58 | 0.19 |\\n| 0.0068 | 0.59 | 1.45 | 0.12 |\\n| 0.0072 | 0.50 | 1.42 | 0.21 |\\n| 0.0076 | 0.50 | 1.42 | 0.22 |\\n| 0.0080 | 0.65 | 1.59 | 0.16 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Excl. cooperation, Incl. all, Excl. migration\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Disabling migration reduces the growth rate compared to the full model.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Migration is more important to sustain growth. The evidence is that when migration is disabled (“excl. migration”), the growth rate collapses to very low values (~0.12–0.22 Å/cycle) at every time point, far below the full model (~0.5–0.75 Å/cycle). In contrast, disabling cooperation (“excl. cooperation”) does not reduce growth.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Cooperation hinders film growth in this model. The evidence is that when cooperation is removed (“excl. cooperation”), the growth rate jumps to the highest values in the plot (~1.4–1.6 Å/cycle) across all times, well above the full model (~0.5–0.75 Å/cycle). By contrast, removing migration lowers growth (~0.12–0.22 Å/cycle), so migration is not what limits growth.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":638,"height":444}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/26/images/figure_15.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/26/images/figure_15.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/26/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/26/Mahdi Shirazi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"26","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":644,"height":450,"image_format":"jpeg","image_sha256":"0ffbd08d89218c00f4c08c46e3cc441ee3bb05a7906faccb90545f624916f575","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_28_fig_10.jpg","caption":"Fig. 10. Surface state of the adsorbed complex  $-\\mathrm{MCl}_x$  of the precursor vs. process temperature. Circles and squares correspond to experiments with  $\\mathrm{ZrO_2}$  from [10] and [11], respectively; triangles correspond to  $\\mathrm{HfO_2}$  experiment [6]; dashed and dashed-dotted lines correspond to the present study modeling for  $\\mathrm{ZrO_2}$  and  $\\mathrm{HfO_2}$ , respectively.","id":"train/atomic-layer-deposition/simulation-usecase/28/fig_10","sample_id":"atomic-layer-deposition/simulation-usecase/28/fig_10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between temperature (T, °C) and the ratio x = Cl / Me. Data points represent experimental measurements, whereas calculations for Zr and Hf are shown with lines..\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| T,°C | x = Cl / Me |\\n|---|---|\\n| 100 | 2.00 |\\n| 200 | 2.25 |\\n| 300 | 2.30 |\\n| 400 | 2.60 |\\n| 500 | 2.80 |\\n| 600 | 3.00 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No the experimental value does not nicely overlap. Only the Hf experiments of J. Aarik come close, however they are still of by .25 on average.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The calculated values are in the error margins of the data points, so it could be that they overlap.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"They both show an s shape, where initially there is an increase with temperature, but this saturates.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No it is not, the upper limit of the error bars is at 3.50, this would mean that out of four chlorine atoms, still more than three are present. This means that non of them reactive with the surface leading to adsorption.\"}]}]","bbox":[{"panel_id":"a","x":10,"y":8,"width":712,"height":497}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/Maxim Deminsky et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":722,"height":508,"image_format":"jpeg","image_sha256":"ad3a411f1eac5146c5d7cc09bb81c0e4464aa61c4124ed004d7a3cd1002c97be","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_28_fig_11.jpg","caption":"Fig. 11. Temperature dependence of the residual chlorine concentration in the growing film; triangles are experimental points [10], and diamonds are simulation results.","id":"train/atomic-layer-deposition/simulation-usecase/28/fig_11","sample_id":"atomic-layer-deposition/simulation-usecase/28/fig_11","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between temperature and chlorine concentration, indicating a negative correlation where chlorine concentration decreases as temperature increases.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature, °C | Chlorine concentration, % |\\n|---|---|\\n| 200 | 2.3 |\\n| 250 | 1.5 |\\n| 300 | 1.0 |\\n| 350 | 0.5 |\\n| 400 | 0.1 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The temperature has a huge influence on the reaction rates that occur during the ALD cycle. This can be the reason for the higher chlorine content.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"XPS is a technique that can be used to determine the chlorine concentration In this technique, electrons are removed from the material by photons that give insight into the materials composition.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.5%\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":0,"y":8,"width":595,"height":411}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/Maxim Deminsky et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":595,"height":419,"image_format":"jpeg","image_sha256":"59e74a6198d4390a0375566f82fdb3833755375867be2fd5485bded3cfd32002","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_28_fig_6.jpg","caption":"Fig. 6.  $\\mathrm{ZrO_2}$  film mass increment as a function of the precursor source temperature. Circles and solid line are experimental [10] and modeling results at the reaction chamber temperature 290  $^\\circ \\mathrm{C}$ ; squares and dashed line are experimental [10] and modeling results at the reaction chamber temperature  $180^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/28/fig_6","sample_id":"atomic-layer-deposition/simulation-usecase/28/fig_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the relationship between the mass increment and the ZrCl₄ source temperature for two different reactor temperatures. As the temperature increases, the mass increment also increases, showing a positive correlation, till there is soft-saturation for both reactor temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ZrCl₄ source temperature, °C | Mass increment, a.u. 290|Mass increment, a.u. 180|\\n|---|---|---|\\n| 130 | 0.2 |-|\\n| 135 | 0.3 |-|\\n| 140 | 0.67 |-|\\n| 145 | 0.85 |0.80|\\n| 150 | 0.91 |1.05|\\n| 155 | 0.95 |1.1|\\n| 160 | 1.0 |1.15|\\n| 165 | - |1.20|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Lower temperature can lead to more ligand incorporation which in its turn leads to higher mass as Cl is heavier than oxygen.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"If this difference is not high enough there can be condensation of the precursor in the reactor, whereas there is always the desire to only work with gas phase species.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For the higher source temperature ranges. This is mostly in the saturated growth regime.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.15 a.u.\"}]}]","bbox":[{"panel_id":"a","x":5,"y":7,"width":584,"height":478}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/Maxim Deminsky et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":592,"height":486,"image_format":"jpeg","image_sha256":"116b4df59daba097bdc33ef8cdac32311361077444ccba3b838ad9b03d305870","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_28_fig_7.jpg","caption":"Fig. 7.  $\\mathrm{ZrO_2}$  film growth rate (average film mass and thickness increments per cycle) as a function of the process temperature: squares and circles correspond to experimental average film thicknesses and mass increments per cycle [10]. Dashed and solid lines correspond to the calculated results by the minimum and extended mechanism. The dashed area corresponds to sensitivity analysis for the extended mechanism.","id":"train/atomic-layer-deposition/simulation-usecase/28/fig_7","sample_id":"atomic-layer-deposition/simulation-usecase/28/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between temperature (T °C) and the thickness and mass increment per cycle, experimentally measured and calculated using minimum and extended mechanisms.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| T °C |average increment |\\n|---|---|\\n| 0 | 0.4 | \\n| 100 | 1.2 | \\n| 200 | 1.0 | \\n| 300 | 0.9 |\\n| 400 | 0.75 | \\n| 500 | 0.7 | \\n| 600 | 0.55 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The points fall in the range of the extended model and can therefore be fitted with such a model.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"There probably is a minimum temperature as the source temperature has to be lower than the processing temperature, but high enough to evaporate the precursor.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"1.2 a.u./cycle\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Except for some outliers in general the model overestimates the experimental data, even when considering the error bars.\"}]}]","bbox":[{"panel_id":"a","x":9,"y":9,"width":582,"height":442}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/Maxim Deminsky et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":592,"height":452,"image_format":"jpeg","image_sha256":"5a3b4afd7e34583b447a8dc3edc84090e153a1769cea7cc984a428a5959134c2","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_28_fig_8.jpg","caption":"Fig. 8.  $\\mathrm{HfO_2}$  film growth rate (average film mass and thickness increments per cycle) as a function of the process temperature: squares and circles correspond to experimental results [6,8]. Dashed and solid lines correspond to the calculated results by the minimum and extended mechanism. Dashed area corresponds to sensitivity analysis for the extended mechanism.","id":"train/atomic-layer-deposition/simulation-usecase/28/fig_8","sample_id":"atomic-layer-deposition/simulation-usecase/28/fig_8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the relationship between temperature (T, °C) and the thickness and mass increments per cycle for an experiment. The data points are compared with calculated values using minimum and extended mechanisms.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| T °C |average increment |\\n|---|---|\\n| 0 | 0.55 | \\n| 100 | 1.2 | \\n| 200 | 1.1 | \\n| 300 | 0.95 |\\n| 400 | 0.8 | \\n| 500 | 0.6 | \\n| 600 | 0.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The simple model does not cover the data points, but the range that the extended model does overlap with all the data points, so the extended model is required.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"175 degrees Celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The maximum temperature that the reactor can achieve, or the stability of this temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":8,"y":0,"width":580,"height":467}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/28/Maxim Deminsky et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"28","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":589,"height":467,"image_format":"jpeg","image_sha256":"99720946e4bbea0fe5e4ebda25bcf7c1424747eaf14c677f1af2949a251d5822","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_29_fig6.jpg","caption":"Fig.6. the experimental growth rate in which deposition was carried out at  $50^{\\circ}C$  substrate temperature using  $3s$  water vapor pulse (Xie et al., 2007). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.)","id":"train/atomic-layer-deposition/simulation-usecase/29/fig6","sample_id":"atomic-layer-deposition/simulation-usecase/29/fig6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot illustrates the validation of a computational model for Atomic Layer Deposition (ALD) by comparing the calculated growth rates (nm/cycle) against pulse time (s)  with experimental data (measured and interpolated). The coefficient of determination is 0.85.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse Time (s) | Growth Rate (nm/cyc) | Data series |\\n|---|---|---|\\n| 0 |  0 | GR exp |  \\n| 1 |  0.121 | GR exp |  \\n| 2 |  0.121 | GR exp |  \\n| 10 |  0.120 | GR exp |\\n| 0.0 | 0.001 | Gr exp (Interp) |\\n| 0.2 | 0.013 | Gr exp (Interp) |\\n| 0.4 | 0.033 | Gr exp (Interp) |\\n| 0.6 | 0.059 | Gr exp (Interp) |\\n| 0.8 | 0.079 | Gr exp (Interp) |\\n| 1.0 | 0.101 | Gr exp (Interp) |\\n| 1.2 | 0.108 | Gr exp (Interp) |\\n| 1.4 | 0.114 | Gr exp (Interp) |\\n| 1.6 | 0.119 | Gr exp (Interp) |\\n| 1.8 | 0.121 | Gr exp (Interp) |\\n| 2.0 | 0.120 | Gr exp (Interp) |\\n| 4.0 | 0.119 | Gr exp (Interp) |\\n| 6.0 | 0.120 | Gr exp (Interp) |\\n| 8.0 | 0.120 | Gr exp (Interp) |\\n| 10.0 | 0.120 | Gr exp (Interp) |\\n| 0.2 | 0.039 | Gr calc |\\n| 0.5 | 0.056 | Gr calc |\\n| 0.6 | 0.066 | Gr calc |\\n| 0.8 | 0.075 | Gr calc |\\n| 1.0 | 0.081 | Gr calc |\\n| 1.2 | 0.089 | Gr calc |\\n| 1.4 | 0.095 | Gr calc |\\n| 1.6 | 0.099 | Gr calc |\\n| 1.8 | 0.104 | Gr calc |\\n| 2.0 | 0.109 | Gr calc |\\n| 4.0 | 0.128 | Gr calc |\\n| 6.0 | 0.135 | Gr calc |\\n| 8.0 | 0.138 | Gr calc |\\n| 10.0 | 0.139 | Gr calc |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The transition occurs between the 2-second and 4-second pulse times. The theoretical model underestimates the growth rate at 2 s. By the 4 s mark, the model surpasses the experimental data.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experimental data reaches saturation faster than the theoretical model .\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experimental data more closely resembles an ideal ALD process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The theoretical model predicts that the substrate requires a pulse time of approximately 4 to 6 seconds to reach full saturation, whereas the experimental data demonstrates that saturation is actually achieved in just 1 second. Consequently, an engineer relying solely on the model would likely program the reactor to dose the precursor for longer than necessary to ensure full coverage. This discrepancy leads to waste because the excess precursor introduced after the 1-second mark cannot react with the already saturated surface and is simply purged from the system.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":6,"width":967,"height":562}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":969,"height":572,"image_format":"jpeg","image_sha256":"9a55ee32710a35fbcbc30149a01ef7a4839d35a9ec283fb87f62a6fef91b36a9","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_29_fig7.jpg","caption":"Fig.7. t t t references to colour in this figure legend, the reader is referred to the web version of this article.)","id":"train/atomic-layer-deposition/simulation-usecase/29/fig7","sample_id":"atomic-layer-deposition/simulation-usecase/29/fig7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot illustrates the validation of a computational model for Atomic Layer Deposition (ALD) with TDEAT by comparing the calculated growth rates (nm/cycle) against pulse time (s) with experimental data (measured and interpolated). The coefficient of determination is 0.99, which indicates a strong correlation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse Time (s) | Growth Rate (nm/cyc) | Data series |\\n| --- | --- | --- |\\n| 0,0 | 0,000 | Gr exp |\\n| 1,0 | 0,023 | Gr exp |\\n| 2,0 | 0,040 | Gr exp |\\n| 4,0 | 0,040 | Gr exp |\\n| 0,0 | 0,000 | Gr exp (interpolated) |\\n| 0,4 | 0,010 | Gr exp (interpolated) |\\n| 0,8 | 0,020 | Gr exp (interpolated) |\\n| 1,0 | 0,023 | Gr exp (interpolated) |\\n| 1,4 | 0,027 | Gr exp (interpolated) |\\n| 1,8 | 0,032 | Gr exp (interpolated) |\\n| 2,0 | 0,033 | Gr exp (interpolated) |\\n| 2,4 | 0,035 | Gr exp (interpolated) |\\n| 2,8 | 0,038 | Gr exp (interpolated) |\\n| 3,0 | 0,038 | Gr exp (interpolated) |\\n| 3,4 | 0,039 | Gr exp (interpolated) |\\n| 3,8 | 0,040 | Gr exp (interpolated) |\\n| 4,0 | 0,040 | Gr exp (interpolated) |\\n| 0,4 | 0,014 | Gr calc |\\n| 0,8 | 0,020 | Gr calc |\\n| 1,0 | 0,023 | Gr calc |\\n| 1,4 | 0,028 | Gr calc |\\n| 1,8 | 0,030 | Gr calc |\\n| 2,0 | 0,032 | Gr calc |\\n| 2,4 | 0,034 | Gr calc |\\n| 2,8 | 0,036 | Gr calc |\\n| 3,0 | 0,037 | Gr calc |\\n| 3,4 | 0,039 | Gr calc |\\n| 3,8 | 0,041 | Gr calc |\\n| 4,0 | 0,041 | Gr calc |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The graph displays an R² value of 0.99. It implies an excellent statistical fit.  This value means that the theoretical model explains 99% of the variance in the experimental data and predicts the growth rate with very high accuracy .\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The data point at 2 seconds.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. At 0.5 seconds, the growth rate is only about ~0.01 nm/cycle. It is below the saturation level of ~0.04 nm/cyc. This indicates the surface is not fully covered, which would lead to poor film quality.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 2s.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":5,"width":961,"height":607}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig7.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":967,"height":619,"image_format":"jpeg","image_sha256":"3f2bebfc418df234be12141438719f3801436ac8c6f06f24201945d1ff1fdecc","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_29_fig8.jpg","caption":"Fig.8. of the references to colour in this figure legend, the reader is referred to the web version of this article.)","id":"train/atomic-layer-deposition/simulation-usecase/29/fig8","sample_id":"atomic-layer-deposition/simulation-usecase/29/fig8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot illustrates the validation of a computational model for Atomic Layer Deposition (ALD) with TDEAH by comparing the calculated growth rates (nm/cycle) against pulse time (s) with experimental data. The coefficient of determination is 0.93, which indicates a strong correlation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse Time (s) | Growth Rate (nm/cyc) | Data series |\\n| --- | --- | --- |\\n| 0,0 | 0,000 | Gr exp |\\n| 1,0 | 0,039 | Gr exp |\\n| 2,0 | 0,088 | Gr exp |\\n| 3,0 | 0,115 | Gr exp |\\n| 4,0 | 0,121 | Gr exp |\\n| 5,9 | 0,133 | Gr exp |\\n| 8,0 | 0,145 | Gr exp |\\n| 12,0 | 0,141 | Gr exp |\\n| 1,0 | 0,064 | Gr calc |\\n| 2,0 | 0,079 | Gr calc |\\n| 3,0 | 0,094 | Gr calc |\\n| 4,0 | 0,109 | Gr calc |\\n| 6,0 | 0,125 | Gr calc |\\n| 8,0 | 0,134 | Gr calc |\\n| 12,0 | 0,150 | Gr calc |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It generally underestimates the experimental values, except for the first and the last data points.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After ~ 8s.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. While the experimental data shows the growth rate reaching a saturation between 8 and 12 seconds, the model predicts a continued increase in growth rate during that same interval. This discrepancy leads the model to overestimate the growth rate at 12 seconds, failing to replicate the self-limiting behavior of an ALD process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 1 and 2 seconds and between 8 and 12 seconds.\"}]}]","bbox":[{"panel_id":"a","x":7,"y":9,"width":964,"height":616}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":970,"height":625,"image_format":"jpeg","image_sha256":"3b1c161d5fc6f87a23dfa3250a2782dadee07cb5d851777bb22e78a53186ba93","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_29_fig9.jpg","caption":"Fig.9. t t references to colour in this figure legend, the reader is referred to the web version of this article.)","id":"train/atomic-layer-deposition/simulation-usecase/29/fig9","sample_id":"atomic-layer-deposition/simulation-usecase/29/fig9","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot illustrates the validation of a computational model for Atomic Layer Deposition (ALD) with TEMAH by comparing the calculated growth rates (nm/cycle) against pulse time (s) with experimental data. The coefficient of determination is 0.93, which indicates a strong correlation.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Pulse time (s) | Growth Rate (nm/cyc) | Data series |\\n|---|---|---|\\n| 0,0 | 0,000 | Gr exp |\\n| 0,2 | 0,030 | Gr exp |\\n| 0,5 | 0,066 | Gr exp |\\n| 0,8 | 0,070 | Gr exp |\\n| 1,2 | 0,071 | Gr exp |\\n| 1,6 | 0,072 | Gr exp |\\n| 2,0 | 0,072 | Gr exp |\\n| 3,0 | 0,074 | Gr exp |\\n| 4,0 | 0,074 | Gr exp |\\n| 0,2 | 0,044 | Gr calc |\\n| 0,5 | 0,054 | Gr calc |\\n| 0,8 | 0,061 | Gr calc |\\n| 1,2 | 0,067 | Gr calc |\\n| 1,6 | 0,069 | Gr calc |\\n| 2,0 | 0,071 | Gr calc |\\n| 3,0 | 0,073 | Gr calc |\\n| 4,0 | 0,075 | Gr calc |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experimental process reaches saturation faster.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The beginning of the curve contributes the most.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"After ~ 1s.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It would have no effect on the film thickness. The experimental data shows a saturation plateau; this confirms the process is self-limiting. Once the surface is saturated (which happens before 2 seconds), adding extra precursor due to an accidental fluctuation does not result in extra growth.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":967,"height":624}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/images/fig9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/29/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"29","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":967,"height":630,"image_format":"jpeg","image_sha256":"365098327a8e02c7c22cdcb80fb0ddbdf8e18871912a8d685bb4db4555db1f5f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_32_fig_16.jpg","caption":"Fig. 16 - Time evolution of the maximum reactant mole fraction inside the reactor chamber.","id":"train/atomic-layer-deposition/simulation-usecase/32/fig_16","sample_id":"atomic-layer-deposition/simulation-usecase/32/fig_16","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart plot the maximum reactant mole fraction as a function of time for TMA and water.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Water Mole Fraction | TMA Mole Fraction |\\n|----------|----------------------|--------------------|\\n| 0.0      | 1.00                | 1.00              |\\n| 0.1      | 0.70                | 0.30              |\\n| 0.2      | 0.40                | 0.20              |\\n| 0.3      | 0.25                | 0.15              |\\n| 0.5      | 0.10                | 0.05              |\\n| 1.0      | 0.02                | 0.02              |\\n| 2.0      | 0.00                | 0.01              |\\n| 5.0      | 0.00                | 0.01             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In order to study the purging efficiency of the reactor, the maximum reactant mole fraction inside the whole chamber was traced during the ALD pulses and purging steps of the ALD cycle, for both reactants.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The maximum mole fraction inside the chamber during the first stages of the exposure time is close to 1, as the computed reactant pulses have a high reactant composition. After the reactant pulses, the maximum mole fraction quickly drops, as the reactor purging step begins.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The results are obtained bysimulating only the transport of chemical species inside thereactor. In order to get a more reliable value for the purging time, chemical reaction kinetics and adsorption/desorption ofspecies on the reactor walls must be taken into account\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":1008,"height":517}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/fig_16.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/fig_16.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/reactorGP Gakis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"32","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1014,"height":525,"image_format":"jpeg","image_sha256":"22f11d17f703bcaedbd84692e28bee36647d115f978110b3465bf935974090f1","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_32_fig_8.jpg","caption":"Fig. 8 - Feeding system model results for the TMA and  $\\mathbf{H}_2\\mathbf{O}$  reactant pulses: a) flow rate, b) molar fraction averaged at the feeding system outlet.","id":"train/atomic-layer-deposition/simulation-usecase/32/fig_8","sample_id":"atomic-layer-deposition/simulation-usecase/32/fig_8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the results from feeding system model of the flow rate of TMA and water over time, with both decreasing rapidly.\"},{\"panel_id\":\"b\",\"text\":\"The line chart displays the averaged molar fraction of TMA and water over time for feeding system outlet, with both fractions decreasing.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Flow Rate - Water (sccm) | Flow Rate - TMA (sccm) |\\n|----------|---------------------------|-------------------------|\\n| 0.00     | 700                       | 450                     |\\n| 0.02     | 500                       | 300                     |\\n| 0.05     | 200                       | 100                     |\\n| 0.10     | 50                        | 20                      |\\n| 0.15     | -                       | -        |\"},{\"panel_id\":\"b\",\"text\":\"| Time (s) | Molar Fraction - Water | Molar Fraction - TMA |\\n|----------|-------------------------|-----------------------|\\n| 0.00     | 0.95                    | 0.90                  |\\n| 0.02     | 0.80                    | 0.60                  |\\n| 0.05     | 0.40                    | 0.20                  |\\n| 0.10     | 0.10                    | 0.05                  |\\n| 0.15    | 0.06                     |0.05                   |\\n| 0.2        | 0.03                  |0.02                   |\\n| 0.25     |     0                        | 0                   |\\n| 0.3       | 0                         |0                   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The calculated TMA and H2O pulses as a function of time, in terms of flow rate  and averaged molar fraction at the feeding system outlet are shown in images a and b respectively\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A higher flow rate for H2O than for TMA for the whole duration of the pulse, leading to an overall higher quantity of the delivered H2O. This is attributed to the higher vapor pressure of H2O inside the reactant bottles and to the longer opening time of the ALD valve above the H2O bottle(60 ms and 25 ms for H2O and TMA, respectively).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The reactant molar fraction is substantial during both pulses, whereas Ar represents only roughly10 wt% of the gas mixture. Notably, it is predicted that although the H2O valve opening time is longer and the quantity of H2O entering the feeding system is higher, the TMA molecules evacuate the feeding system slower, i.e the feeding system takes more time to purge. This is attributed to the slower diffusion rate in N2of TMA compared to H2O, since the molecules of the former are bigger and heavier than the latter\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1. The feeding system is purged from both reactants in 0.3 s. \\n2. The total calculated amounts entering the reactor are 0.63 mg for TMA and 0.408 mg for H2O.\\n3. The flow rate of the feeding system after the reactant pulses are 30 sccm\"}]}]","bbox":[{"panel_id":"a","x":0,"y":2,"width":508,"height":403},{"panel_id":"b","x":514,"y":7,"width":492,"height":391}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/fig_8.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/reactorGP Gakis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"32","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1012,"height":458,"image_format":"jpeg","image_sha256":"6da731400c5af3cad2d32f6f968de9da5cbd8276c8e8f7dcf71fa684b775617a","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-deposition_simulation-usecase_32_fig_9.jpg","caption":"Fig. 9 - Outlet pressure variation vs time: model predictions vs Pirani gauge measurements.","id":"train/atomic-layer-deposition/simulation-usecase/32/fig_9","sample_id":"atomic-layer-deposition/simulation-usecase/32/fig_9","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart compares the outlet pressure over time for experimental data and a model.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Time (s) | Outlet Pressure - Experiments (Torr) | Outlet Pressure - Model (Torr) |\\n|----------|---------------------------------------|---------------------------------|\\n| 0        | 0.049                                | 0.049                          |\\n| 5        | 0.075 (H₂O peak)                     | 0.074                          |\\n| 10       | 0.058 (TMA peak)                     | 0.057                          |\\n| 15       | 0.075 (H₂O peak)                     | 0.074                          |\\n| 20       | 0.049                                | 0.049                          |\\n| 25       | 0.075 (H₂O peak)                     | 0.074                          |\\n| 30       | 0.058 (TMA peak)                     | 0.057                          |\\n| 35       | 0.075 (H₂O peak)                     | 0.074                          |\\n| 40       | 0.049                                             |  0.049                        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The predicted outlet pressure with the Pirani conversion(Pconv) variation during three ALD cycles shown in the figure is compared with the experimental measurements.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In particular, the outlet peak heights are captured and thus the dynamic response to the reactant pulses. This validates the coupling of the two CFD models\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the pressure reading given by the Pirani gauge is gas dependent, it will vary with the gas composition. The pressure reading is performed via the measurement of the gas thermal conductivity, which is inversely proportional to the molecular mass of the gas. As the gauge is calibrated for nitrogen, a gas with a higher molecular mass than N2(28 g/mol) will lead to a lower pressure reading, while a gas with a lower molecular mass will have the opposite effect\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":1184,"height":594}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/fig_9.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/images/fig_9.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/32/reactorGP Gakis et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"32","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1181,"height":598,"image_format":"jpeg","image_sha256":"66f20de20a119a045a9c2ef9d3cff2ee604e1e811b7b4489c7b0260709a2bc1a","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_35_fig_2.jpg","caption":"Fig. 2. Effect of substrate heat-treatment temperature on the amount of aluminum adsorbed  $\\Delta c_{\\mathrm{Al}}$  in the  $\\mathrm{AlMe}_3$  reaction (a) with alumina (reaction at  $150^{\\circ}\\mathrm{C}$ :  $(\\Delta \\Omega)$  [21]) and (b) with silica (reaction at  $150^{\\circ}\\mathrm{C}$ :  $(\\bigcirc)$  [20], reaction at  $250^{\\circ}\\mathrm{C}$ :  $(\\square)$  [19]). Lines have been fitted to the experimental data; the confidence limits represent one standard deviation. Surface OH group concentrations  $c_{(\\mathrm{O})\\mathrm{H}}(\\times , + )$  are shown for reference [20-22].","id":"train/atomic-layer-deposition/simulation-usecase/35/fig_2","sample_id":"atomic-layer-deposition/simulation-usecase/35/fig_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the effect of alumina heat-treatment temperature on both the aluminum adsorbed per cycle (Δc_Al, triangles) and the surface OH group concentration (c_(O)H, X markers). Both quantities decrease with increasing pretreatment temperature (200–800°C). The linear fit to Δc_Al yields a = 5.9 ± 0.2 nm⁻² and b = −0.0050 ± 0.0003, indicating that higher pretreatment temperatures reduce aluminum uptake.\"},{\"panel_id\":\"b\",\"text\":\"The scatter plot shows analogous data for silica substrates, with Δc_Al measured at two reaction temperatures (150°C: circles; 250°C: squares) and c_(O)H shown as + markers. The fit gives a = 4.7 ± 0.3 nm⁻² and b = −0.0033 ± 0.0005. Both Δc_Al and c_(O)H decrease with heat-treatment temperature, with c_(O)H declining more steeply at higher temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Alumina heat-treatment[°C] | ΔC<sub>Al</sub><sup>(OH)</sup>[nm<sup>-2</sup>] |\\n|---|---|\\n| 200 | 6 |\\n| 400 | 5 |\\n| 600 | 4 |\\n| 800 | 2 |\"},{\"panel_id\":\"b\",\"text\":\"| Silica heat-treatment[°C] | ΔC<sub>Si</sub><sup>(OH)</sup>[nm<sup>-2</sup>] |\\n|---|---|\\n| 200 | 7 |\\n| 400 | 6 |\\n| 600 | 5 |\\n| 800 | 3 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both Δc_Al and c_(O)H decrease with increasing heat-treatment temperature on both alumina and silica substrates. However, c_(O)H decreases more rapidly than Δc_Al, particularly at higher temperatures. At elevated pretreatment temperatures, c_(O)H can even fall below Δc_Al, indicating that the simple model where each OH group binds one aluminum atom cannot fully explain the observed growth. This divergence suggests that AlMe₃ can also adsorb through mechanisms other than ligand exchange with OH groups, such as dissociation on oxygen bridges.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Higher heat-treatment temperatures cause dehydroxylation, where pairs of surface OH groups combine to form oxygen bridges and release water, thereby reducing the concentration of reactive OH sites available for AlMe₃ chemisorption.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Circles represent Δc_Al measured at 150°C reaction temperature, squares represent Δc_Al at 250°C reaction temperature, and + markers represent the surface OH group concentration c_(O)H.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Alumina shows higher initial aluminum uptake, with a fitted intercept of a = 5.9 ± 0.2 nm⁻² compared to a = 4.7 ± 0.3 nm⁻² for silica. This difference correlates with the higher initial OH group concentration on alumina surfaces. Additionally, alumina and silica have different surface chemistries and OH group types (isolated, bridged, hydrogen-bonded), which affect their reactivity toward AlMe₃. The steeper slope for alumina (b = −0.0050) versus silica (b = −0.0033) also indicates that alumina's aluminum uptake is more sensitive to dehydroxylation.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":288,"height":230},{"panel_id":"b","x":307,"y":0,"width":271,"height":230}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/Riikka L. Puurunen.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"35","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":578,"height":230,"image_format":"jpeg","image_sha256":"5bff4bb9b7d6d4e56d7930d4b5e5c8c76d9957b202c8844161287e66c2d943ef","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_35_fig_3.jpg","caption":"Fig. 3. Effect of the OH group surface concentration  $c_{(\\mathrm{O})\\mathrm{H}}$  on the reaction of  $\\mathrm{AlMe}_3$ : (a) aluminum atoms adsorbed,  $\\Delta c_{\\mathrm{Al}}$ , (b) methyl groups adsorbed,  $\\Delta c_{\\mathrm{Me}}$ , and (c) the average  $\\mathrm{Me / Al}$  ratio in the adsorbed species  $(\\Delta c_{\\mathrm{Me}} / \\Delta c_{\\mathrm{Al}})$  (data for  $\\Delta c_{\\mathrm{Al}}$  and  $c_{(\\mathrm{O})\\mathrm{H}}$  from Figs. 1 and 2; data for  $\\Delta c_{\\mathrm{Me}}$  from references [19-21]). In panels (a) and (b), solid lines show fits to the experimental data; the confidence limits represent 1 standard deviation. The dashed line in panels (a-c) show the results expected for the chemistry of Eq. (6) (with  $\\Delta c_{\\mathrm{Me}}^{\\mathrm{max}} = 6.0 \\mathrm{nm}^{-2}$ ).","id":"train/atomic-layer-deposition/simulation-usecase/35/fig_3","sample_id":"atomic-layer-deposition/simulation-usecase/35/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows Δc_Al (aluminum adsorbed per cycle) versus surface OH concentration c_(O)H, combining data from alumina (triangles, 80–300°C) and silica (circles 150°C, squares 250°C). The solid line is a linear fit yielding a = 1.68 ± 0.09 nm⁻² and b = 0.37 ± 0.02. The dashed line represents the theoretical prediction assuming one OH binds one Al atom (Eq. 6), which clearly fails to describe the data.\"},{\"panel_id\":\"b\",\"text\":\"The scatter plot shows Δc_Me (methyl groups adsorbed) versus c_(O)H. The solid fit gives a = 4.9 ± 0.2 nm⁻² and b = 0.14 ± 0.06, indicating Δc_Me is nearly constant (~5 nm⁻²) regardless of OH concentration. The dashed theoretical line again fails to match observations.\"},{\"panel_id\":\"c\",\"text\":\"The scatter plot shows the average Me/Al ratio (Δc_Me/Δc_Al) versus c_(O)H. The ratio decreases from ~3 at low c_(O)H toward ~2 at high c_(O)H. The solid curve fits the experimental trend while the dashed line shows the inadequate theoretical prediction.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| c_(O)H, nm⁻² | Δc_Al 80-300°C alumina (nm⁻²) | Δc_Al 150°C silica (nm⁻²) | Δc_Al 250°C silica (nm⁻²) |\\n|---|---|---|---|\\n| 1 | 2.1 | - | - |\\n| 2 | 2.4 | 2.9 | - |\\n| 3 | 2.8 | - | 2.7 |\\n| 4 | 3.2 | - | 3.0 |\\n| 5 | 3.5 | - | - |\\n| 6 | 3.9 | - | - |\\n| 8 | 4.6 | - | - |\\n| 9 | 5.0 | - | - |\"},{\"panel_id\":\"b\",\"text\":\"| c_(O)H, nm⁻² | Δc_Me 80-300°C alumina (nm⁻²) | Δc_Me 150°C silica (nm⁻²) | Δc_Me 250°C silica (nm⁻²) |\\n|---|---|---|---|\\n| 1 | 5.0 | 5.2 | - |\\n| 2 | 5.2 | - | - |\\n| 3 | 5.5 | - | 5.7 |\\n| 4 | 5.6 | - | 5.5 |\\n| 5 | 5.8 | - | - |\\n| 6 | 5.9 | - | - |\\n| 8 | 6.0 | - | - |\\n| 9 | 6.1 | - | - |\"},{\"panel_id\":\"c\",\"text\":\"| c_(O)H, nm⁻² | Δc_Me/Δc_Al 80-300°C alumina | Δc_Me/Δc_Al 150°C silica | Δc_Me/Δc_Al 250°C silica |\\n|---|---|---|---|\\n| 1 | 2.4 | 1.8 | - |\\n| 2 | 2.2 | - | - |\\n| 3 | 2.0 | - | 2.1 |\\n| 4 | 1.8 | - | 1.9 |\\n| 5 | 1.7 | - | - |\\n| 6 | 1.5 | - | - |\\n| 8 | 1.4 | - | - |\\n| 9 | 1.3 | - | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The intercept represents the amount of aluminum that would adsorb on an oxide surface with zero OH groups. Since AlMe₃ retains all three methyl ligands when no OH groups are present (Me/Al = 3), this aluminum must chemisorb through dissociation or association reactions with oxygen bridges rather than ligand exchange. Multiplying a by 3 gives ~5.0 nm⁻², which matches the observed Δc_Me at low OH concentrations, confirming internal consistency of the model.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The dashed lines represent predictions from a simplified model (Eq. 6) assuming each OH group binds exactly one aluminum atom via ligand exchange. The systematic deviation from experimental data proves this simple one-to-one correspondence is invalid for the AlMe₃/H₂O ALD process.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At low c_(O)H, AlMe₃ adsorbs primarily through dissociation or association on oxygen bridges, retaining all three methyl groups (Me/Al ≈ 3). As c_(O)H increases, more aluminum atoms attach via ligand exchange reactions that release methyl groups as methane. Each OH group that reacts removes one methyl ligand, so the average number of methyl groups per adsorbed aluminum decreases. At high OH concentrations, the Me/Al ratio approaches 2, indicating that on average one methyl is released per adsorbed aluminum.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 5 to 6 methyl groups per nm² are adsorbed at saturation. This value is nearly constant because steric hindrance by the bulky methyl groups, not OH availability, determines the saturation coverage.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":732,"height":256},{"panel_id":"b","x":0,"y":288,"width":329,"height":243},{"panel_id":"c","x":356,"y":288,"width":323,"height":243}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/35/Riikka L. Puurunen.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"35","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":736,"height":531,"image_format":"jpeg","image_sha256":"cfdc893c77c72be02aeeb6360df8247ce66228f04cb0809238c915b7b0159fa9","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_36_figure_2.jpg","caption":"Figure 2. (a) WCA values as a function of ALD  $\\mathrm{Al}_2\\mathrm{O}_3$  cycles with a general recipe at  $150^{\\circ}\\mathrm{C}$  for four different substrates: ODPA/Ti, Ti,  $\\mathrm{SiO}_2$ , and HF-treated Si. The insets show WCA images captured using a three-times-magnified optical lens in a WCA analyzer. (b) Al 2p XPS spectra of  $\\mathrm{Al}_2\\mathrm{O}_3$  on ODPA/Ti for 50, 70, and 100 ALD cycles.","id":"train/atomic-layer-deposition/simulation-usecase/36/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/36/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multi spectra chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The WCA plot shows distinct initial contact angles for the four substrates, followed by progressive convergence with increasing ALD cycles. ODPA/Ti maintains a high angle through early cycles before dropping sharply between 100 and 150 cycles. HF-treated Si reaches its final angle by ~25 cycles, while SiO₂ and Ti follow nearly identical trajectories after that point. By 150 cycles, all surfaces converge near ~45°.\"},{\"panel_id\":\"b\",\"text\":\"The Al 2p spectra display a common peak centered near ~75 eV for all cycle counts. Peak intensity increases monotonically from 50 to 100 cycles, while peak position and width remain essentially unchanged. The spectra therefore differ primarily in amplitude rather than energy shift.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | ODPA/Ti (deg) | HF-treated Si (deg) | SiO₂ (deg) | Ti (deg) |\\n|-----------|----------------|----------------------|------------|----------|\\n| 0         | ~110           | ~90                  | ~70        | ~60      |\\n| 25        | ~110           | ~45                  | ~45        | ~45      |\\n| 50        | ~105           | ~45                  | ~45        | ~45      |\\n| 100       | ~90            | ~45                  | ~45        | ~45      |\\n| 150       | ~45            | ~45                  | ~45        | ~45      |\"},{\"panel_id\":\"b\",\"text\":\"| Binding energy (eV) | Relative intensity trend |\\n|---------------------|--------------------------|\\n| ~73.0               | Very low                 |\\n| ~74.0               | Rising                   |\\n| ~75.0               | Maximum (100 > 70 > 50)  |\\n| ~76.0               | Decreasing               |\\n| ~77.0               | Very low                 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Each substrate begins with a distinct contact-angle value, but the curves progressively collapse toward a common range as ALD cycles increase. ODPA/Ti remains hydrophobic through early cycles before declining sharply, whereas HF-treated Si reaches its final angle quickly and remains constant. SiO₂ and Ti show nearly identical behavior after the initial cycles. Overall, increasing ALD cycles reduce substrate-dependent differences.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ODPA/Ti.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"-All spectra peak near ~75 eV.\\n, -Peak intensity increases from 50 to 100 cycles.\\n, -Peak widths and positions remain nearly unchanged.\\n, -Differences are dominated by amplitude, not energy shift.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The stable peak position indicates that the chemical state of aluminum remains consistent as the film grows. Increasing cycle count primarily increases signal intensity rather than altering bonding environment. This suggests uniform film growth rather than a change in aluminum coordination or oxidation state.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":510,"height":384},{"panel_id":"b","x":1,"y":414,"width":506,"height":396}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/Seunggi Seo et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":513,"height":812,"image_format":"jpeg","image_sha256":"1bb47c8d3abdf4ff2623aa6afade85251904a9d172bc8f79b91971b29efb0d38","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_36_figure_5.jpg","caption":"Figure 5. WCA values as a function of the number of ALD  $\\mathrm{Al}_2\\mathrm{O}_3$  cycles for up to 150 cycles (a) using an ALD process with decreased TMA pressure (from 8.9 to 1.3 Torr) and (b) using a decreased TMA pressure process with increased Ar purging (TMA injection pressure decreased from 8.9 to 1.3 Torr) and no purging pressure increased from 1.3 to 8.9 Torr) over 370 cycles at  $150^{\\circ}\\mathrm{C}$  for three different substrates, ODPA/Ti, Ti, and  $\\mathrm{SiO}_2$  and HF-treated Si. The insets show WCA images captured using a  $3\\times$  magnified optical lens in a WCA analyzer.","id":"train/atomic-layer-deposition/simulation-usecase/36/figure_5","sample_id":"atomic-layer-deposition/simulation-usecase/36/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Under reduced TMA pressure, HF-treated Si, SiO₂, and Ti exhibit a rapid decrease in WCA to ~45° within the first 25 ALD cycles, after which their values remain constant. ODPA/Ti behaves differently, maintaining a high contact angle (~110°) through early cycles and decreasing gradually to ~90° by 150 cycles.\"},{\"panel_id\":\"b\",\"text\":\"With reduced TMA pressure and increased Ar purge, the same three substrates again converge rapidly to ~45° and remain stable over the entire 370-cycle range. ODPA/Ti shows only a slight decrease from ~110° to ~105°, indicating strong resistance to long-term hydrophilization under these conditions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD cycles | ODPA/Ti (deg) | HF-treated Si (deg) | SiO₂ (deg) | Ti (deg) |\\n|-----------|----------------|----------------------|------------|----------|\\n| 0         | ~110           | ~90                  | ~70        | ~60      |\\n| 25        | ~110           | ~45                  | ~45        | ~45      |\\n| 50        | ~110           | ~45                  | ~45        | ~45      |\\n| 100       | ~105           | ~45                  | ~45        | ~45      |\\n| 150       | ~90            | ~45                  | ~45        | ~45      |\"},{\"panel_id\":\"b\",\"text\":\"| ALD cycles | ODPA/Ti (deg) | HF-treated Si (deg) | SiO₂ (deg) | Ti (deg) |\\n|-----------|----------------|----------------------|------------|----------|\\n| 0         | ~110           | ~90                  | ~70        | ~60      |\\n| 25        | ~110           | ~45                  | ~45        | ~45      |\\n| 150       | ~105           | ~45                  | ~45        | ~45      |\\n| 370       | ~105           | ~45                  | ~45        | ~45      |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"-HF-treated Si stabilizes near ~45°.\\n-SiO₂ and Ti follow nearly identical trajectories to ~45°.\\n-ODPA/Ti remains significantly higher than the others.\\n-The separation between ODPA/Ti and the remaining substrates persists over subsequent cycles.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A fast, stable WCA response reduces uncertainty in downstream wet processing, adhesion, or interfacial layer formation. If most substrates quickly reach a consistent wetting state, process tuning becomes simpler. The persistent exception (ODPA/Ti) also highlights where selectivity might be achievable.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The curves indicate those substrates still transition rapidly to ~45° and then remain stable, even with the modified purge scheme. That suggests the early surface change is robust to this process adjustment once the low-WCA state is reached. The main visible effect is that the long-cycle behavior remains steady rather than drifting.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"ODPA/Ti.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":512,"height":381},{"panel_id":"b","x":1,"y":406,"width":515,"height":390}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/36/Seunggi Seo et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":517,"height":795,"image_format":"jpeg","image_sha256":"34d3c30b795ac4426ceaa0d6ee9125349f3b8a942c82466eec07079ab9660ecc","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_42_fig_6.jpg","caption":"Fig. 6 (a)  $t_{\\mathrm{phy}} - t_{\\mathrm{ox}}$  plot, (b)  $J - V$  plot, and (c)  $J - t_{\\mathrm{ox}}$  plot of the MIM capacitor with  $20 \\mathrm{nm}$ -thick  $\\mathrm{TiO_2}$  deposited by (black) the conventional sequence and (red) the  $\\mathrm{H}_2\\mathrm{O}$ -added sequence with a  $\\mathrm{H}_2\\mathrm{O}$  feeding time of  $10 \\mathrm{s}$ .","id":"train/atomic-layer-deposition/simulation-usecase/42/fig_6","sample_id":"atomic-layer-deposition/simulation-usecase/42/fig_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"A line chart showing the relationship between t_{hyb} [nm] and l_{0} [µm], comparing the effect of the conventional vs. the H2O-added sequence.\"},{\"panel_id\":\"b\",\"text\":\"A scatter plot displaying leakage current density against applied voltage, comparing the effect of the conventional vs. the H2O-added sequence.\"},{\"panel_id\":\"c\",\"text\":\"A scatter plot showing leakage current density at -0.8 V against t_{hyb} [nm], comparing the effect of the conventional vs. the H2O-added sequence.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| t_{hyb} [nm] | l_{0} [µm] Conventional | l_{0} [µm] H2O-added |\\n|---|---|---|\\n| 0 | 0 | 0 |\\n| 5 | 0.2 | 0.2 |\\n| 10 | 0.4 | 0.4 |\\n| 15 | 0.6 | 0.6 |\\n| 20 | 0.8 | 0.8 |\\n| 25 | 1 | 1 |\"},{\"panel_id\":\"b\",\"text\":\"| Applied Voltage [V] | Leakage current density [A/cm^{2}] Conventional | Leakage current density [A/cm^{2}] H2O-added |\\n|---|---|---|\\n|-1.5 | 10^{-6} | 10^{-7} |\\n|-1.0 | 10^{-7} | 10^{-8} |\\n|-0.5 | 10^{-8} | 10^{-8} |\\n| 0.0 | 10^{-9} | 10^{-8} |\\n| 0.5 | 10^{-6} | 10^{-6} |\\n| 1.0 | 10^{-4} | 10^{-4} |\\n| 1.5 | 10^{-3} | 10^{-3} |\"},{\"panel_id\":\"c\",\"text\":\"| t_{hyb} [nm] | Leakage current density @ -0.8 V [A/cm^{2}] |\\n|---|---|\\n| 0.2 | 10^{-7} |\\n| 0.4 | 10^{-6} |\\n| 0.6 | 10^{-5} |\\n| 0.8 | 10^{-4} |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"93.5.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"97.8.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The y-intercept reflects the interfacial properties between the TiO₂ film and the top/bottom electrodes, implying that the additional H2O step does not induce any change at the interface of the TE or the BE.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both the J–V curves and leakage current trends show that the TiO₂ films produced by the conventional and H₂O-enhanced processes behave nearly identically, indicating that their bulk electrical and interfacial properties are the same. The similar dielectric constants, unchanged y-intercepts, and overlapping leakage current behavior suggest that adding an H₂O feeding step does not negatively impact film quality or electrode interfaces, and that both films exhibit comparable crystallinity consistent with the rutile TiO₂ phase.\"}]}]","bbox":[{"panel_id":"a","x":7,"y":12,"width":199,"height":189},{"panel_id":"b","x":227,"y":6,"width":215,"height":195},{"panel_id":"c","x":453,"y":11,"width":207,"height":189}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/42/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/42/images/fig_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/42/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/42/Yeonchoo Cho et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"42","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":661,"height":200,"image_format":"jpeg","image_sha256":"20de0fccead6229318554edc62eb7449917deffa1a07866c162814606cf28ae8","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_5_9e00b88115926e47232a0241f1247464af11d2b941b12f93e9ea7d022654029e.jpg","caption":"","id":"train/atomic-layer-deposition/simulation-usecase/5/9e00b88115926e47232a0241f1247464af11d2b941b12f93e9ea7d022654029e","sample_id":"atomic-layer-deposition/simulation-usecase/5/9e00b88115926e47232a0241f1247464af11d2b941b12f93e9ea7d022654029e","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the thickness of ZnO as a function of ALD cycles at different temperatures over silicon and amorphous silicon substrates.\"},{\"panel_id\":\"b\",\"text\":\"The figure illustrates the selectivity of a process as a function of ALD cycles at two different temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD Cycles | 100°C on SiO₂ (nm) | 250°C on SiO₂ (nm) | 100°C on a-Si:H (nm) | 250°C on a-Si:H (nm) |\\n|------------|----------------------|----------------------|------------------------|------------------------|\\n| 0          | 0.0                 | 0.0                 | 0.0                   | 0.0                   |\\n| 20         | 2.0                 | 3.0                 | 0.0                   | 0.0                   |\\n| 40         | 4.0                 | 6.0                 | 0.0                   | 0.0                   |\\n| 60         | 6.0                 | 9.0                 | 0.0                   | 0.0                   |\\n| 80         | 8.0                 | 12.0                | 0.5                   | 0.0                   |\\n| 100        | 10.0                | —                   | 1.5                   | 0.2                   |\\n| 120        | —                   | —                   | 3.0                   | 0.5                   |\\n| 160        | —                   | —                   | 6.0                   | 1.0                   |\\n| 200        | —                   | —                   | 8.0                   | 2.0                   |\"},{\"panel_id\":\"b\",\"text\":\"| ALD Cycles | Selectivity @ 100°C | Selectivity @ 250°C |\\n|------------|----------------------|----------------------|\\n| 0          | 1.00                | 1.00                |\\n| 40         | 0.95                | 1.00                |\\n| 80         | 0.85                | 1.00                |\\n| 120        | 0.65                | 0.95                |\\n| 160        | 0.45                | 0.85                |\\n| 200        | 0.30                | 0.75                |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For a deposition temperature of 100 °C, the selectivity decreases from 1.0 to 0.96 at 80 ALD cycles. By further increasing the number of cycles, the selectivity drops to 0.85 (at 100 cycles), which corresponds to a ZnO thickness of ∼1 nm on the NGA. For a deposition temperature of 250 °C, the selectivity, S, remains 1 until 110 cycles and then decreases to 0.95 at 140 ALD cycles.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The selectivity  is abruptly lost at 350 °C.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"A larger nucleation delay is observed for deposition at 250 °C with a-Si:H substrate\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":14,"y":0,"width":497,"height":375},{"panel_id":"b","x":8,"y":393,"width":505,"height":413}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/images/9e00b88115926e47232a0241f1247464af11d2b941b12f93e9ea7d022654029e.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/images/9e00b88115926e47232a0241f1247464af11d2b941b12f93e9ea7d022654029e.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/Alfredo Mameli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"not_found"},"width":511,"height":809,"image_format":"jpeg","image_sha256":"870798f14a2da06899d656a16d3ec6fa9f9b1117bea4b5df3f620b8f1a2228ef","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_5_figure_2.jpg","caption":"Figure 2.  $\\mathrm{ZnO}$  film thickness measured by in situ SE as a function of the number of ALD cycles performed on substrates with a native  $\\mathrm{SiO}_2$  surface (closed squares) and with a  $\\sim 10 \\mathrm{nm}$  thick a-Si/H layer (open squares). On  $\\mathrm{SiO}_2$ , the  $\\mathrm{ZnO}$  film thickness increases linearly after a short delay of a few cycles, whereas on a-Si/H, the  $\\mathrm{ZnO}$  deposition experiences a relatively long nucleation delay. Only after 80 ALD cycles, the growth also starts on the a-Si/H and enters a linear regime at about 120 cycles. Both depositions were carried out at  $100^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-deposition/simulation-usecase/5/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/5/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the thickness of ZnO deposited on SiO2 (closed squared) and a-Si:H (open squares) substrates as a function of ALD cycles. The thickness increases linearly with increasing cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALD Cycles | ZnO Thickness on SiO₂ (nm) | ZnO Thickness on a-Si:H (nm) |\\n|------------|-----------------------------|--------------------------------|\\n| 0          | 0.0                         | 0.0                            |\\n| 10         | 2.0                         | 0.0                            |\\n| 20         | 4.0                         | 0.0                            |\\n| 30         | 6.0                         | 0.0                            |\\n| 40         | 8.0                         | 0.2                            |\\n| 50         | 10.0                        | 0.5                            |\\n| 60         | 12.0                        | 1.0                            |\\n| 70         | 14.0                        | 2.0                            |\\n| 80         | 16.0                        | 3.0                            |\\n| 90         | —                           | 4.0                            |\\n| 100        | —                           | 5.0                            |\\n| 110        | —                           | 6.0                            |\\n| 120        | —                           | 7.0                            |\\n| 130        | —                           | 8.0                            |\"}]","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\":\"On SiO2, the ZnO film thickness increases linearly after a short delay of a few cycles, whereas on a-Si/H, the ZnO deposition experiences a relatively long nucleation delay. Only after 80 ALD cycles, the growth also starts on the a-Si/H and enters a linear regime at about 120 cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Both depositions were carried out at 100 °C.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Area-selective ALD of ZnO can be obtained by exploiting the difference in nucleation delays on the two surfaces. Furthermore, the nucleation curves suggest a selectivity window (expressed as the number of ALD cycles for which no deposition is obtained on the a-Si/H) of about 80 ALD cycles before deposition occurs on both surfaces and the selectivity is lost. The key reason for the difference in nucleation delay lies in the surface chemistry because dissimilar surface groups have distinct kinetic barriers toward various thermodynamically favorable end states.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":430,"height":359}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/Alfredo Mameli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":431,"height":361,"image_format":"jpeg","image_sha256":"9358e6304379746b7f1d2c8f54fe3acc53e2d8c2412a5ebe0a231eb097cdaf19","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_5_figure_6.jpg","caption":"Figure 6. Selectivity for  $80~\\mathrm{ZnO}$  ALD cycles as a function of the deposition temperature. The insets show the top-view SEM images of patterns realized using the same EBID parameters and number of ALD cycles, but different  $\\mathrm{ZnO}$  growth temperatures (100, 150, and  $250^{\\circ}\\mathrm{C}$ ); scale bars are  $500~\\mathrm{nm}$ . The horizontal dotted line at  $S = 1$  indicates perfect selectivity.","id":"train/atomic-layer-deposition/simulation-usecase/5/figure_6","sample_id":"atomic-layer-deposition/simulation-usecase/5/figure_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"image panel"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The plot shows selectivity versus deposition temperature where selectivity increases with increasing deposition temperature.\"},{\"panel_id\":\"b\",\"text\":\"SEM Microscopy images of samples deposited at different temperatures demonstrating selectivity\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Deposition temperature (°C) | Selectivity |\\n|---|---|\\n| 100 | 0.82 |\\n| 150 | 0.96 |\\n| 200 | 0.99 |\\n| 250 | 1.00 |\"},{\"panel_id\":\"b\",\"text\":\"| 100 °C | 150 °C | 250 °C |\\n|---|---|---|\\n| ![](100°C_image.png) | ![](150°C_image.png) | ![](250°C_image.png) |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"the selectivity was calculated using the ZnO surface coverage in and outside the patterned area, as measured by top-view SEM.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The SEM-measured selectivity data confirmed the results obtained using SE: the selectivity increases with the deposition temperature up to 250 °C\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The selectivity increases as follows\\na.  0.83 at 100 °C\\nb. 0.97 at 150 °C\\nc. 0.99 at 200  °C\\nd. 0.99 at  250 °C\"}]}]","bbox":[{"panel_id":"b","x":5,"y":308,"width":551,"height":152},{"panel_id":"a","x":2,"y":2,"width":552,"height":306}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/5/Alfredo Mameli et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"5","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":556,"height":461,"image_format":"jpeg","image_sha256":"f50b66641f2a179ee0915cd091af838eccaf9caf2962e830ffecd844c1839f08","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_2.jpg","caption":"Figure 2. Deposition rate versus exposure of  $\\mathrm{SiO}_2$  (red diamonds) and  $\\mathrm{Si}_3\\mathrm{N}_4$  (green circles) films deposited from chloro-silane precursors by ALD at  $450^{\\circ}\\mathrm{C}$  with less than  $2\\%$  absolute variation on per cycle deposition rate.","id":"train/atomic-layer-deposition/simulation-usecase/53/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/53/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the per cycle deposition rate of SiO2 and Si3N4 as a function of exposure time in Torr·s.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Exposure (Torr·s) | Per cycle Dep. Rate (Å) SiO₂ | Per cycle Dep. Rate (Å) Si₃N₄ |\\n|--------------------|-------------------------------|---------------------------------|\\n| 0.02              | 0.4                           | -                               |\\n| 0.05              | 0.8                           | -                               |\\n| 0.10              | 1.0                           | -                               |\\n| 1.0               | -                             | 0.5                             |\\n| 3.0               | -                             | 0.9                             |\\n| 10.0              | -                             | 1.1                             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Films deposited from chloro-silane precursors by ALD at 450 °C with less than 2% absolute variation on per cycle deposition rate.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure compares the relative reactivity for simple chloro-silane precursors (e.g., DCS,HCDS, etc.)\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Precursor exposure was varied by changing the exposure time, the exposure partial pressure, or both. Deposition rate was determined by measurement of the thickness after deposition and dividing by the number of ALD cycles performed. The very long precursor exposure for deposition of silicon nitride makes this process economically unviable, due to both the excessive throughput time per film deposited and the unacceptably high volume of silicon precursors consumed.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":2,"width":592,"height":456}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/53/images/figure_2.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 scatter plot","caption_source":"content.json:image_caption"},"width":594,"height":461,"image_format":"jpeg","image_sha256":"463616b24ad45d70701af9a4941c96b51fe4f54e8c584aa1d2ae08648402517e","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_6_figure_2.jpg","caption":"Figure 2. Duo-linear plot of  $\\mathrm{TiO_2}$  film thickness versus the number of cycles and a silicon reference (reprinted with permission from [6]).","id":"train/atomic-layer-deposition/simulation-usecase/6/figure_2","sample_id":"atomic-layer-deposition/simulation-usecase/6/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the thickness of TiO2 ALD deposited on PMMA and silicon substrates as a function of the number of cycles, showing cycle dependent GPCs.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of cycles | Thickness (Å) on PMMA |Thickness (Å) on silicon |\\n|------------------|---------------|---------------|\\n| 0               | 0             |0|\\n| 100             | 150           |75|\\n| 150             | 250           |100|\\n| 200             | 300           |120|\\n| 250             | 350           |150|\\n| 300             | 360           |160|\\n| 400             | 400           |200|\\n| 500             | 450           |250|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It could be that PMMA as a substrate enables enhanced growth, whereas TiO2 growth on TiO2 does not enable this.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SiO2 looks more like TiO2 in structure than PMMA does so it is likely that the growth on SiO2 looks more like the growth on TiO2 than on PMMA.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"After these number of cycles the substrate does not have an influence on the growth anymore and just the TiO2 acts as a starting substrate. As this is the same for both cases it is likely that the same GPC is acquired.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The GPC is 0.74 angstrom/cycle.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":694,"height":470}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/6/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/6/images/figure_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/6/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/6/Mina Shahmohammadi et al.pdf"],"main_category":"atomic-layer-deposition","sub_category":"simulation-usecase","paper_id":"6","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":695,"height":472,"image_format":"jpeg","image_sha256":"98c70cc45ed040ffd673f42915a6dead412f09ab62bc934c719fa8d9184c3bcc","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_11.jpg","caption":"Figure 11. Simulated GPC data, estimated from both the thickness of the simulated thin film and from the deposited mass, compared to the growth rate reported in ref 26. The GPC estimated from the average mass includes the error bars, obtained as a standard deviation from series of simulations.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_11","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_11","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between temperature and growth-per-cycle in angstroms. Data points represent measurements from Yousfi et al., kMC (average mass), and kMC (average thickness). The trend shows increasing GPC with increasing temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature / °C | Growth-per-cycle / angstrom Yousfi|Growth-per-cycle / angstrom Model|\\n|---|---|---|\\n| 0 | 0.0 |0.0 |\\n| 25 | 0.0 |0.0 |\\n| 50 | 0.25 |1.3|\\n| 75 | 0.6|2.2 |\\n| 100 |1.3| 2.6 |\\n| 125 | 2.0|2.6 |\\n| 150 | 2.1|2.6 |\\n| 175 | 2.1|2.6 |\\n| 200 | 2.0|- |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No they do not overlap in max GPC, however, the trend is similar. In addition, both methods result in an optimal deposition temperature of about 175 degrees celcius.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For too high temperatures, there can be desorption of surface species, this can result in lower GPC.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"That the temperature of the process is always higher than the bubbler temperature, such that condensation does not occur.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":613,"height":492}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_11.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 scatter plot","caption_source":"content.json:img_caption"},"width":614,"height":497,"image_format":"jpeg","image_sha256":"cf2894806b97d0c0f1b7ca4aff2aa75c4302585331c8cb2b7eca4d4c2ffc0724","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_12.jpg","caption":"Figure 12. GPC as a function of temperature in the case where the barriers for the second ligand elimination reaction (LE2) have been increased by  $0.10\\mathrm{eV}$ . The elimination of the MEZ from the surface is clearly the crucial step in the deposition process. Increasing the barrier shifts the observed behavior to higher temperature as the ligand-elimination reaction can occur only at elevated temperatures.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_12","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_12","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between temperature and growth-per-cycle in angstroms. This is shown for experimental data, a kMC model, and an additional model with different input variables.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature / °C | Growth-per-cycle / angstrom Yousfi|Growth-per-cycle / angstrom Model|Growth-per-cycle / angstrom Model LE2 +0.1 eV|\\n|---|---|---|---|\\n| 0 | 0.0 |0.0 |0.0 |\\n| 25 | 0.0 |0.0 |0.0 |\\n| 50 | 0.25 |1.3|0.0 |\\n| 75 | 0.6|2.2 |0.6|\\n| 100 |1.3| 2.6 |1.8|\\n| 125 | 2.0|2.6 |2.4|\\n| 150 | 2.1|2.6 |2.5|\\n| 175 | 2.1|2.6 |2.7|\\n| 200 | 2.0|- |-|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes, this is especially the case for the lower temperatures where there seems to be a short delay or thresshold temperature.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This needs to be overcome by added energy via increased temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"2.3 angstrom per cycle\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This is not yet visible in the graph for this temperature range, but if the model behaves according to the actual physical situation this should be the case.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":608,"height":497}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_12.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 scatter plot","caption_source":"content.json:img_caption"},"width":613,"height":500,"image_format":"jpeg","image_sha256":"4a25a24c72f9493d94b95848ba38212b35d60ee654b88ee6897d47ad32d1b117","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_13.jpg","caption":"Figure 13. Fraction of persisting ligands as a function of temperature in the simulation. Experimental data, obtained from ref 12, are measured for low and high water exposures. Low and high water exposures correspond to  $10^{4}$  and  $10^{7} \\mathrm{~L}$ , respectively.","id":"train/atomic-layer-deposition/simulation-usecase/8/figure_13","sample_id":"atomic-layer-deposition/simulation-usecase/8/figure_13","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the fraction of ethyl-ligands persisting at various temperatures. The simulated data points are represented by red squares, while the experimental data points are shown by gray triangles. The graph shows the fraction of ethyl-ligands persisting decreasing with increasing temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature / °C | Fraction of ethylligands persisting Simulated |Fraction of ethylligands persisting Long exposure |Fraction of ethylligands persisting short exposure |\\n|---|---|---|---|\\n| 25 | 0.83 |-|-|\\n| 50 | 0.6 |-|-|\\n| 75 | 0.5 |-|-|\\n| 100 | 0.45 |0.3|0.75|\\n| 120 | 0.45 |-|-|\\n| 150 | 0.45 |0.15|0.55|\\n| 180 | 0.45 |-|-|\\n| 200 | - |0.05|0.12|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both for long and short water exposures the experimental data does not overlap with the simulated data indicating that the model is incomplete.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Water is supposed to react with the ligands and remove them. If the exposure is shorter, it could be that there are less ligand removing reactions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The fraction is close to one, this means that the ligands are not removed and therefore also not replaced with reactive surface sites to which new precursor molecules can adsorb. This means it is unlikely there will be any substantial growth at this temperature.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The layers grown with ALD are usually used in devices where carbon from ligands can degrade the material properties making them incompatible with the application of interest.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":619,"height":516}],"source":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-deposition/simulation-usecase/8/images/figure_13.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 scatter plot","caption_source":"content.json:img_caption"},"width":620,"height":514,"image_format":"jpeg","image_sha256":"5607376bc51e7f7ae661ad7e3de4848aab024b58874a7ed58415c791967f5fdb","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_14_figure_14.jpg","caption":"Figure 14. W ALE using  $\\mathrm{O}_2 / \\mathrm{O}_2$ ,  $\\mathrm{BCl}_3$ , and HF as reactants at  $207^{\\circ}\\mathrm{C}$ . (a)  $\\mathrm{WO}_3$  thickness increases after  $\\mathrm{O}_2 / \\mathrm{O}_3$  exposure and decreases after  $\\mathrm{BCl}_3 / \\mathrm{HF}$  exposure. (b) W etching occurs with an etch rate of  $2.44\\mathrm{\\AA}/$  cycle. Reproduced with permission from ref 32. Copyright 2017 American Chemical Society.","id":"train/atomic-layer-etching/experimental-usecase/14/figure_14","sample_id":"atomic-layer-etching/experimental-usecase/14/figure_14","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the thickness of WO₃ after exposure to O₂/O₃ and BCl₃/HF as a function of the number of half-cycles.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the thickness of W as a function of the number of half-cycles at 207 °C, with O2/O3, BCl3 and HF exposure. The figure shows an etch rate of 2.44 Å/cycle.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Half-Cycles | WO₃ Thickness (Å) After O₂/O₃ | WO₃ Thickness (Å) After BCl₃/HF |\\n| --- | --- | --- |\\n| 0 | 38 | 27 |\\n| 10 | 43 | 35 |\\n| 20 | 37 | 30 |\\n| 30 | 33 | 25 |\\n| 40 | 30 | 23 |\\n| 50 | 26 | 20 |\\n| 60 | 26 | 17 |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Half-Cycles | W Thickness (Å) |\\n|---|---|\\n| 0 | 230 |\\n| 10 | 215 |\\n| 20 | 200 |\\n| 30 | 190 |\\n| 40 | 175 |\\n| 50 | 170 |\\n| 60 | 155 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The experiment was performed at 207 °C with the following exposure: 3150 mTorr s of O2/O3, 329 mTorr s of BCl3, and 2800 mTorr s of HF.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The W thickness decreases as the number of half-cycles increases.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seenin the figure, the etch per cycle is 2.44 Å/cycle.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The thickness of the WO3 increases after exposure to O2/O3 because it can oxidize the surface. However, it decreases after exposure to BCl3/HF.\"}]}]","bbox":[{"panel_id":"b","x":1,"y":397,"width":642,"height":482},{"panel_id":"a","x":4,"y":0,"width":643,"height":417}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/Mechanisms of Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"14","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":653,"height":873,"image_format":"jpeg","image_sha256":"b9c9f373c395dfe158457970207748a2347e9f94ee84de6a9bb55b5fa67b53f5","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_14_figure_5.jpg","caption":"Figure 5. Film thickness versus number of ALE cycles for various materials using HF as fluorine reactant and (a) TMA or (b) DMAF as metal precursor for ligand-exchange. Reproduced with permission from ref 23. Copyright 2016 American Chemical Society.","id":"train/atomic-layer-etching/experimental-usecase/14/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/14/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows how the film thickness of various materials (TiN, ZrO₂, SiO₂, Si₃N₄, HfO₂, Al₂O₃) changes over cycles at 300°C using TMA.\"},{\"panel_id\":\"b\",\"text\":\"The chart shows how the film thickness of various materials (TiN, SiO₂, Si₃N₄, ZrO₂, HfO₂, Al₂O₃) changes over cycles at 250°C using DMAC.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | TiN | ZrO₂ | SiO₂ | Si₃N₄ | HfO₂ | Al₂O₃ |\\n|---|---|---|---|---|---|---|\\n| 0 | 68 | 50 | 48 | 40 | 55 | 44 |\\n| 50 | - | 48 | - | - | 50 | 20 |\\n| 100 | - | 53 | - | - | 43 | 0 |\\n| 200 | 70 | 48 | 43 | 42 | 37 | 0 |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Cycles | TiN | SiO₂ | Si₃N₄ | ZrO₂ | HfO₂ | Al₂O₃ |\\n|---|---|---|---|---|---|---|\\n| 0 | 70 | 50 | 40 | 50 | 50 | 48 |\\n| 25 | 70 | 51 | 40 | 22 | 40 | 40 |\\n| 50 | 70 | 50 | 40 | 5 | 20 | 38 |\\n| 100 | 70 | 50 | 40 | 0 | 0 | 18 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Only Al2O3, HfO2, and ZrO2 exhibit etching behavior. For the other materials, the film thickness remains constante over the number of cycles.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN has the highest film thickness values.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The TMA ALE cycle was performed at 300 °C.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen in the figure, Al2O3 exhibit the highest etch rates using TMA as reactant.\"}]}]","bbox":[{"panel_id":"b","x":49,"y":539,"width":604,"height":506},{"panel_id":"a","x":17,"y":5,"width":630,"height":526}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/Mechanisms of Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"14","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":658,"height":1072,"image_format":"jpeg","image_sha256":"4b524612e6fdd367954e4901d6b187e0eb9a9ce7c4819377844aab951c454e7b","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_14_figure_6.jpg","caption":"Figure 6. Thickness change versus number of ALE cycles for crystalline and amorphous  $\\mathrm{HfO}_2$  using HF and  $\\mathrm{TiCl}_4$  as reactants. Reproduced with permission from ref 29. Copyright 2020 American Vacuum Society.","id":"train/atomic-layer-etching/experimental-usecase/14/figure_6","sample_id":"atomic-layer-etching/experimental-usecase/14/figure_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the thickness change of crystalline and amorphous HfO₂ as a function of the number of ALE cycles at 250°C using HF/TiCl₄. The etch rate is 0.02 Å/cycle for the crystalline sample and 0.35 Å/cycle for the amorphous.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of ALE Cycles | crystalline | amorphous |\\n| --- | --- | --- |\\n| 0 | 0 | 0 |\\n| 50 | 0 | -18 |\\n| 100 | 0 | -30 |\\n| 150 | -2 | -45 |\\n| 200 | -3 | -80 |\\n| 300 | -5 | -95 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At 250 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"In both cases the film thickness decreases, but the change is more significant for the amorphous HfO2 film.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"HF and TiCl4 were used as reactants.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The etch rate is 0.02 Å/cycle for the crystalline film and for the amorphous film: 0.36 Å/\"}]}]","bbox":[{"panel_id":"a","x":22,"y":2,"width":629,"height":488}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/images/figure_6.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/14/Mechanisms of Thermal Atomic Layer Etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"14","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":658,"height":500,"image_format":"jpeg","image_sha256":"7ffdad84762b9db9d15b5ce4b9f3435a62cb9b4c134537ccda6314668c9fd3c6","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_17_figure_7.jpg","caption":"Figure 7. Temperature dependence of  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE using  $\\mathrm{SF}_4$  and  $\\mathrm{Sn(acac)}_2$ . Etch rate increases from  $0.04 \\mathrm{Å / cycle}$  at  $150^{\\circ}\\mathrm{C}$  to  $0.25 \\mathrm{Å / cycle}$  at  $225^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/17/figure_7","sample_id":"atomic-layer-etching/experimental-usecase/17/figure_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure presents the film thickness of Al₂O₃ as a function of the number of atomic layer etching (ALE) cycles at four different temperatures: 150 °C, 175 °C, 200 °C, and 225 °C. Each dataset shows a monotonic decrease in thickness with increasing cycles, but the rate of etching significantly depends on temperature. At 150 °C, the change is minimal, indicating very slow etching, while the steepest decrease is observed at 200 °C and 225 °C, demonstrating a higher etch rate. The linearity in most cases suggests a consistent per-cycle removal, particularly at elevated temperatures.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | Thickness at 150 °C (Å) | Thickness at 175 °C (Å) | Thickness at 200 °C (Å) | Thickness at 225 °C (Å) |\\n|------------------|--------------------------|--------------------------|--------------------------|--------------------------|\\n| 0                | 190                      | 190                      | 190                      | 190                      |\\n| 50               | 189                      | 187                      | 182                      | 180                      |\\n| 100              | 188                      | 184                      | 174                      | 170                      |\\n| 200              | 187                      | 179                      | 160                      | 155                      |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Al₂O₃ thickness consistently decreases with increasing ALE cycles at all temperatures, but the rate of this decrease varies. At 150 °C, the reduction in thickness is minimal, indicating limited etching. In contrast, at 200 °C and 225 °C, the film is etched more aggressively, with a sharp decline in thickness over 200 cycles. This suggests that higher temperatures enhance the etch efficiency, likely by improving precursor reactivity or facilitating the removal of reaction byproducts.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"150 °C.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Increased SF₄ and Sn(acac)₂ reactivity at higher temperatures, enhanced byproduct desorption, improved surface diffusion.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":672,"height":547}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/images/figure_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/17/SF4 as the Fluorination Reactant for Al2O3 and VO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":547,"image_format":"jpeg","image_sha256":"db4540cfe0dc3c8695a08295734287de8b6efe2e0e6214f581b6b6676edc8843","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_2_figure_10.jpg","caption":"Figure 10. Review of literature data of physical sputter rates reported for Si versus the square-root of Ar ion energy up to energies of  $400\\mathrm{eV}^{95,118 - 132}$  along with SRIM simulation results. The threshold for physical sputtering of Si is  $\\approx 20\\mathrm{eV}$ . The large scatter with regard to the threshold energy of physical sputtering for Si shows the difficulty of controlling experimental conditions.","id":"train/atomic-layer-etching/experimental-usecase/2/figure_10","sample_id":"atomic-layer-etching/experimental-usecase/2/figure_10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure presents a comparative scatter plot of experimentally reported physical sputter yields of silicon (Si) as a function of the square root of Ar⁺ ion energy, compiled from multiple literature sources. The wide dispersion of data points highlights strong experiment-to-experiment variability, particularly near the sputtering threshold. A SRIM simulation curve is included as a reference, indicating an approximate threshold energy of ~20 eV for Si sputtering. The figure emphasizes the inherent difficulty in achieving precise control in physical sputtering processes.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Energy [eV] | Square Root of Ion Energy [eV^1/2] | Sputter Yield [Si/Ar^+] |\\n|---|---|---|\\n| 0 | 0 | 0 |\\n| 25 | 5 | 0.05 |\\n| 100 | 10 | 0.1 |\\n| 225 | 15 | 0.2 |\\n| 400 | 20 | 0.3 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This comparison highlights the sensitivity of physical sputtering to ion energy, especially near the threshold region. In ALE, precise control of ion energy is essential to remain within the ALE window, avoiding uncontrolled sputtering. The observed scatter emphasizes the need for carefully tuned plasma and ion conditions to achieve repeatable, layer-by-layer etching.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~20 eV, as suggested by both experimental data and SRIM simulations.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. Higher ion energies generally lead to increased physical sputtering of Si.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Variations in ion energy calibration, Differences in surface condition and crystallinity, Experimental setup and measurement uncertainty, Ion incidence angle and beam conditions\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":575,"height":603}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/images/figure_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/Atomic Layer Etching at the Tipping Point An Overview.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"2","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":575,"height":603,"image_format":"jpeg","image_sha256":"f97b84f2943eb3ef0be66fcfe4cd282f2f0c34854630766a1e82df7125df6729","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_2_figure_11.jpg","caption":"Figure 11. Review of literature data of physical sputter rates for  $\\mathrm{SiO_2}$  versus the square-root of Ar ion energy up to energies of  $400\\mathrm{eV}^{95,122,124,128,133 - 137}$  The threshold for physical sputtering of  $\\mathrm{SiO_2}$  is  $\\approx 45\\mathrm{eV}$ .","id":"train/atomic-layer-etching/experimental-usecase/2/figure_11","sample_id":"atomic-layer-etching/experimental-usecase/2/figure_11","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure compiles experimental literature data and SRIM simulation results showing the physical sputter yield of SiO₂ as a function of the square root of Ar⁺ ion energy. The data reveal a clear sputtering threshold around ~45 eV, above which the sputter yield increases approximately linearly with ion energy. The wide scatter among experimental points highlights strong sensitivity to experimental conditions such as surface state, angle of incidence, and ion flux, emphasizing challenges in precise sputter control at low energies.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Energy [eV] | Sputter Yield [SiO₂/Ar⁺] |\\n|---|---|\\n| 0 | 0.0 |\\n| 5 | 0.0 |\\n| 10 | 0.0 |\\n| 15 | 0.0 |\\n| 20 | 0.0 |\\n| 25 | 0.0 |\\n| 30 | 0.0 |\\n| 35 | 0.0 |\\n| 40 | 0.0 |\\n| 45 | 0.0 |\\n| 50 | 0.0 |\\n| 55 | 0.0 |\\n| 60 | 0.0 |\\n| 65 | 0.0 |\\n| 70 | 0.0 |\\n| 75 | 0.0 |\\n| 80 | 0.0 |\\n| 85 | 0.0 |\\n| 90 | 0.0 |\\n| 95 | 0.0 |\\n| 100 | 0.0 |\\n| 105 | 0.0 |\\n| 110 | 0.0 |\\n| 115 | 0.0 |\\n| 120 | 0.0 |\\n| 125 | 0.0 |\\n| 130 | 0.0 |\\n| 135 | 0.0 |\\n| 140 | 0.0 |\\n| 145 | 0.0 |\\n| 150 | 0.0 |\\n| 155 | 0.0 |\\n| 160 | 0.0 |\\n| 165 | 0.0 |\\n| 170 | 0.0 |\\n| 175 | 0.0 |\\n| 180 | 0.0 |\\n| 185 | 0.0 |\\n| 190 | 0.0 |\\n| 195 | 0.0 |\\n| 200 | 0.0 |\\n| 205 | 0.0 |\\n| 210 | 0.0 |\\n| 215 | 0.0 |\\n| 220 | 0.0 |\\n| 225 | 0.0 |\\n| 230 | 0.0 |\\n| 235 | 0.0 |\\n| 240 | 0.0 |\\n| 245 | 0.0 |\\n| 250 | 0.0 |\\n| 255 | 0.0 |\\n| 260 | 0.0 |\\n| 265 | 0.0 |\\n| 270 | 0.0 |\\n| 275 | 0.0 |\\n| 280 | 0.0 |\\n| 285 | 0.0 |\\n| 290 | 0.0 |\\n| 295 | 0.0 |\\n| 300 | 0.0 |\\n| 305 | 0.0 |\\n| 310 | 0.0 |\\n| 315 | 0.0 |\\n| 320 | 0.0 |\\n| 325 | 0.0 |\\n| 330 | 0.0 |\\n| 335 | 0.0 |\\n| 340 | 0.0 |\\n| 345 | 0.0 |\\n| 350 | 0.0 |\\n| 355 | 0.0 |\\n| 360 | 0.0 |\\n| 365 | 0.0 |\\n| 370 | 0.0 |\\n| 375 | 0.0 |\\n| 380 | 0.0 |\\n| 385 | 0.0 |\\n| 390 | 0.0 |\\n| 395 | 0.0 |\\n| 400 | 0.0 |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Understanding the sputtering threshold and energy-dependent yield is critical for ALE because it defines the narrow ion-energy window where material removal is controllable and self-limiting. Operating below the physical sputtering regime enables selective, damage-minimized etching, while exceeding it risks uncontrolled material loss and reduced process selectivity.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The sputter yield of SiO₂ increases approximately linearly with the square root of ion energy once the sputtering threshold is exceeded.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The sputter yield remains near zero at low ion energies and increases only after exceeding a threshold of approximately 45 eV.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Surface condition of SiO₂, Ion incidence angle, Ion energy distribution, Experimental setup variations, Measurement methodology differences\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":572,"height":600}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/images/figure_11.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/images/figure_11.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/2/Atomic Layer Etching at the Tipping Point An Overview.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"2","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":572,"height":600,"image_format":"jpeg","image_sha256":"af1e24a90182f1b29ed4f1ee1ade03091899201c0f4f4b905dde823f3aee6061","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_22_figure_5.jpg","caption":"Figure 5. Thickness measurements were acquired after etching of amorphous as-deposited ALD  $\\mathrm{MoS}_2$  films on  $\\mathrm{SiO}_2$  coupons at  $200^{\\circ}\\mathrm{C}$ . Both SE and XRR thickness measurements show a linear decrease in film thickness. A linear fit of the data produces an etching rate of 0.05 nm/cycle for amorphous  $\\mathrm{MoS}_2$ .","id":"train/atomic-layer-etching/experimental-usecase/22/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/22/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Panel (a) shows no change in thickness when only Hhfac is used, even after 20 pulses.\"},{\"panel_id\":\"b\",\"text\":\"Similarly, panel (b) indicates no measurable thickness reduction with H₂ plasma alone.\"},{\"panel_id\":\"c\",\"text\":\"In contrast, panel (c) demonstrates a steady linear decrease in film thickness when both Hhfac and H₂ plasma are applied in combination, highlighting the synergistic effect of this treatment for etching. These results suggest that neither precursor nor plasma alone is sufficient to etch the surface, but their combination results in controlled material removal.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of pulses/cycles | Thickness change (nm) |\\n|--------------------------|------------------------|\\n| 0                        | 0                      |\\n| 5                        | 0                      |\\n| 10                       | 0                      |\\n| 15                       | 0                      |\\n| 20                       | 0                      |\"},{\"panel_id\":\"b\",\"text\":\"| Number of pulses/cycles | Thickness change (nm) |\\n|--------------------------|------------------------|\\n| 0                        | 0                      |\\n| 5                        | 0                      |\\n| 10                       | 0                      |\\n| 15                       | 0                      |\\n| 20                       | 0                      |\"},{\"panel_id\":\"c\",\"text\":\"| Number of pulses/cycles | Thickness change (nm) |\\n|--------------------------|------------------------|\\n| 0                        | 0                      |\\n| 5                        | -0.5                   |\\n| 10                       | -1.0                   |\\n| 15                       | -1.5                   |\\n| 20                       | -2.0                   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The thickness decreases by approximately 2.0 nm after 20 cycles.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"(a): Hhfac, no thickness change, (b): H₂ plasma, no thickness change , (c): Hhfac + H₂ plasma, linear decrease in thickness\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In atomic layer etching, precursor adsorption alone often does not result in material removal due to the lack of activation energy. Similarly, plasma exposure alone may not modify the surface sufficiently for etching to occur. The combination of Hhfac and H₂ plasma enables a two-step mechanism where Hhfac modifies the surface chemically, and the H₂ plasma provides energetic species to drive desorption or reaction of surface-bound fragments.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":483,"height":389}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/images/figure_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/22/Thermal Atomic Layer Etching of MoS2 Using MoF6 and H2O.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"22","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":483,"height":389,"image_format":"jpeg","image_sha256":"5051dfebc7e58adcabf584de46f39386c987bd7e00dd2c11f7a3988ff58373a5","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_4.jpg","caption":"Figure 4. Expansion showing thickness of  $\\mathrm{ZnS}$  film during the first 5  $\\mathrm{ZnS}$  ALE cycles in Figure 3.","id":"train/atomic-layer-etching/experimental-usecase/23/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows an expanded view of the ZnS thickness evolution over the first few ALE cycles. It plots the ZnS film thickness as a function of ZnS ALE cycle number using HF and TMA as reactants at 300 °C, focusing only on the 0–5 cycle region. The initial ZnS films used for ALE were deposited by ALD at 100 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALE Cycles | ZnS Thickness (Å) | Stage                    | \\n|------------|--------------------|-----------------------------|\\n| 0          | 203             | Initial ZnS thickness       |                              \\n| 0          |   -                 | After HF Exposure           |                              \\n| 0          |     -               | After TMA Exposure          |                              \\n| 0.5        |     -               | Initial ZnS thickness       |                              \\n| 0.5        | 203.52      | After HF Exposure        |                             \\n| 0.5        |       -             | After TMA Exposure          |                              \\n| 1          |         -           | Initial ZnS thickness       |                              \\n| 1          |         -           | After HF Exposure           |                              \\n| 1          | 203.62        | After TMA Exposure       |                              \\n| 1.5        |       -             | Initial ZnS thickness       |                              \\n| 1.5        | 201.9         | After HF Exposure           |                              \\n| 1.5        |       -             | After TMA Exposure          |                              \\n| 2          |         -           | Initial ZnS thickness       |                              \\n| 2          |         -           | After HF Exposure           |                              \\n| 2          | 201.58        | After TMA Exposure       |                              \\n| 2.5        |        -            | Initial ZnS thickness       |                              \\n| 2.5        | 200             | After HF Exposure           |                              \\n| 2.5        |         -           | After TMA Exposure          |                              \\n| 3          |        -            | Initial ZnS thickness       |                              \\n| 3          |          -          | After HF Exposure           |                              \\n| 3          | 199.4          | After TMA Exposure       |                             \\n| 3.5        |           -         | Initial ZnS thickness       |                              \\n| 3.5        | 198             | After HF Exposure           |                              \\n| 3.5        |        -            | After TMA Exposure          |                              \\n| 4          |          -          | Initial ZnS thickness       |                              \\n| 4          |          -          | After HF Exposure           |                              \\n| 4          | 197.25        | After TMA Exposure      |                              \\n| 4.5        |        -            | Initial ZnS thickness       |                              \\n| 4.5        | 196             | After HF Exposure           |                              \\n| 4.5        |      -              | After TMA Exposure          |                              \\n| 5          |        -            | Initial ZnS thickness       |                              \\n| 5          |         -           | After HF Exposure           |                              \\n| 5          | 195.1            | After TMA Exposure          |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1) HF Exposure: Fluorinates/modifies the ZnS surface.\\n\\n2) Thickness Measurement.\\n\\n3) TMA Exposure: Removes the modified layer.\\n\\n4) Thickness Measurement. Measuring after each step verifies the self-limiting nature of each half-reaction and quantifies the individual contribution to the total etch per cycle.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"HF/TMA ALE: Pro: Chemically selective to ZnS; SiO₂ is resistant to HF under these conditions.\\n\\nThermal Decomposition: Con: Non-selective; relies on volatility, likely attacking both ZnS and SiO₂.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The larger drop after HF indicates it causes the major physical removal or volumetric change, likely by converting crystalline ZnS into a porous, volatile, or low-density zinc fluoride (ZnF₂) compound. The subsequent TMA step then cleans up this modified layer, causing a smaller but essential thickness change to complete the cycle.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This enables digital etch trimming of quantum dot layers with sub-nanometer precision. Engineers can remove exact numbers of monolayers to tune quantum confinement and emission wavelength after growth, or to expose a specific subsurface dot layer with minimal damage to adjacent layers.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":2,"width":666,"height":536}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_4.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 scatter plot","caption_source":"content.json:img_caption"},"width":670,"height":539,"image_format":"jpeg","image_sha256":"5e0df4a3a6d2e6e820904418c60ac7e577bde4bd21ccd2de756ab7edcede1e35","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_5.jpg","caption":"Figure 5. Thickness change of  $\\mathrm{ZnS}$  film vs number of  $\\mathrm{ZnS}$  ALE cycles at 225, 250, 275, and  $300^{\\circ}\\mathrm{C}$ . Initial  $\\mathrm{ZnS}$  ALE films were grown at  $100^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/23/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/23/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The multiple line chart shows the reduction in ZnS film thickness with increasing ALE cycles (0–20) at temperatures from 225 °C to 300 °C for films initially deposited at 100 °C. All values display a linear decrease in thickness, confirming a constant etch per cycle and a well-controlled, self-limiting ALE process. The slope increases with temperature, indicating thermally activated etching, with the highest etch rate at 300 °C (~2.1 Å/cycle) and the lowest at 225 °C (~1.25 Å/cycle).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of ALE Cycles | Change in ZnS Thickness (Å)  | Temperature |\\n|------------|------------------|-------------|\\n| 0          | 0                | 225 °C      |\\n| 0          | 0                | 250 °C      |\\n| 0          | 0                | 275 °C      |\\n| 0          | 0                | 300 °C      |\\n| 2          | -1.23            | 225 °C      |\\n| 2          | -1.86            | 250 °C      |\\n| 2          | -2.35            | 275 °C      |\\n| 2          | -3.59            | 300 °C      |\\n| 4          | -3.59            | 225 °C      |\\n| 4          | -5.20            | 250 °C      |\\n| 4          | -6.56            | 275 °C      |\\n| 4          | -7.80            | 300 °C      |\\n| 6          | -6.68            | 225 °C      |\\n| 6          | -8.91            | 250 °C      |\\n| 6          | -10.77           | 275 °C      |\\n| 6          | -11.88           | 300 °C      |\\n| 8          | -10.27           | 225 °C      |\\n| 8          | -12.25           | 250 °C      |\\n| 8          | -14.72           | 275 °C      |\\n| 8          | -16.71           | 300 °C      |\\n| 10         | -12.37           | 225 °C      |\\n| 10         | -15.84           | 250 °C      |\\n| 10         | -19.31           | 275 °C      |\\n| 10         | -20.79           | 300 °C      |\\n| 12         | -15.84           | 225 °C      |\\n| 12         | -19.18           | 250 °C      |\\n| 12         | -23.02           | 275 °C      |\\n| 12         | -24.87           | 300 °C      |\\n| 14         | -18.56           | 225 °C      |\\n| 14         | -23.02           | 250 °C      |\\n| 14         | -27.47           | 275 °C      |\\n| 14         | -29.94           | 300 °C      |\\n| 16         | -21.16           | 225 °C      |\\n| 16         | -27.22           | 250 °C      |\\n| 16         | -32.04           | 275 °C      |\\n| 16         | -34.27           | 300 °C      |\\n| 18         | -23.02           | 225 °C      |\\n| 18         | -30.93           | 250 °C      |\\n| 18         | -35.76           | 275 °C      |\\n| 18         | -38.24           | 300 °C      |\\n| 20         | -26.48           | 225 °C      |\\n| 20         | -34.27           | 250 °C      |\\n| 20         | -40.46           | 275 °C      |\\n| 20         | -43.18           | 300 °C      |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The process is thermally activated. Increasing the temperature from 225°C to 300°C nearly doubles the material removal: at 225°C, roughly -25 Å is removed after 20 cycles, whereas at 300°C, roughly -42 Å is removed.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"300°C Process: Advantage: Faster, higher-throughput etch. Disadvantage : Risk of thermal damage to the underlying sensitive layer.\\n\\n225°C Process: Advantage: Lower thermal budget, safer for the underlying layer. Disadvantage: Much slower etch rate, reducing throughput.\\n\\nFor a temperature-sensitive substrate, the 225°C process is preferable. The trade-off is accepting a slower etch rate to preserve the integrity of the underlying material.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The constant slope proves that the surface chemistry resets perfectly after every cycle. There is no \\\"memory effect,\\\" residue accumulation, or surface roughening that inhibits (or accelerates) the reaction over time; the 20th cycle removes exactly as much material as the 1st cycle, confirming the robust, self-limiting nature of the ALE mechanism.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This enables the precise tuning of optical cavity lengths in ZnS-based micro-resonators or filters. By running an exact number of ALE cycles, the thickness of a ZnS waveguide layer can be reduced with sub-nanometer accuracy, allowing for post-fabrication trimming of the device's resonant wavelength to a specific target value without degrading surface smoothness.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":665,"height":577}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/23/images/figure_5.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 scatter plot","caption_source":"content.json:img_caption"},"width":667,"height":583,"image_format":"jpeg","image_sha256":"87f7a8eb795c5f28a0bfe24b8618b29a5cac7cbdee9abe55a69be65021579f28","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_3.jpg","caption":"FiG. 3. Surface roughness variation of AlGaN as a function of etch depth of the AlGaN film after (a) BT step, (b) RIE, (c) ALE using  $\\mathrm{Cl}_2$  plasma, and (d) ALE using  $\\mathrm{Cl}_2$  gas.","id":"train/atomic-layer-etching/experimental-usecase/29/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/29/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between etch depth and RMS roughness variation for different treatments, including RIE, ALE with Cl2 gas or Cl2 plasma, with the reference after BT.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Etch Depth (nm) | RMS Roughness Variation (nm) RIE |RMS Roughness Variation (nm) Cl2 gas |RMS Roughness Variation (nm) Cl2 plasma |\\n|---|---|---|---|\\n| 0 | 0 |0 | 0 |\\n| 3 | 0.05|0.01|0|\\n| 5 | 0.15 |0.05|0|\\n| 7 | 0.2|0.05|0|\\n| 10 | - |-|-0.01|\\n| 15 | - |-|-|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"RIE is increasing the roughness of the substrate, this is likely due to the sort of random distribution of ion influences on the substrate.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the case of perfect ALE, all bumps and holes on a substrate are equally etched so the average roughness does not change.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The roughness changes with 0.15 nm.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The Cl2 plasma process seems to slightly decrease the roughness but remains constant, the Cl2 gas process increases the roughness on the other hand.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":538,"height":501}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/29/images/fig_3.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 scatter plot","caption_source":"content.json:image_caption"},"width":539,"height":505,"image_format":"jpeg","image_sha256":"555566156ae829fe484902fa42b67611173633967ab2d909e07ce9cb388792ed","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_34_fig_7.jpg","caption":"FIG. 7.  $\\mathrm{Al}_2\\mathrm{O}_3$  film thickness vs cycle number of pALE for various process temperatures.","id":"train/atomic-layer-etching/experimental-usecase/34/fig_7","sample_id":"atomic-layer-etching/experimental-usecase/34/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure compares Al₂O₃ film thickness across increasing pALE cycle numbers at four different temperatures. Higher temperatures (350–400°C) show significantly greater film removal per cycle, while lower temperatures (250–300°C) show slower and less pronounced thinning, indicating a clear temperature-dependent etch rate.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle number of pALE|250°C|300°C|350°C|400°C|\\n|---------------------|-----|-----|-----|-----|\\n|0|0|0|0|0|\\n|50|-5|-7|-9|-18|\\n|100|-7|-10|-12|-25|\\n|150|-8|-15|-18|-34|\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure indicates that higher temperatures significantly accelerate Al₂O₃ removal during pALE, allowing for deeper etching in fewer cycles. Conversely, lower temperatures produce minimal thinning and offer finer control for delicate processing. Understanding this temperature-dependent response helps identify the optimal balance between etch rate, precision, and thermal stability for specific applications.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"At higher temperatures such as 400°C, the film becomes much more negative (up to −34 nm), while at 250°C the thickness only decreases slightly (around −5 to −8 nm).\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Higher thermal energy increases etch reactivity, Faster surface removal at elevated temperatures, Lower temperatures show slower, limited thinning\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":675,"height":523}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/34/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/34/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/34/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/34/CF4 plasma-based atomic layer etching of Al2O3.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"34","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":675,"height":523,"image_format":"jpeg","image_sha256":"d66e5a44f77fe217ca6add20c07c2fb5c5bb436523268e627a6909f1ca735da4","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_36_fig_10.jpg","caption":"Fig. 10. Cyclic number dependence of the blanket  $\\mathrm{TiO_2}$  etched amount at different stage temperatures: (O) both steps had temperatures of  $60^{\\circ}\\mathrm{C}$ ,  $(\\Delta)$  the  $\\mathrm{C_4F_8}$  plasma had a temperature of  $20^{\\circ}\\mathrm{C}$  and the  $\\mathrm{O_2}$  plasma a temperature of  $60^{\\circ}\\mathrm{C}$ ,  $(\\square)$  and the  $\\mathrm{C_4F_8}$  plasma had a temperature of  $60^{\\circ}\\mathrm{C}$  and the  $\\mathrm{O_2}$  plasma a temperature of  $20^{\\circ}\\mathrm{C}$ . The plasma treatment time was  $30\\mathrm{s}$  under all conditions.","id":"train/atomic-layer-etching/experimental-usecase/36/fig_10","sample_id":"atomic-layer-etching/experimental-usecase/36/fig_10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the etched amount of TiO₂ against the number of cycles for different plasma conditions (\\\"60 °C (C₄F₈ plasma), 60 °C (O₂ plasma)\\\", \\\"20 °C (C₄F₈ plasma), 60 °C (O₂ plasma)\\\", \\\"60 °C (C₄F₈ plasma), 20 °C (O₂ plasma)\\\"). The condition \\\"60 °C (C₄F₈ plasma), 60 °C (O₂ plasma)\\\" results in the highest amount etched over cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of cycles | blencket TiO2 etched amount (nm) | Measurement |\\n|---|---|---|\\n| 0 | 0 | 60 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 1 | 0.79 | 60 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 4 | 2.94 | 60 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 8 | 5.58 | 60 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 12 | 8.68 | 60 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 0 | 0 | 20 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 1 | 0.96  | 20 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 3 | 2.19  | 20 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 5 | 3.86 | 20 °C (C₄F₈ plasma), 60 °C (O₂ plasma) |\\n| 0 | 0 | 60 °C (C₄F₈ plasma), 20 °C (O₂ plasma) |\\n| 1 | 0.6 | 60 °C (C₄F₈ plasma), 20 °C (O₂ plasma) |\\n| 3 | 1.03 | 60 °C (C₄F₈ plasma), 20 °C (O₂ plasma) |\\n| 6 | 1.54 | 60 °C (C₄F₈ plasma), 20 °C (O₂ plasma) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"O₂ plasma step is more temperature dependent. Lowering C₄F₈ plasma step temperature has little effect. Lowering the temperature for the O₂ plasma step significantly reduces the etch amount.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"This suggests the step (the removal of the modified layer)  is thermally activated.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"60 °C C₄F₈ plasma, 20 °C O₂ plasma (smallest error bars and least fluctuation).\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"60 °C C₄F₈ plasma, 60 °C O₂ plasma (largest error bars and highest variability).\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":661,"height":438}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/images/fig_10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/Cyclic C4F8 and O2 plasma etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":669,"height":448,"image_format":"jpeg","image_sha256":"65d9ad11d60f4fc2a419d63a0b83a1d1a535b99cfba544ad1b889c4e24707fe3","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_36_fig_14.jpg","caption":"Fig. 14. Sidewall  $\\mathrm{TiO_2}$  thicknesses of the trench pattern samples as deposited and after the cyclic processes.","id":"train/atomic-layer-etching/experimental-usecase/36/fig_14","sample_id":"atomic-layer-etching/experimental-usecase/36/fig_14","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between depth and sidewall TiO2 thickness as deposited and  for different plasma treatments (30s and 60s oxygen plasma treatment, after 8 cycles).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Depth (nm) | Sidewall TiO2 thickness (nm) | Measurement | \\n|---|---|---|\\n| 681 |  9.04 | as TiO2 deposited |\\n| 1391 |  9.55 | as TiO2 deposited |\\n| 2091 |  9.78 | as TiO2 deposited |\\n| 2786 |  9.20 | as TiO2 deposited |\\n| 3496 | 9.69 | as TiO2 deposited |\\n| 51 |  8.88 | after 8 cycles, 30 s oxygen plasma |\\n| 695 |  5.35 | after 8 cycles, 30 s oxygen plasma |\\n| 1400 |  5.88 | after 8 cycles, 30 s oxygen plasma |\\n| 2095 |  6.33 | after 8 cycles, 30 s oxygen plasma |\\n| 2800 |  6.15 | after 8 cycles, 30 s oxygen plasma |\\n| 3496 |  5.59 | after 8 cycles, 30 s oxygen plasma |\\n| 27.8 |  5.88 | after 8 cycles, 60 s oxygen plasma |\\n| 690 |  4.47 | after 8 cycles, 60 s oxygen plasma |\\n| 1391 |  5.21 | after 8 cycles, 60 s oxygen plasma |\\n| 2100 |  4.74 | after 8 cycles, 60 s oxygen plasma |\\n| 2800 |  5.10 | after 8 cycles, 60 s oxygen plasma |\\n| 3496 |  5.82 | after 8 cycles, 60 s oxygen plasma |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Not uniform, as the top remains unetched while the deeper sections show significant material removal.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"60 s plasma treatment.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"~ 4.4 nm.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The protective layer is thicker at the top of the trench. This is demonstrated by the significantly smaller amount of etched TiO₂ at the trench opening compared to the deeper sidewalls under identical plasma conditions. The thicker protective layer at the top prevents  the underlying material from being etched for longer, until the layer is fully removed.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":663,"height":433}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/images/fig_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/images/fig_14.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/Cyclic C4F8 and O2 plasma etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":666,"height":437,"image_format":"jpeg","image_sha256":"c07628641567f17b181cc1a90f1d293eece792df37bfb6d8c8e56d83c5478db0","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_36_fig_3.jpg","caption":"Fig. 3. (a) Depth dependence of the sidewall  $\\mathrm{TiO_2}$  thicknesses before and after  $\\mathrm{CF_4}$  plasma treatment for  $30~\\mathrm{s}$  and (b) the sidewall etched amount.","id":"train/atomic-layer-etching/experimental-usecase/36/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/36/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart compares the sidewall TiO2 thickness in nm for pristine TiO2 and TiO2 after plasma etching as a function of depth. Sidewall thickness after etching is lower than before.\"},{\"panel_id\":\"b\",\"text\":\"The scatter plot shows the sidewall TiO2 etched amount in nm as a function of depth. The etching at lower depths is higher.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| depth (nm) | Sidewall TiO2 thickness (nm): Pristine | Sidewall TiO2 thickness (nm): after RIE |\\n| --- | --- | --- |\\n| 20 | 9,83 | 4,35 |\\n| 39 | 9,79 | 7,76 |\\n| 60 | 9,65 | 7,81 |\\n| 80 | 9,97 | 8,08 |\\n| 101 | 9,60 | 8,13 |\\n| 119 | 9,83 | 8,13 |\\n| 139 | 9,83 | 8,13 |\\n| 160 | 9,56 | 8,27 |\\n| 179 | 9,60 | 8,45 |\\n| 199 | 9,65 | 8,41 |\\n| 219 | 9,65 | 8,31 |\\n| 240 | 9,79 | 8,36 |\\n| 257 | 9,60 | 8,45 |\\n| 280 | 9,79 | 8,50 |\\n| 299 | 9,47 | 8,04 |\"},{\"panel_id\":\"b\",\"text\":\"| depth (nm) | sidewall TiO2 etched amount (nm) |\\n|---|---|\\n| 19 | 5,46 |\\n| 38 | 2,08 |\\n| 58 | 1,92 |\\n| 78 | 1,92 |\\n| 97 | 1,43 |\\n| 119 | 1,62 |\\n| 139 | 1,57 |\\n| 159 | 1,41 |\\n| 178 | 1,11 |\\n| 200 | 1,22 |\\n| 218 | 1,43 |\\n| 239 | 1,43 |\\n| 259 | 1,22 |\\n| 280 | 1,27 |\\n| 300 | 1,41 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It was uniform.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Tapered (because the etching is stronger at the top opening than deep inside).\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Top 20 nm.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etching gets weaker. The etched amount drops from ~5.5 nm at the top to ~1.5 nm at the bottom. This happens because plasma is mostly being consumed at the top and can't reach the bottom effectively.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":3,"width":673,"height":464},{"panel_id":"b","x":5,"y":471,"width":670,"height":460}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/images/fig_3.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/36/Cyclic C4F8 and O2 plasma etching of TiO2.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"36","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":677,"height":933,"image_format":"jpeg","image_sha256":"6485613f3dee907980fdfb2f2b9b7c166cc20c9ea54deaa8bf239b9ac710ad1f","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_experimental-usecase_38_fig4.jpg","caption":"FIG.4. (a) Plted etch depth vs ALE cycle. Dashed line marks the transition between AlGaN and GaN layers. AFM scans of various steps (b) 25, (c) 100, (d) 300, and (e) 400 cycles.","id":"train/atomic-layer-etching/experimental-usecase/38/fig4","sample_id":"atomic-layer-etching/experimental-usecase/38/fig4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"unknown"},{"panel_id":"c","label":"unknown"},{"panel_id":"d","label":"unknown"},{"panel_id":"e","label":"unknown"}],"summarization":"[]","data_extraction":"[]","vqa":"[]","bbox":[{"panel_id":"a","x":76,"y":6,"width":1081,"height":470},{"panel_id":"b","x":6,"y":539,"width":644,"height":291},{"panel_id":"c","x":681,"y":546,"width":652,"height":291},{"panel_id":"d","x":10,"y":890,"width":630,"height":280},{"panel_id":"e","x":690,"y":890,"width":643,"height":289}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/images/fig4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/images/fig4.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/38/High synergy atomic layer etching of AlGaNGaN.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"38","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":1344,"height":1192,"image_format":"jpeg","image_sha256":"9bc43ca2f343a8ac0c5ef28ed4b02b61f0e0fb9122b4fc36f60f570b26b143c3","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_40_fig5.jpg","caption":"FIG.5. Atomic concentrations of Ti, N, O, F, and C vs Ar milling time and estimated depth for (a) original and (b) ALE-treated TiN thin films. The dashed lines are guides to the eye.","id":"train/atomic-layer-etching/experimental-usecase/40/fig5","sample_id":"atomic-layer-etching/experimental-usecase/40/fig5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Depth-profiling XPS data for the original TiN film showing atomic concentrations versus Ar sputtering time and estimated depth. At the surface, oxygen (~16%) and carbon (~12%) are elevated due to native oxide and adventitious carbon. After ~60 s of sputtering (~2 nm depth), concentrations plateau to bulk values: Ti (~49%), N (~42%), O (~6%), C (~2%), and F (~1%).\"},{\"panel_id\":\"b\",\"text\":\"Depth-profiling XPS data for the ALE-treated TiN film. The surface shows significantly lower oxygen (~8%) compared to the original film, while fluorine is slightly elevated (~6.5%). Bulk composition remains essentially unchanged from the original film, confirming that ALE affects only the top few nanometers.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputtering time, s | Estimated depth, nm | Ti (%) | N (%) | F (%) | O (%) | C (%) |\\n|---|---|---|---|---|---|---|\\n| 0 | 0 | 32 | 38 | 3 | 15 | 12 |\\n| 60 | 2 | 46 | 45 | 3 | 5 | 2 |\\n| 120 | 3 | 48 | 43 | 2 | 6 | 2 |\\n| 180 | 5 | 48 | 43 | 2 | 6 | 2 |\\n| 240 | 7 | 48 | 42 | 2 | 6 | 2 |\\n| 300 | 8 | 50 | 42 | 2 | 6 | 2 |\\n| 360 | 10 | 48 | 41 | 2 | 6 | 2 |\"},{\"panel_id\":\"b\",\"text\":\"| Sputtering time, s | Estimated depth, nm | Ti (%) | N (%) | F (%) | O (%) | C (%) |\\n|---|---|---|---|---|---|---|\\n| 0 | 0 | 34 | 41 | 6 | 7 | 12 |\\n| 60 | 2 | 47 | 46 | 2 | 5 | 2 |\\n| 120 | 3 | 49 | 44 | 2 | 5 | 2 |\\n| 180 | 5 | 49 | 43 | 2 | 5 | 2 |\\n| 240 | 7 | 50 | 43 | 1 | 6 | 2 |\\n| 300 | 8 | 50 | 43 | 1 | 6 | 2 |\\n| 360 | 10 | 49 | 42 | 1 | 6 | 2 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The compositional gradient in the first 2 nm reflects the presence of a modified surface layer on the TiN film. At the surface, native oxide formation from atmospheric exposure causes elevated oxygen content, while adventitious carbon contamination adds to the carbon signal. As the Ar ion beam sputters through this surface layer, it reaches the bulk TiN film where the composition stabilizes at the stoichiometric values determined by the ALD growth process. This depth-dependent behavior confirms that surface contamination and oxidation are confined to a thin surface region.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Ti, N, F, O, C\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The surface oxygen decreases from approximately 16% in the original film to about 8% in the ALE-treated film, representing a 49% reduction.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No, the ALE process does not alter the bulk composition. Both the original and ALE-treated films show nearly identical bulk atomic concentrations after sputtering beyond 2 nm depth: approximately 49% Ti, 42% N, 6% O, 2% C, and 1% F. This indicates that the plasma-thermal ALE process selectively modifies only the surface region while preserving the intrinsic film properties. This characteristic is important for applications where maintaining bulk superconducting or electronic properties is essential.\"}]}]","bbox":[{"panel_id":"a","x":12,"y":13,"width":657,"height":488},{"panel_id":"b","x":18,"y":528,"width":651,"height":480}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/images/fig5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/images/fig5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/Isotropic plasma-thermal atomic layer etching of.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":669,"height":1008,"image_format":"jpeg","image_sha256":"db34d255221b4ea2146e04e45e267f82408f1dfbe7f7ce2b4da4dbe9f713b3af","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_40_fig_2.jpg","caption":"FIG. 2. (a) Etch rate of  $\\mathrm{TiO}_2$  (red circles) and TiN (blue squares) vs the  $\\mathrm{SF}_6{:}\\mathrm{H}_2$  flow rate ratio. The green shaded area represents the flow rate ratios for which selective etching of  $\\mathrm{TiO}_2$  over TiN was achieved. The vertical dashed black line at a ratio of 0.2 represents the ratio used in the ALE experiments. (b) TiN thickness change vs number of cycles with exposure only to  $\\mathrm{O}_2$  gas (red triangles), in situ HF (green squares), full ALE process at  $200^{\\circ}\\mathrm{C}$  (purple circles), and  $300^{\\circ}\\mathrm{C}$  (blue diamonds). The dashed lines are guides to the eye.","id":"train/atomic-layer-etching/experimental-usecase/40/fig_2","sample_id":"atomic-layer-etching/experimental-usecase/40/fig_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch rate of TiO₂ and TiN as a function of SF₆:H₂ flow rate ratio.\"},{\"panel_id\":\"b\",\"text\":\"The thickness change of materials with different treatments as a function of the number of cycles\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| SF₆:H₂ flow rate ratio | TiO₂ (Å s⁻¹) | TiN (Å s⁻¹) |\\n|---|---|---|\\n| 0 | 0.00 | -0.05 |\\n| 0.05 | 0.00 | -0.05 |\\n| 0.10 | 0.18 | -0.03 |\\n| 0.15 | 0.45 | -0.02 |\\n| 0.20 | 0.58 | 0.00 |\\n| 0.25 | 0.65 | 0.16 |\\n| 0.30 | 0.73 | 0.33 |\"},{\"panel_id\":\"b\",\"text\":\"| Number of cycles | O₂ gas (nm) | In-situ HF (nm) | ALE 200°C (nm) | ALE 300°C (nm) |\\n|---|---|---|---|---|\\n| 0 | 0 | 0 | 0 | 0 |\\n| 20 | 0 | 0 | -5 | -6.5 |\\n| 40 | 0 | 0 | -10 | -13 |\\n| 60 | 0 | 0 | -15 | -19 |\\n| 80 | 0 | -1 | -20 | -26 |\\n| 100 | 0 | -1 | -24 | -33 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"A ratio of 0.2 was selected for the ALE experiments, as indicated by the vertical dashed line in the figure. This ratio was chosen because it lies at the upper boundary of the selective etching window (0.1 to 0.2) where TiO2 is etched while TiN remains unetched. At this ratio, the SF6/H2 plasma produces sufficient in-situ HF to etch the oxide layer, but the fluorine radical concentration remains low enough to avoid spontaneous etching of TiN. This maximizes the etch rate of the modified layer while maintaining selectivity.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"When the SF6:H2 ratio exceeds 0.2, the concentration of fluorine radicals in the plasma becomes sufficient to spontaneously etch TiN. At lower ratios, hydrogen radicals combine with fluorine radicals to form HF molecules, which do not etch TiN due to its 3+ oxidation state. However, at higher ratios, excess fluorine radicals are available that can directly attack the TiN surface, leading to uncontrolled etching of both TiO2 and TiN and loss of selectivity.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The green shaded region represents the range of SF6:H2 flow rate ratios (0.1 to 0.2) where selective etching of TiO2 over TiN is achieved.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The etch rate at 300°C (3.2 Å/cycle) is approximately 33% higher than at 200°C (2.4 Å/cycle).\"}]}]","bbox":[{"panel_id":"a","x":1,"y":41,"width":650,"height":525},{"panel_id":"b","x":11,"y":586,"width":648,"height":525}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/images/fig_2.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/Isotropic plasma-thermal atomic layer etching of.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":675,"height":1117,"image_format":"jpeg","image_sha256":"2bb69647bfc4e2690a1e30795d286d79c3e41894b5b83bfdb26b0d07e8aaa3a3","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_40_fig_7.jpg","caption":"FIG. 7. Resistivity vs temperature for an original  $60~\\mathsf{nm}$  TiN film (blue squares), an ALE-treated film of  $50~\\mathsf{nm}$  thickness (red circles), and a  $50~\\mathsf{nm}$  ALD TiN film (black diamonds) for comparison. The difference in  $T_{c}$  between the  $50~\\mathsf{nm}$  ALD film and the ALE-treated  $50~\\mathsf{nm}$  film is negligible. The dashed lines are guides to the eye.","id":"train/atomic-layer-etching/experimental-usecase/40/fig_7","sample_id":"atomic-layer-etching/experimental-usecase/40/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Resistivity versus temperature (1.7 to 6 K) showing superconducting transitions for three TiN films. The 60 nm ALD film (blue) has Tc ≈ 3.22 K and resistivity ~222 μΩcm. The ALE-treated 50 nm film (red) and 50 nm ALD reference (black) show nearly identical Tc ≈ 3.1 K, demonstrating that ALE does not degrade superconducting properties beyond the expected thickness dependence.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature, K | 60 nm ALD (μΩcm) | 50 nm after ALE (μΩcm) | 50 nm ALD (μΩcm) |\\n|---|---|---|---|\\n| 2.0 | 0 | 0 | 0 |\\n| 2.5 | 0 | 0 | 0 |\\n| 2.9 | 0 | 0 | 0 |\\n| 3.0 | 0 | 0 | 85 |\\n| 3.05 | 0 | 5 | 110 |\\n| 3.1 | 15 | 150 | 150 |\\n| 3.15 | 65 | 165 | 175 |\\n| 3.2 | 150 | 185 | 190 |\\n| 3.3 | 185 | 195 | 210 |\\n| 3.5 | 210 | 200 | 220 |\\n| 4.0 | 215 | 202 | 225 |\\n| 5.0 | 218 | 203 | 227 |\\n| 6.0 | 220 | 205 | 228 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ALE-treated 50 nm film has Tc ≈ 3.13 K, which is nearly identical to the 50 nm ALD reference film at Tc ≈ 3.11 K. This negligible difference demonstrates that the ALE process does not introduce damage that would degrade superconducting properties. The small Tc reduction from the 60 nm film (Tc ≈ 3.22 K) is consistent with expected thickness-dependent behavior.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure demonstrates that the plasma-thermal ALE process is low-damage. Conventional etching methods often degrade superconducting properties through subsurface damage or contamination. Here, the ALE-treated film maintains the same Tc as an as-deposited film of equal thickness, confirming that ALE removes material without harming the remaining film quality.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The ALE-treated film has lower resistivity (~201 μΩcm) compared to the 50 nm ALD film (~227 μΩcm) due to reduced native oxide after etching. The authors hypothesize that fluorine incorporation during ALE may also act as a diffusion barrier, slowing reoxidation when exposed to atmosphere.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Four-point resistivity measurements using a Quantum Design DynaCool Physical Property Measurement System (PPMS) from 6 K down to 1.7 K.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":3,"width":668,"height":537}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/images/fig_7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/40/Isotropic plasma-thermal atomic layer etching of.pdf"],"main_category":"atomic-layer-etching","sub_category":"experimental-usecase","paper_id":"40","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":542,"image_format":"jpeg","image_sha256":"dc38a50162b60d977b7f24fc47c3d3159bb61574b450b9f00a79347ae0751e66","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_2.jpg","caption":"Figure 2. (a) Film thickness versus number of HF and  $\\mathrm{Sn(acac)}_2$  reaction cycles at  $200^{\\circ}\\mathrm{C}$  for a variety of materials. (b) Film thickness versus number of HF and  $\\mathrm{Sn(acac)}_2$  reaction cycles at  $200^{\\circ}\\mathrm{C}$  for  $\\mathrm{Al}_2\\mathrm{O}_3$ ,  $\\mathrm{HfO}_2$ , and  $\\mathrm{ZrO}_2$  showing etch rates.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_2","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the film thickness of various materials (TiN, SiO₂, Si₃N₄, HfO₂, ZrO₂, Al₂O₃) at 200°C when Sn(acac)2 is used for etching.\"},{\"panel_id\":\"b\",\"text\":\"The figure illustrates the atomic layer deposition (ALE) rates when Sn(acac)2 is used for etching for HfO₂, ZrO₂, and Al₂O₃ at 200°C over cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Film Thickness (Å) | TiN | SiO₂ | Si₃N₄ | HfO₂ | ZrO₂ | Al₂O₃ |\\n|-------------------|-----|------|-------|------|------|-------|\\n| 0                | 70  | 50   | 40    | 30   | 0    | 30    |\\n| 1-30-1-30         | 68  | 48   | 38    | 32   | 0    | 28    |\\n| 200°C             | 66  | 46   | 36    | 30   | 0    | 26    |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Cycles | HfO₂ ALE: 0.06 Å/cycle | ZrO₂ ALE: 0.14 Å/cycle | Al₂O₃ ALE: 0.23 Å/cycle |\\n|------------------|------------------------|------------------------|-------------------------|\\n| 0                | 50                     | 50                     | 50                      |\\n| 100              | 45                     | 45                     | 40                      |\\n| 200              | 40                     | 40                     | 25                      |\\n| 400              | 35                     | 35                     | 0                       |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Sequential exposures of HF and acetylacetone (acacH, Sigma-Aldrich >99%) did not lead to the etching of Al2O3 at 200 °C. Without the metal acetylacetonate precursor, acetylacetone alone does not yield thermal ALE.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"From the film thicknesses after 50, 100, 200, and 400 ALE cycles using sequential HF and Sn(acac)2 exposures at 200 °C, Al2O3, ZrO2, and HfO2 films were etched linearly versus the number of ALE cycles. In contrast, there were no measurable\\nthickness changes for the SiO2, Si3N4, and TiN films.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The slopes from the linear least-squares fittings of the data in image b yielded etch rates of 0.23 Å/cycle, 0.14 Å/cycle, and 0.06 Å/cycle for the Al2O3, ZrO2, and HfO2 films, respectively\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Additional control experiments were performed on the Al2O3, ZrO2, and HfO2 films to determine if both HF and Sn(acac)2 were necessary for the thermal ALE. Experiments using 200 cycles of HF exposures observed negligible thickness changes for the Al2O3, ZrO2, and HfO2 films.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":4,"width":671,"height":454},{"panel_id":"b","x":0,"y":464,"width":665,"height":534}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_2.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 scatter plot","caption_source":"content.json:image_caption"},"width":673,"height":1003,"image_format":"jpeg","image_sha256":"8e969cb2dbf584acd89641ce66d575dfdd841aaa9ad1b55912724f73559f7059","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_3.jpg","caption":"Figure 3. (a) Film thickness versus number of HF and TMA reaction cycles at  $300^{\\circ}\\mathrm{C}$  for a variety of materials. (b) Film thickness versus number of HF and TMA reaction cycles at  $300^{\\circ}\\mathrm{C}$  for  $\\mathrm{Al}_2\\mathrm{O}_3$ ,  $\\mathrm{HfO}_2$ , and  $\\mathrm{ZrO}_2$  showing etch rates.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the film thickness of various materials (TiN, ZrO₂, SiO₂, Si₃N₄, HfO₂, Al₂O₃) at 300°C using TMA and HF for ALE etching.\"},{\"panel_id\":\"b\",\"text\":\"The figure illustrates the atomic layer etching (ALE) rates of ZrO₂, HfO₂, and Al₂O₃ at 300°C using TMA and HF\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | TiN (Å) | SiO₂ (Å) | Al₂O₃ (Å) | Si₃N₄ (Å) | HfO₂ (Å) | ZrO₂ (Å) |\\n|-------------------|---------|----------|-----------|-----------|----------|----------|\\n| 0                | 65      | 50       | 45        | 45        | 55       | 50       |\\n| 50               | 65      | 49       | 30        | 44        | 50       | 49       |\\n| 100              | 65      | 48       | 15        | 43        | 48       | 48       |\\n| 150              | 65      | 47       | 5         | 42        | 40       | 47       |\\n| 200              | 65      | 46       | 0         | 42        | 38       | 46       |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Cycles | Al₂O₃ (Å) | HfO₂ (Å) | ZrO₂ (Å) |\\n|-------------------|-----------|----------|----------|\\n| 0                | 45        | 55       | 50       |\\n| 50               | 20        | 50       | 49       |\\n| 100              | 0        | 48       | 48       |\\n| 150              | -         | 40       | 47       |\\n| 200              | 0         | 38       | 47       |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"All the ALE processes on different materials with HF and TMA are thermal processes conducted at 300 °C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The film thicknesses after 25, 50, 100, and 200 ALE cycles using sequential HF and TMA exposures at 300 °C. The Al2O3 and HfO2 films were etched linearly versus the number of ALE cycles. There were negligible thickness changes during 200 ALE cycles for the ZrO2, SiO2, Si3N4, and TiN films.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch rates of 0.45 Å/cycle and 0.10 Å/cycle were measured for the Al2O3 and HfO2 films, respectively. The etch rate of Al2O3 is in agreement with the value measured earlier by quartz crystal microbalance\\n(QCM) and SE experiments.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":669,"height":455},{"panel_id":"b","x":0,"y":471,"width":669,"height":538}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_3.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 scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":1009,"image_format":"jpeg","image_sha256":"8324b60255a0acd9120ea88d7bed815d40e0a82cb6273ef6a33d15380937a7af","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_4.jpg","caption":"Figure 4. (a) Film thickness versus number of HF and DMAC reaction cycles at  $250^{\\circ}\\mathrm{C}$  for a variety of materials. (b) Film thickness versus number of HF and DMAC reaction cycles at  $250^{\\circ}\\mathrm{C}$  for  $\\mathrm{Al}_2\\mathrm{O}_3$ ,  $\\mathrm{HfO}_2$ , and  $\\mathrm{ZrO}_2$  showing etch rates.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_4","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_4","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the film thickness of various materials (TiN, SiO₂, Si₃N₄, ZrO₂, HfO₂, Al₂O₃) at 250°C for various DMAC and HF etch cycles.\"},{\"panel_id\":\"b\",\"text\":\"The figure illustrates the atomic layer etching (ALE)  etch rates of HfO₂, Al₂O₃, and ZrO₂ at 250°C\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Film Thickness (Å) | TiN | SiO₂ | Si₃N₄ | ZrO₂ | HfO₂ | Al₂O₃ |\\n|-------------------|-----|------|-------|------|------|-------|\\n| 0                | 65  | 50   | 40    | 40   | 50   | 45    |\\n| 25               | 65  | 50   | 40    | 30   | 40   | 40    |\\n| 50               | 65  | 50   | 40    | 20   | 30   | 35    |\\n| 100              | 65  | 50   | 40    | 0    | 0    | 20     |\"},{\"panel_id\":\"b\",\"text\":\"| Film Thickness (Å) | HfO₂ ALE: 0.77 Å/cycle | Al₂O₃ ALE: 0.32 Å/cycle | ZrO₂ ALE: 0.96 Å/cycle |\\n|-------------------|-------------------------|-------------------------|-------------------------|\\n| 0                 | 50                      | 45                      | 50                      |\\n| 25                | 40                      | 40                      | 35                      |\\n| 50                |20                      | 35                      | 10                      |\\n| 100               | 0                       | 20                       | 0                       |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The film thicknesses after 10, 25, 50, and 100 ALE cycles using sequential HF and DMAC exposures at 250 °C show linear etching\\n for the ZrO2, HfO2, and Al2O3 films. While no measurable thickness changes were  observed for the SiO2, Si3N4, and TiN films\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"250 °C\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. Al2O3\\n, 2. HfO2\\n, 3. ZrO2\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etch rates of 0.96 Å/cycle, 0.77 Å/cycle, and 0.32 Å/cycle were determined for the ZrO2, HfO2, and Al2O3 films, respectively. Al2O3 etched at the slowest rate compared to the other two materials.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":673,"height":451},{"panel_id":"b","x":0,"y":464,"width":671,"height":538}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_4.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_4.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 scatter plot","caption_source":"content.json:image_caption"},"width":673,"height":1005,"image_format":"jpeg","image_sha256":"685e938ba593f3319495ae96ac1a4eadaeffd0c0e2944819df52f53421f16306","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_5.jpg","caption":"Figure 5. (a) Film thickness versus number of HF and  $\\mathrm{SiCl}_4$  reaction cycles at  $350^{\\circ}\\mathrm{C}$  for a variety of materials. (b) Film thickness versus number of HF and  $\\mathrm{SiCl}_4$  reaction cycles at  $350^{\\circ}\\mathrm{C}$  for  $\\mathrm{Al}_2\\mathrm{O}_3$ ,  $\\mathrm{HfO}_2$ , and  $\\mathrm{ZrO}_2$  showing etch rates.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the film thickness of various materials (TiN, Al₂O₃, SiO₂, Si₃N₄, HfO₂, ZrO₂) at 350°C using SiCl₄ as function of number of cycles\"},{\"panel_id\":\"b\",\"text\":\"The figure compares the film thickness of Al₂O₃, HfO₂, and ZrO₂ with different atomic layer deposition (ALE) rates at 350°C using SiCl₄.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | TiN (Å) | SiO₂ (Å) | Al₂O₃ (Å) | Si₃N₄ (Å) | HfO₂ (Å) | ZrO₂ (Å) |\\n|-------------------|---------|----------|-----------|-----------|----------|----------|\\n| 0                | 65      | 50       | 45        | 45        | 55       | 50       |\\n| 100              | 65      | 48       | 44        | 44        | 50       | 35       |\\n| 200              | 65      | 47       | 44        | 43        | 48       | 20       |\\n| 300              | 65      | 46       | 43        | 42        | 46       | 10       |\\n| 400              | 65      | 45       | 43        | 42        | 44       | 5        |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Cycles | Al₂O₃ (Å) | HfO₂ (Å) | ZrO₂ (Å) |\\n|-------------------|-----------|----------|----------|\\n| 0                | 45        | 55       | 50       |\\n| 100              | 45        | 50       | 35       |\\n| 200              | 45        | 48       | 20       |\\n| 300              | 45        | 46       | 10       |\\n| 400              | 45        | 44       | 5        |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The film thicknesses after 50, 100, 200, and 400 ALE cycles using sequential HF and SiCl4 exposures at 350 °C. The ZrO2 and HfO2 films were etched linearly versus the number of ALE cycles. In contrast, there were no measurable thickness changes during 400 ALE cycles for the Al2O3, SiO2, Si3N4, and TiN films\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Al₂O₃ ALE: 0 Å/cycle  \\n, HfO₂ ALE: 0.05 Å/cycle  \\n, ZrO₂ ALE: 0.14 Å/cycle\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Image b shows only those materials which get etched with HF anf SiCl4 from the group shown in image a\"}]}]","bbox":[{"panel_id":"a","x":2,"y":0,"width":669,"height":468},{"panel_id":"b","x":2,"y":470,"width":664,"height":528}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_5.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 scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":1005,"image_format":"jpeg","image_sha256":"24098daf06b854c9163821f2cf3c8b0f8a7c67472fc77be0b6040dd61f2417fc","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_6.jpg","caption":"Figure 6. (a) Etch rates for  $\\mathrm{ZrO}_2$  and  $\\mathrm{HfO}_2$  versus  $\\mathrm{SiCl}_4$  exposure time with constant HF exposure time of  $1.0~\\mathrm{s}$ . (b) Etch rates for  $\\mathrm{ZrO}_2$  and  $\\mathrm{HfO}_2$  versus HF exposure time with constant  $\\mathrm{SiCl}_4$  exposure time of  $2.0~\\mathrm{s}$ .","id":"train/atomic-layer-etching/experimental-usecase/42/figure_6","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The etch rate of ZrO₂ and HfO2 increases slightly with increasing SiCl₄ exposure time at 350°C.\"},{\"panel_id\":\"b\",\"text\":\"The etch rate of ZrO₂ increases significantly while HfO2 only slightly with increasing HF exposure time at 350°C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| SiCl₄ Exposure Time (s) | ZrO₂ Etch Rate (Å/cycle) | HfO₂ Etch Rate (Å/cycle) |\\n|--------------------------|---------------------------|---------------------------|\\n| 0.0                      | 0.00                     | 0.00                     |\\n| 0.5                      | 0.13                     | 0.05                     |\\n| 1.0                      | 0.14                     | 0.06                     |\\n| 1.5                      | 0.14                     | 0.06                     |\\n| 2.0                      | 0.14                     | 0.06                     |\"},{\"panel_id\":\"b\",\"text\":\"| HF Exposure Time (s)     | ZrO₂ Etch Rate (Å/cycle) | HfO₂ Etch Rate (Å/cycle) |\\n|--------------------------|---------------------------|---------------------------|\\n| 0.0                      | 0.00                     | 0.00                     |\\n| 0.5                      | 0.07                     | 0.03                     |\\n| 1.0                      | 0.12                     | 0.05                     |\\n| 1.5                      | 0.15                     | 0.06                     |\\n| 2.0                      | 0.17                     |  0.06                     |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rates for both ZrO2 and HfO2 are self-limiting versus SiCl4 exposure. The etch rates are constant at progressively larger SiCl4 exposure times. The etch rates at each SiCl4 exposure time were determined using at least 3−4 measured film\\nthicknesses after different numbers of ALE cycles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"1. Saturated etch rate for ZrO2 = 0.14 Å/cycle\\n, 2. Saturated etch rate for HfO2 = 0.06 Å/cycle\\n, at HF exposure of 1 s and SiCl4 exposure of 1s\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The etch rate for ZrO2 begins to level off at higher HF exposures. The slight increases at larger HF exposures for ZrO2 can be attributed to the difficulty purging HF from the reactor after longer HF exposures. Residual HF remaining in the reactor will lead to chemical vapor etching (CVE) during  the SiCl4 exposure. Similar influences of CVE on thermal ALE have been observed earlier after longer HF exposures.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":6,"width":669,"height":366},{"panel_id":"b","x":0,"y":377,"width":673,"height":376}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_6.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 scatter plot","caption_source":"content.json:image_caption"},"width":675,"height":755,"image_format":"jpeg","image_sha256":"20569855f7d316f1dff5d5cb45a6e82fb45e7830b4503a4e6e222c2993aeacdf","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_8.jpg","caption":"Figure 8. Thermochemical calculations for the ligand-exchange reactions between  $\\mathrm{SiCl_4}$  and  $\\mathrm{AlF}_3,$ $\\mathrm{HfF_4}$  and  $\\mathrm{ZrF_4}$  Ligand exchange is assumed to proceed completely to  $\\mathrm{SiF_4}$  and the corresponding metal chloride.","id":"train/atomic-layer-etching/experimental-usecase/42/figure_8","sample_id":"atomic-layer-etching/experimental-usecase/42/figure_8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the change in Gibbs free energy (ΔG) for three different chemical reactions as a function of temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | ΔG for SiCl₄ + AlF₃ / SiF₄ + AlCl₃ (kcal/mol) | ΔG for SiCl₄ + HfF₄ / SiF₄ + HfCl₄ (kcal/mol) | ΔG for SiCl₄ + ZrF₄ / SiF₄ + ZrCl₄ (kcal/mol) |\\n|------------------|-----------------------------------------------|-----------------------------------------------|-----------------------------------------------|\\n| 0                | 45                                           | 8                                            | 5                                            |\\n| 100              | 40                                           | 5                                            | 2                                            |\\n| 200              | 35                                           | 2                                            | 0                                            |\\n| 300              | 30                                           | -2                                           | -5                                           |\\n| 400              | 25                                           | -6                                           | -10                                          |\\n| 500              | 20                                           | -10                                          | -15                                          |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The thermochemistry of the ligand exchange was determined assuming complete ligand exchange to form SiF4 and the fully chlorinated metal in the original metal fluoride.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The thermochemical results for the ligand exchange between SiCl4 and AlF3, ZrF4 and HfF4 can be seen in the image. These\\nresults reveal that the ligand exchange between SiCl4 and ZrF4 or HfF4 is thermochemically favorable at >200 °C. In contrast,\\nthe ligand exchange between SiCl4 and AlF3 is not thermchemically favorable at any temperature.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"ZrF4 and HfF4 can both undergo ligand exchange with SiCl4 at 350 °C because these reactions are spontaneous with ΔG ≈ −10 kcal/mol. On the other hand, Al2O3 can not be etched at 350 °C because the ligand exchange between SiCl4 and AlF3 is not spontaneous with ΔG ≈ + 26 kcal/mol.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]}]","bbox":[{"panel_id":"a","x":1,"y":4,"width":661,"height":556}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/42/images/figure_8.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 scatter plot","caption_source":"content.json:image_caption"},"width":666,"height":564,"image_format":"jpeg","image_sha256":"51e58fa207e82e7bac33ca8518e5f4617b4efd3b04f6ce7fa461eb5684f4b228","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_3.jpg","caption":"Fig. 3. Mass change per cycle as a function of (a)  $\\mathrm{NF_3}$  remote plasma time or (b) TMA pulse time at  $250^{\\circ}\\mathrm{C}$ . The TMA pulse time was  $3\\mathrm{~s~}$  in (a), and the  $\\mathrm{NF_3}$  remote plasma time was  $4\\mathrm{~s~}$  in (b).","id":"train/atomic-layer-etching/experimental-usecase/43/fig_3","sample_id":"atomic-layer-etching/experimental-usecase/43/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This two-panel figure shows Mass Change Per Cycle (MCPC) (ng/(cm²·cycle)) data for the Atomic Layer Etching (ALE) of Al₂O₃, MCPC varies the NF₃ plasma time (with fixed 3s TMA)\"},{\"panel_id\":\"b\",\"text\":\"Mass Change Per Cycle (MCPC) (ng/(cm²·cycle)) varies the TMA pulse time (with fixed 4s NF₃). Both plots show the mass loss saturating, proving each half-reaction is self-limiting.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| NF₃ Remote Plasma Time (s) | MCPC (ng/(cm²·cycle)) | Symbol    |\\n|---|---|---|\\n| 0 | 0 |NF₃ remote plasma (square) |\\n| 1.5 | -25 | NF₃ remote plasma (square) |\\n| 2.0 | -60 | NF₃ remote plasma (square) |\\n| 3.0 | -70 | NF₃ remote plasma (square) |\\n| 4.0 | -60 | NF₃ remote plasma (square) |\\n| 4.0 | -2  | NF₃ gas (circle) |\"},{\"panel_id\":\"b\",\"text\":\"| TMA Pulse Time (s) | MCPC (ng/(cm²·cycle)) |\\n|---|---|\\n| 0.0 | 0 |\\n| 0.5 | -25 |\\n| 1.5 | -45 |\\n| 3.0 | -55 |\\n| 4.5 | -58 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The curve would still rise but plateau at a lower maximum mass loss. With an undersaturated TMA dose, not all the fluorinated layer is removed each cycle. The process would not be fully self-limiting or stable, as the etch depth would depend on the incomplete removal step, leading to poor control and possible residue accumulation.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The grey region represents an incubation or threshold period. During these short exposures, the incoming fluorine radical flux is too low to fully fluorinate the Al₂O₃ surface to the depth needed for the subsequent ligand-exchange reaction with TMA. Because a sufficient coverage of surface fluoride (AlFₓ) species has not yet formed, TMA cannot remove material, resulting in nearly zero MCPC.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"60 ng/(cm²·cycle corresponds to roughly 2.0 Å/cycle\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"NF₃ Remote Plasma Exposure (4 s)\\n\\n, Purge\\n\\n, TMA Exposure (3 s)\\n\\n, Purge\"}]}]","bbox":[{"panel_id":"a","x":0,"y":4,"width":627,"height":575},{"panel_id":"b","x":5,"y":600,"width":619,"height":575}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/43/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/43/images/fig_3.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 scatter plot","caption_source":"content.json:image_caption"},"width":628,"height":1183,"image_format":"jpeg","image_sha256":"c28a2d1d5a2e74c0305b445f4263a5894060af962d843b0fe423cce36af4f698","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_10.jpg","caption":"Figure 10. X-ray reflectivity and spectroscopic ellipsometry measurements of TiN film thickness versus number of  $\\mathrm{H}_2\\mathrm{O}_2$  and HF reaction cycles at  $250^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/45/figure_10","sample_id":"atomic-layer-etching/experimental-usecase/45/figure_10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the film thickness in Ångströms (Å) over the number of cycles for XRR and SE methods at 250°C using H₂O₂ and HF\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | XRR (Film Thickness, Å) | SE (Film Thickness, Å) |\\n|---|---|---|\\n| 0 | 63 | 73 |\\n| 50 | 58 | 66 |\\n| 100 | 51 | 61 |\\n| 200 | 39 | 46 |\\n| 400 | 6 | 12 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Using both X-ray reflectivity and spectroscopic ellipsometry provides independent verification of the thickness measurements. XRR also yields additional information about film density and surface roughness. The close agreement between techniques (0.14 vs 0.15 Å/cycle) confirms measurement reliability.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The H₂O₂/HF etch rate (0.15 Å/cycle) is slightly lower than the O₃/HF etch rate (0.20 Å/cycle) at 250°C, reflecting O₃'s stronger oxidizing power.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Linearity confirms that each ALE cycle removes a constant, reproducible amount of material, demonstrating true self-limiting behavior rather than uncontrolled continuous etching.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"SE measured ~73 Å initial thickness; XRR measured ~63 Å. This offset is maintained throughout but both yield the same etch rate slope.\"}]}]","bbox":[{"panel_id":"a","x":1,"y":0,"width":662,"height":529}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_10.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 scatter plot","caption_source":"content.json:image_caption"},"width":669,"height":533,"image_format":"jpeg","image_sha256":"fa978ceca71019483da724568888267957195c7e03c4b4dae16c22dfdfeb5f1b","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_3.jpg","caption":"Figure 3. X-ray reflectivity and spectroscopic ellipsometry measurements of TiN film thickness versus number of  $\\mathrm{O}_3$  and HF reaction cycles at  $250^{\\circ}\\mathrm{C}$ .","id":"train/atomic-layer-etching/experimental-usecase/45/figure_3","sample_id":"atomic-layer-etching/experimental-usecase/45/figure_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the decrease in film thickness over cycles for TiN ALE at 250°C using O₃ & HF\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | SE (Film Thickness, Å) | XRR (Film Thickness, Å) |\\n|---|---|---|\\n| 0 | 73 | 63 |\\n| 25 | 68 | 61 |\\n| 50 | 63 | 57 |\\n| 100 | 52 | 47 |\\n| 150 | 45 | 26 |\\n| 200 | 34 | 27 |\\n| 300 | 16 | 6 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SE gives 0.19 Å/cycle and XRR gives 0.20 Å/cycle—essentially identical within experimental uncertainty, validating both measurement approaches.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The offset (~10 Å difference in initial thickness) arises from different optical/structural models used by each technique. SE uses a Lorentz optical model while XRR measures electron density contrast. Despite this absolute offset, both yield the same etch rate slope.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The optimized sequence was 3-30-1-30: 3 s O₃ exposure, 30 s N₂ purge, 1 s HF exposure, 30 s N₂ purge.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 60 Å of TiN is removed after 300 cycles, reducing the film from ~70 Å to ~10 Å (based on XRR data).\"}]}]","bbox":[{"panel_id":"a","x":2,"y":4,"width":664,"height":524}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_3.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 scatter plot","caption_source":"content.json:image_caption"},"width":670,"height":533,"image_format":"jpeg","image_sha256":"8db63fc4f91082c39c489cb9489e931881fa002014e0139c8f1873f7f7763710","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_5.jpg","caption":"Figure 5. Film thickness versus number of  $\\mathrm{O}_3$  and HF reaction cycles at  $250^{\\circ}\\mathrm{C}$  for  $\\mathrm{Al}_2\\mathrm{O}_3$ ,  $\\mathrm{HfO}_2$ ,  $\\mathrm{ZrO}_2$ ,  $\\mathrm{SiO}_2$ ,  $\\mathrm{Si}_3\\mathrm{N}_4$ , and TiN.","id":"train/atomic-layer-etching/experimental-usecase/45/figure_5","sample_id":"atomic-layer-etching/experimental-usecase/45/figure_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure demonstrates selective TiN etching using O₃/HF at 250°C. TiN (gray diamonds) shows linear thickness reduction while Al₂O₃, HfO₂, ZrO₂, SiO₂, and Si₃N₄ remain essentially unchanged over 400 cycles. This confirms high selectivity of the ALE process for TiN over common dielectrics and barrier materials.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | TiN | Al₂O₃ | Si₃N₄ | SiO₂ | ZrO₂ | HfO₂ |\\n|-----------------|-----|-------|------|-----|------|------|\\n| 0              | 75  | 40    | 40   | 55  | 55   | 55   |\\n| 100            | 40  | 40    | 40   | 55  | 55   | 55   |\\n| 200            | 20  | 40    | 40   | 55  | 55   | 55   |\\n| 300            | 10  | 40    | 40   | 55  | 55   | 55   |\\n| 400            | 5   | 40    | 40   | 55  | 55   | 55   |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Etched: TiN only. Unchanged: Al₂O₃, HfO₂, ZrO₂, SiO₂, and Si₃N₄.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"These metal oxides form stable, non-volatile fluorides (AlF₃, HfF₄, ZrF₄) when exposed to HF. Unlike TiF₄ which sublimes at 284°C, these fluorides remain on the surface and cannot be removed thermally under ALE conditions.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Dry HF (without H₂O present) cannot etch SiO₂. The HF/SiO₂ reaction requires water as a catalyst for effective etching.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Selective etching allows precise removal of one material while preserving adjacent structures. For TiN, selective ALE enables controlled thinning or removal without damaging surrounding dielectrics like HfO₂ gate oxides or SiO₂ interlayer dielectrics.\"}]}]","bbox":[{"panel_id":"a","x":2,"y":1,"width":662,"height":533}],"source":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/experimental-usecase/45/images/figure_5.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 scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":537,"image_format":"jpeg","image_sha256":"ee300273ded21a5a8c7e01786806a647f4df35db537fe6752c2d863decd5609d","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_1.jpg","caption":"FIG. 1. Amount of silicon etched as a function of cycle number. For these results, one cycle corresponds to 2255  $\\mathrm{Cl}_2$  impacts and  $1000\\mathrm{Ar}^+$  impacts. Each line corresponds to results for a different  $\\mathrm{Ar}^+$  ion energy: blue (lowest curve)  $-40\\mathrm{eV}$ , green (second lowest curve)  $-60\\mathrm{eV}$ , red (second highest curve)  $-80\\mathrm{eV}$ , and black (highest curve)  $-100\\mathrm{eV}$ . The vertical dashed lines separate the  $\\mathrm{Cl}_2$  and  $\\mathrm{Ar}^+$  portions of the ALE cycles.","id":"train/atomic-layer-etching/simulation-usecase/15/fig_1","sample_id":"atomic-layer-etching/simulation-usecase/15/fig_1","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows silicon ALE using cyclic Cl₂ adsorption and Ar⁺ bombardment, producing a characteristic stair-step etch profile. Material removal occurs only during the ion step, confirming the self-limiting ALE sequence. The etch per cycle increases with ion energy, from ~3 Å at 40 eV to ~11 Å at 100 eV after 8 cycles, indicating that higher energies may exceed the monolayer-limited regime.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle number | Si etched (Å) | Ion energy (eV) |\\n|-------------:|--------------:|---------------:|\\n| 1 | 0.3 | 40  |\\n| 1 | 0.6 | 60  |\\n| 1 | 1.0 | 80  |\\n| 1 | 1.4 | 100 |\\n| 2 | 0.5 | 40  |\\n| 2 | 1.0 | 60  |\\n| 2 | 1.7 | 80  |\\n| 2 | 2.6 | 100 |\\n| 3 | 0.8 | 40  |\\n| 3 | 1.6 | 60  |\\n| 3 | 2.5 | 80  |\\n| 3 | 3.6 | 100 |\\n| 4 | 1.2 | 40  |\\n| 4 | 2.2 | 60  |\\n| 4 | 3.5 | 80  |\\n| 4 | 5.0 | 100 |\\n| 5 | 1.7 | 40  |\\n| 5 | 3.0 | 60  |\\n| 5 | 4.8 | 80  |\\n| 5 | 6.7 | 100 |\\n| 6 | 2.1 | 40  |\\n| 6 | 3.9 | 60  |\\n| 6 | 5.8 | 80  |\\n| 6 | 8.0 | 100 |\\n| 7 | 2.6 | 40  |\\n| 7 | 4.5 | 60  |\\n| 7 | 6.8 | 80  |\\n| 7 | 9.2 | 100 |\\n| 8 | 3.1 | 40  |\\n| 8 | 5.2 | 60  |\\n| 8 | 7.7 | 80  |\\n| 8 | 10.8| 100 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Spontaneous Chemical Etching. Ideally, Cl_2 should only adsorb. If the line rises, it means the chlorine is etching the silicon spontaneously without ion activation, destroying the precision of the ALE process.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"40 eV (Blue). Although slower, the lower energy ensures the etch is strictly limited to the surface surface. The 100 eV ions would cause sidewall sputtering or corner rounding, ruining the shape of the delicate nano-fin.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"100 eV (Black). It removes >1.3 Å per cycle (steeper steps). This poses a risk of sub-surface damage or amorphization, as the high-energy ions penetrate deeper than the chemically modified surface layer.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The line would stay flat. Without the kinetic energy of the Ar^+ ions to knock off the chlorinated layer, no material would be removed, resulting in a 'missed step' in the staircase pattern.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":3,"width":1002,"height":458}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_1.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_1.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 scatter plot","caption_source":"content.json:img_caption"},"width":1009,"height":461,"image_format":"jpeg","image_sha256":"8c234eeb7004b5bc6086345f378bc5bd03fd0f61573b137f3267a6a24302217e","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_2.jpg","caption":"FIG. 2. Chlorine uptake (in units of equivalent monolayers) as a function of cycle number for the cycles described in Fig. 1. Each line corresponds to results for a different  $\\mathrm{Ar}^+$  ion energy: blue (lowest curve)  $-40\\mathrm{eV}$ , green (second lowest curve)  $-60\\mathrm{eV}$ , red (second highest curve)  $-80\\mathrm{eV}$ , and black (highest curve)  $-100\\mathrm{eV}$ . The vertical dashed lines separate the  $\\mathrm{CI}_2$  and  $\\mathsf{Ar}^+$  portions of the ALE cycles.","id":"train/atomic-layer-etching/simulation-usecase/15/fig_2","sample_id":"atomic-layer-etching/simulation-usecase/15/fig_2","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This graph illustrates the cyclic nature of the Atomic Layer Etching (ALE) process over 8 cycles. The \\\"sawtooth\\\" pattern represents the alternating addition of Chlorine (rising slope) and removal of material via Argon ion bombardment (falling slope). The distinct colored lines show that higher ion energies (e.g., 100 eV) result in deeper removal of the chlorinated layer compared to lower energies (e.g., 40 eV).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Cycle number | Cu uptake (ML) | Ion energy (eV) |\\n|--------------|---------------:|-----------------:|\\n| 1            | ~1.39          | 40               |\\n| 1            | ~1.05          | 60               |\\n| 1            | ~1.17          | 80               |\\n| 1            | ~1.04         | 100              |\\n| 2            | ~1.50          | 40               |\\n| 2            | ~1.45          | 60               |\\n| 2            | ~1.40          | 80               |\\n| 2            | ~1.35          | 100              |\\n| 3            | ~1.80          | 40               |\\n| 3            | ~1.70          | 60               |\\n| 3            | ~1.60          | 80               |\\n| 3            | ~1.50          | 100              |\\n| 4            | ~1.95          | 40               |\\n| 4            | ~1.85          | 60               |\\n| 4            | ~1.75          | 80               |\\n| 4            | ~1.60          | 100              |\\n| 5            | ~2.00          | 40               |\\n| 5            | ~1.95          | 60               |\\n| 5            | ~1.85          | 80               |\\n| 5            | ~1.55          | 100              |\\n| 6            | ~2.05          | 40               |\\n| 6            | ~2.00          | 60               |\\n| 6            | ~1.95          | 80               |\\n| 6            | ~1.60          | 100              |\\n| 7            | ~2.20          | 40               |\\n| 7            | ~2.10          | 60               |\\n| 7            | ~1.95          | 80               |\\n| 7            | ~1.70          | 100              |\\n| 8            | ~2.10          | 40               |\\n| 8            | ~2.15          | 60               |\\n| 8            | ~1.60          | 80               |\\n| 8            | ~1.40          | 100              |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Chlorination (Cl_2 Exposure): Represented by the rising edge where Cl uptake increases.\\nIon Bombardment (Ar^+Irradiation): Represented by the steep downward slope.\\nPhysical Action: The downward slope corresponds to the removal (etching) of the chlorinated surface layer by the argon ions.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Higher Residual Contamination. They are accepting a trade-off where the surface retains significantly more Chlorine (approx. 0.5–0.6 ML vs 0.2 ML) after every cycle, which could alter the starting conditions for the next layer of growth or etch.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The 40 eV ions possess lower kinetic energy, which is insufficient to break all the bonds holding the chlorinated species to the surface. This results in an incomplete etch, leaving a \\\"residue\\\" of chemically modified material (0.8 ML). In contrast, 100 eV ions have enough energy to sputter away nearly the entire modified layer, resulting in a cleaner surface.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It implies process drift and a loss of the \\\"self-limiting\\\" characteristic essential for ALE. If the residual chlorine accumulates with every cycle, the surface chemistry changes continuously, likely leading to variations in etch rate or surface roughness over time. A 500-cycle run would likely fail to meet tight tolerance specifications due to this cumulative error.\"}]}]","bbox":[{"panel_id":"a","x":17,"y":10,"width":1002,"height":454}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_2.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_2.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 scatter plot","caption_source":"content.json:img_caption"},"width":1019,"height":464,"image_format":"jpeg","image_sha256":"fb41ffcfca6c7e138a95a8dc61344badf15f9c5fe833f2bb4c0b49f0e9e415fd","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_5.jpg","caption":"FIG. 5. Etch per cycle as a function of  $\\mathsf{Ar}^+$  ion energy. For the \"High Fluence\" simulation results, one cycle corresponds to  $2255\\mathrm{Cl}_2$  impacts and  $1000\\mathrm{Ar}^+$  impacts, while for \"Low Fluence,\" one cycle corresponds to  $2255\\mathrm{Cl}_2$  impacts and  $68\\mathrm{Ar}^+$  impacts. Solid red circles are physical sputtering results from the MD simulations, open red circles are ALE results from the \"High Fluence\" MD simulations, and open red triangles are ALE results from the \"Low Fluence\" MD simulations. Solid blue squares are experimental physical sputtering results, while open blue squares are experimental ALE results, both from Park et al. (Ref. 11). The dotted lines indicate where the supposed ALE window is located from experimental results.","id":"train/atomic-layer-etching/simulation-usecase/15/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/15/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot represents that Atomic Layer Etching produces a much higher EPC than physical sputtering at the same ion energies, demonstrating the advantage of chlorine surface modification. An optimal “ALE window” (~70–90 eV) allows selective removal of the modified layer without damaging bulk silicon, while sufficient ion fluence is essential, as low fluence yields negligible etching despite chlorination. Etch per cycle is shown as a function of Ar⁺ ion energy. For the “High Fluence” simulation results, one cycle corresponds to 2255 Cl₂ impacts and 1000 Ar⁺ impacts, while for “Low Fluence,” one cycle corresponds to 2255 Cl₂ impacts and 68 Ar⁺ impacts. Solid red circles represent physical sputtering results from MD simulations, open red circles represent ALE results from the “High Fluence” MD simulations, and open red triangles represent ALE results from the “Low Fluence” MD simulations. Solid blue squares are experimental physical sputtering results, while open blue squares are experimental ALE results, both from Park et al. (Ref. 11). The dotted lines indicate the experimentally identified ALE window.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ar⁺ Energy (eV) | Etch per Cycle (Å/cycle) | Process / Condition |\\n|-----------------|--------------------------:|---------------------|\\n| 40              | 0.031                     | Physical Sputtering (Simulation) |\\n| 40              | 0.458                     | ALE (Simulation, High Fluence) |\\n| 40              | 0.181                     | ALE (Simulation, Low Fluence) |\\n| 40              | 0                         | Physical Sputtering (Experiment) |\\n| 40              |                            | ALE (Experiment) |\\n| 50              | 0.083                     | Physical Sputtering (Simulation) |\\n| 50              | 0.5941                    | ALE (Simulation, High Fluence) |\\n| 50              | 0.233                     | ALE (Simulation, Low Fluence) |\\n| 50              | 0                         | Physical Sputtering (Experiment) |\\n| 50              |                            | ALE (Experiment) |\\n| 60              |                            | Physical Sputtering (Simulation) |\\n| 60              | 0.739                     | ALE (Simulation, High Fluence) |\\n| 60              | 0.205                     | ALE (Simulation, Low Fluence) |\\n| 60              | 0.167                     | Physical Sputtering (Experiment) |\\n| 60              | 0.233                     | ALE (Experiment) |\\n| 70              | 0.289                     | Physical Sputtering (Simulation) |\\n| 70              | 0.894                     | ALE (Simulation, High Fluence) |\\n| 70              |                            | ALE (Simulation, Low Fluence) |\\n| 70              | 0.200                     | Physical Sputtering (Experiment) |\\n| 70              | 1.105                     | ALE (Experiment) |\\n| 80              | 0.425                     | Physical Sputtering (Simulation) |\\n| 80              | 1.020                     | ALE (Simulation, High Fluence) |\\n| 80              |                            | ALE (Simulation, Low Fluence) |\\n| 80              | 0.360                     | Physical Sputtering (Experiment) |\\n| 80              | 1.076                     | ALE (Experiment) |\\n| 90              | 0.598                     | Physical Sputtering (Simulation) |\\n| 90              | 1.175                     | ALE (Simulation, High Fluence) |\\n| 90              |                            | ALE (Simulation, Low Fluence) |\\n| 90              |                            | Physical Sputtering (Experiment) |\\n| 90              |                            | ALE (Experiment) |\\n| 100             | 0.763                     | Physical Sputtering (Simulation) |\\n| 100             | 1.320                     | ALE (Simulation, High Fluence) |\\n| 100             |                            | ALE (Simulation, Low Fluence) |\\n| 100             | 0.894                     | Physical Sputtering (Experiment) |\\n| 100             | 1.316                     | ALE (Experiment) |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The \\\"Low Fluence\\\" step uses only 68 $Ar^+ impacts, which is mathematically insufficient to cover and clear the surface sites generated by the 2255 Cl₂ impacts.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"At 100 eV, the selectivity decreases because Physical Sputtering (uncontrolled removal) becomes a dominant fraction of the total etch. Operating at 70 eV maintains a larger gap between the ALE rate and the Sputtering rate (Solid Square is low, ~0.2 Å/cycle). This ensures the etching is primarily driven by chemical modification, not brute-force kinetic damage.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Chlorination weakens the surface binding energy. The adsorption of Chlorine atoms draws electron density away from the silicon-silicon backbonds, lowering the energy threshold required to break them. This allows the same 80 eV ions to remove significantly more material than they could from a pristine, unmodified surface.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Above the window, the physical sputtering component becomes too high, causing ions to penetrate and damage the underlying lattice even after the chlorinated layer is removed.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":672,"height":461}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_5.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 scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":461,"image_format":"jpeg","image_sha256":"3a57d83c51a33f658ea076b7ec2e33b7ebb4124c106d1be89463774605e885dd","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_6.jpg","caption":"FIG. 6. Etch per cycle as a function of irradiation time for  $70\\mathrm{eV}\\mathrm{Ar}^+$  ions. Red circles are from MD simulations, while blue squares are experimental results from Park et al. (Ref. 11). For the simulation results, the chlorination step consists of  $2255\\mathrm{Cl}_2$  impacts, while for experiments  $20\\mathrm{SCCM}$  of  $\\mathrm{Cl}_2$  is supplied for  $20\\mathrm{s}$ . It is assumed a  $30\\mathrm{s}$  ion irradiation time corresponds to  $1000$  impacts in the simulation. The dotted line is the experimentally saturated EPC (1.36 Å/cycle).","id":"train/atomic-layer-etching/simulation-usecase/15/fig_6","sample_id":"atomic-layer-etching/simulation-usecase/15/fig_6","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot compares simulated and experimental etch per cycle (EPC) as a function of Ar⁺ ion irradiation time at 70 eV. Both datasets show EPC increasing with time before saturating, with the simulation (red circles) closely matching the trend and saturation value (~1.36 Å/cycle) of the experimental results (blue squares).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ar⁺ Irradiation (s) | Etch per Cycle (Å/cycle) | Source      |\\n|--------------------|--------------------------:|-------------|\\n| 8                  | 0.49                      | Simulation  |\\n| 8                  |                           | Experiment  |\\n| 10                 |                           | Simulation  |\\n| 10                 | 0.20                      | Experiment  |\\n| 15                 | 0.62                      | Simulation  |\\n| 15                 |                           | Experiment  |\\n| 20                 |                           | Simulation  |\\n| 20                 | 0.97                      | Experiment  |\\n| 30                 | 0.89                      | Simulation  |\\n| 30                 | 1.12                      | Experiment  |\\n| 35                 |                           | Simulation  |\\n| 35                 | 1.29                      | Experiment  |\\n| 37                 | 1.06                      | Simulation  |\\n| 37                 |                           | Experiment  |\\n| 40                 |                           | Simulation  |\\n| 40                 | 1.40                      | Experiment  |\\n| 45                 | 1.02                      | Simulation  |\\n| 45                 | 1.35                      | Experiment  |\\n| 50                 |                           | Simulation  |\\n| 50                 | 1.37                      | Experiment  |\\n| 60                 | 1.31                      | Simulation  |\\n| 60                 |                           | Experiment  |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"1.    Simulate the chlorination step with 2255 Cl₂ impacts.\\n\\n2.     Simulate the bombardment step with Ar⁺ ions (70 eV) for a duration corresponding to the target irradiation time.\\n\\n3.    Calculate the etch depth by comparing surface positions before and after the simulated cycle.\\nThe correlation (30 s ≈ 1000 impacts) calibrates the simulation's timescale to real-world tool parameters, enabling direct quantitative comparison between the model and physical experiment.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes. The close agreement validates the model's predictive power for this chemistry. It can rapidly and inexpensively screen irradiation times to identify the saturation point, significantly reducing the number of costly experimental trials needed for initial reactor recipe development.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"EPC increases with irradiation time then saturates. This occurs because the Ar⁺ ions can only remove the finite-thickness chlorinated (Si-Clₓ) layer created in the prior step. Once all the chlorinated material is sputtered away, additional ion bombardment does not increase the etch depth, leading to the observed saturation.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"This implies precise depth control is achievable. By running a calculated number of cycles (e.g., ~15 cycles), one can etch to a target depth of ~20 Å with atomic-scale accuracy. The self-limiting, saturating behavior ensures each cycle removes a consistent thickness, which is critical for such a shallow etch.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":8,"width":670,"height":496}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_6.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/15/images/fig_6.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 scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":506,"image_format":"jpeg","image_sha256":"444ca98bc4ccb0b32fa1a886272039174d539d9045f72c2fba4b3366ced57161","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_12.jpg","caption":"Fig. 12. The vacuum pressure versus etching per cycle relationship for several substrate velocities.","id":"train/atomic-layer-etching/simulation-usecase/16/fig_12","sample_id":"atomic-layer-etching/simulation-usecase/16/fig_12","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot shows the relationship between vacuum pressure (in Pa) and etching per cycle (in Å/cycle) at different speeds (60 mm/s, 90 mm/s, and 120 mm/s).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Vacuum pressure (Pa) | Etching per cycle (Å/cycle) 60mm/s |Etching per cycle (Å/cycle) 90mm/s |Etching per cycle (Å/cycle) 120mm/s |\\n|---|---|---|---|\\n|-200 | 0.37 |0.25|0.13|\\n|-150 | 0.40 |0.30|0.20|\\n|-100 | 0.40 |0.33|0.22|\\n|-50 | 0.40 |0.33|0.26|\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In the ideal case the vacuum pressure between the sample and the table does not have an influence on the trend. I.e. a straightline should be observed.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It could be that the process gases are pumped away during the dosing, due to the low vacuum pressure.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.41 Å/cycle and 0.33 Å/cycle respectively.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Linear increase\"}]}]","bbox":[{"panel_id":"a","x":0,"y":1,"width":662,"height":506}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/16/images/fig_12.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 scatter plot","caption_source":"content.json:img_caption"},"width":661,"height":506,"image_format":"jpeg","image_sha256":"26d167579d142bc49248daa60ef3d50a47e794e27d8de6c9eb63335a9796650c","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_17_fig7.jpg","caption":"Fig.7. t  t divided wafer in Fig. 2b to the bottom.","id":"train/atomic-layer-etching/simulation-usecase/17/fig7","sample_id":"atomic-layer-etching/simulation-usecase/17/fig7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"},{"panel_id":"c","label":"multiple scatter plot"},{"panel_id":"d","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The partial pressure of HF is plotted against substrate position for reactors G0 under different conditions. Each has four different dwell times (1.0 s, 0.3 s, 0.2 s, 0.1 s) represented by different markers.\"},{\"panel_id\":\"b\",\"text\":\"The partial pressure of HF is plotted against substrate position for reactors G1 under different conditions.\"},{\"panel_id\":\"c\",\"text\":\"The partial pressure of HF is plotted against substrate position for reactors G2 under different conditions.\"},{\"panel_id\":\"d\",\"text\":\"The partial pressure of HF is plotted against substrate position for reactors G3 under different conditions.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Substrate Position | 0.1 s | 0.2 s | 0.3 s | 1.0 s |\\n|--------------------|-------|-------|-------|-------|\\n| 1                  | 08    | 38    | 58    | 68    |\\n| 2                  | 9       | 39    | 59    | 68    |\\n| 3                  | 10    | 40    | 60    | 68   |\\n| 4                  | 11    | 41    | 61    | 68    |\\n| 5                  | 12    | 42    | 62    | 68   |\\n| 6                  | 13    | 43    | 62    | 68    |\\n| 7                  | 13    | 43    | 62    | 68    |\\n| 8                  | 12    | 42    | 62    | 68    |\\n| 9                  | 11    | 41    | 61    | 68    |\\n| 10                 | 10    | 40    | 60    | 68    |\\n| 11                 | 9    | 39    | 59    | 68    |\\n| 12                 | 8    | 38    | 58    | 68    |\"},{\"panel_id\":\"b\",\"text\":\"| Substrate Position | 0.1 s | 0.2 s | 0.3 s | 1.0 s |\\n|--------------------|-------|-------|-------|-------|\\n| 1                  | 08    | 38    | 58    | 68    |\\n| 2                  | 9       | 39    | 59    | 68    |\\n| 3                  | 10    | 40    | 60    | 68   |\\n| 4                  | 11    | 41    | 61    | 68    |\\n| 5                  | 12    | 42    | 62    | 68   |\\n| 6                  | 13    | 43    | 62    | 68    |\\n| 7                  | 13    | 43    | 62    | 68    |\\n| 8                  | 12    | 42    | 62    | 68    |\\n| 9                  | 11    | 41    | 61    | 68    |\\n| 10                 | 10    | 40    | 60    | 68    |\\n| 11                 | 9    | 39    | 59    | 68    |\\n| 12                 | 8    | 38    | 58    | 68    |\"},{\"panel_id\":\"c\",\"text\":\"| Substrate Position | 0.1 s | 0.2 s | 0.3 s | 1.0 s |\\n|--------------------|-------|-------|-------|-------|\\n| 1                  | 08    | 38    | 58    | 68    |\\n| 2                  | 9       | 39    | 59    | 68    |\\n| 3                  | 10    | 40    | 60    | 68   |\\n| 4                  | 11    | 41    | 61    | 68    |\\n| 5                  | 12    | 42    | 62    | 68   |\\n| 6                  | 13    | 43    | 62    | 68    |\\n| 7                  | 13    | 43    | 62    | 68    |\\n| 8                  | 12    | 42    | 62    | 68    |\\n| 9                  | 11    | 41    | 61    | 68    |\\n| 10                 | 10    | 40    | 60    | 68    |\\n| 11                 | 9    | 39    | 59    | 68    |\\n| 12                 | 8    | 38    | 58    | 68    |\"},{\"panel_id\":\"d\",\"text\":\"| Substrate Position | 0.1 s | 0.2 s | 0.3 s | 1.0 s |\\n|--------------------|-------|-------|-------|-------|\\n| 1                  | 08    | 38    | 58    | 68    |\\n| 2                  | 9       | 39    | 59    | 68    |\\n| 3                  | 10    | 40    | 60    | 68   |\\n| 4                  | 11    | 41    | 61    | 68    |\\n| 5                  | 12    | 42    | 62    | 68   |\\n| 6                  | 13    | 43    | 62    | 68    |\\n| 7                  | 13    | 43    | 62    | 68    |\\n| 8                  | 12    | 42    | 62    | 68    |\\n| 9                  | 11    | 41    | 61    | 68    |\\n| 10                 | 10    | 40    | 60    | 68    |\\n| 11                 | 9    | 39    | 59    | 68    |\\n| 12                 | 8    | 38    | 58    | 68   |\"}]","vqa":"[{\"panel_id\":\"d\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The reactor with the inclined plate (G3) has the most uniform flow pattern at every time step, in which the largest pressure deviation is 2.9 Pa at 0.2 s.\"}]},{\"panel_id\":\"c\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Among G0, G1 and G2, G2 reactor has uniform flow at the shortest time of 0.1s\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"For every group (G0–G3), partial pressure increases strongly with longer HF pulse time: 0.1 s < 0.2 s < 0.3 s ≪ 1.0 s.\\nThe 1.0 s trace is near the upper limit (~68–70 Pa) and essentially flat. Across-substrate uniformity: Pressures are lower at the edges and higher near the center, especially at the shorter pulses (0.1–0.3 s). The center‑to‑edge difference is ~4–10 Pa depending on dose and group.\"}]}]","bbox":[{"panel_id":"a","x":10,"y":0,"width":588,"height":399},{"panel_id":"b","x":627,"y":0,"width":584,"height":401},{"panel_id":"c","x":8,"y":472,"width":591,"height":386},{"panel_id":"d","x":624,"y":472,"width":587,"height":389}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/images/fig7.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/17/Multiscale computational fluid dynamics modeling of thermal atomic layer etching.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"17","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1222,"height":861,"image_format":"jpeg","image_sha256":"78201347ea2339d9d3d2b954e37f511f78db95d53a42153a7b1d9f0dca9b7062","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_24_fig10.jpg","caption":"Fig.10. Selective etching of TiN at  $250^{\\circ}C$  using  $\\mathrm{O_3}$  and HF in the presence of other surrounding materials of  $\\mathrm{Al}_2\\mathrm{O}_3$ $\\mathrm{HfO_2}$ $\\mathrm{ZrO_2}$ $\\mathrm{SiO_2}$  and  $\\mathrm{Si_3N_4}$  . Reproduced from Ref. [20].","id":"train/atomic-layer-etching/simulation-usecase/24/fig10","sample_id":"atomic-layer-etching/simulation-usecase/24/fig10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the film thickness of different materials (TiN, Al₂O₃, Si₃N₄, HfO₂, ZrO₂, and SiO₂) after 0 to 400 cycles. There is a linear decrease for TiN, while the others show a plateau.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | TiN | Al₂O₃ | Si₃N₄ | HfO₂ | ZrO₂ | SiO₂ |\\n|---|---|---|---|---|---|---|\\n| 0 | 75 | - | 40 | 55 | 50 | 50 |\\n| 100 | 50 | 50 | 40 | 55 | 50 | 50 |\\n| 200 | 35 | 50 | 40 | 55 | 50 | 50 |\\n| 300 | 20 | 48 | 40 | 55 | 50 | 50 |\\n| 400 | - | 42 | 40 | 55 | 50 | 50 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The etching of TiN was studied at 250 °C in the presence of other materials such as Al2O3, Si3N4, HfO2, ZrO2, and SiO2, using O2 and HF.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The film thickness of TiN decreases from approximately 75 Å to 20 Å as the number of cycles increases from 0 to 400.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"TiN.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen, the Si3N4 film, shows a slight increase in thickness that could be related to the growth of the material.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":3,"width":650,"height":501}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/images/fig10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/images/fig10.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/Review Paper -- Thermal atomic layer etching Mechanism, materials and prospects.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"24","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":652,"height":509,"image_format":"jpeg","image_sha256":"0f857608ac03d50912cb75a65bd8a1e68093bc9d4fca8e3a4625459d1505d300","metadata_license":"CC BY 4.0","image_license":"source_publisher_rights_reserved","image_reuse_status":"non_commercial_research_use_only","schema_version":"1.0.0"}
{"file_name":"images/train_atomic-layer-etching_simulation-usecase_24_fig_5.jpg","caption":"Fig. 5. Selective etching of  $\\mathrm{Al_2O_3}$ ,  $\\mathrm{HfO_2}$ ,  $\\mathrm{ZrO_2}$ ,  $\\mathrm{SiO_2}$ ,  $\\mathrm{Si_3N_4}$ , and TiN by the metal ALE using fluorination and ligand-exchange reactions: (a)  $\\mathrm{Sn(acac)}_2$  at  $200^{\\circ}\\mathrm{C}$  and (b) TMA at  $300^{\\circ}\\mathrm{C}$ . Reproduced from Ref. [29].","id":"train/atomic-layer-etching/simulation-usecase/24/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/24/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The figure shows the changes on the film thickness of various materials (TiN, SiO_2, Si_3N_4, HfO_2, Al_2O_3, ZrO_2) over cycles with Sn(acac)2 at 200 °C.\"},{\"panel_id\":\"b\",\"text\":\"The figure shows the changes on the film thickness of various materials (TiN, SiO_2, Si_3N_4, HfO_2, Al_2O_3, ZrO_2) over cycles with TMA at 300 °C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Number of Cycles | TiN | SiO_2 | Si_3N_4 | HfO_2 | Al_2O_3 | ZrO_2 |\\n|---|---|---|---|---|---|---|\\n| 0 | 70 | 50 | 40 | 50 | 48 | 50 | \\n| 100 | 70 | 50 | 40 | 48 | 20 | 40 | \\n| 200 | 70 | 50 | 40 | 45 | 0 | 30 | \\n| 300 | - | - | - | - | - | - | \\n| 400 | 70 | 50 | 40 | 38 | 0 | 0 |\"},{\"panel_id\":\"b\",\"text\":\"| Number of Cycles | TiN | SiO_2 | Si_3N_4 | HfO_2 | Al_2O_3 | ZrO_2 |\\n|---|---|---|---|---|---|---|\\n| 0 | 70 | 50 | 40 | 50 | 48 | 50 |\\n| 100 | - | - | - | 48 | 20 | 50 | \\n| 200 | - | - | - | 45 | 0 | 53 | \\n| 300 | - | - | - | - | - | - | \\n| 400 | 75 | 48 | 38 | 35 | 0 | 50 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The film thickness for TiN remains approximately constant for the ALE cycle with Sn(acac)2, but increases for the ALE cycle with TMA.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The film thickness was studied as a function of the number of cycles.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As the number of cycles increases the thickness of the HfO2, ZrO2 and Al2O3 film decreases, but the thickness of the TiN, SiO2 and Si3N4 films remains approximately constant.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"As can be seen in the subfigure a, Al2O3 has the lowest film thickness at 100 and 200 cycles. Then, at 300 cycles, both Al2O3 and ZrO3 exhibit a value of 0.\"}]}]","bbox":[{"panel_id":"b","x":631,"y":4,"width":573,"height":463},{"panel_id":"a","x":1,"y":3,"width":621,"height":467}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/24/Review Paper -- Thermal atomic layer etching Mechanism, materials and prospects.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"24","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:img_caption"},"width":1208,"height":472,"image_format":"jpeg","image_sha256":"14a33fd792402c20fc731655ec88d855f475ca813b8fc61bc58cef54a8b1f9a9","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_10.jpg","caption":"FIG. 10. Atomic percentages of F, O, S, and C at the etched surface from XPS as a function of ALE cycles at  $300^{\\circ}\\mathrm{C}$ . Lines are a guide to the eye.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_10","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the atomic percent of O1s, C1s, F1s, and S2p elements as a function of ALE cycles\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALE Cycles | O1s (at.%) | C1s (at.%) | F1s (at.%) | S2p (at.%) |\\n|---|---|---|---|---|\\n| 0 | 15.5 | 14.3 | 2.0 | 0.0 |\\n| 10 | 10.5 | 5.8 | 5.8 | 2.0 |\\n| 20 | 9.9 | 5.9 | 6.2 | 1.8 |\\n| 30 | 9.0 | 6.3 | 6.0 | 1.7 |\\n| 50 | 9.0 | 6.1 | 5.5 | 2.0 |\\n| 75 | 7.6 | 6.2 | 6.4 | 1.4 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both oxygen and carbon contamination decrease significantly during ALE processing. Oxygen drops from ~16 at.% to ~8 at.% while carbon decreases from ~14 at.% to ~6 at.%. Most of this reduction occurs within the first 10 ALE cycles, after which the concentrations stabilize. The continued decrease in oxygen with additional cycles reflects removal of the initially oxidized surface layer present on the as-deposited GaN film.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Fluorine increases due to residual fluorination from the SF6 plasma half-cycle used in the ALE process.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"No. The sulfur content remains below 2 at.% throughout the ALE process, indicating minimal S incorporation from the SF6 plasma.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Approximately 6 at.%.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":4,"width":674,"height":516}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_10.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 scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":523,"image_format":"jpeg","image_sha256":"7d4203d2e3fc71c45b8db07875f185d4938aca0c30b59898cbbec80ccec72bea","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_5.jpg","caption":"FIG. 5. Saturation curves at 150 and  $300^{\\circ}\\mathrm{C}$  for (a)  $\\mathrm{SF}_6$  plasma exposure using  $10\\times 100\\mathrm{ms}$  TMA exposures and (b) varied number of  $100\\mathrm{ms}$  TMA pulses with a  $10\\mathrm{s}$ $\\mathrm{SF}_6$  plasma step, determined over 30 ALE cycles. Lines are guides for the eye.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Saturation curve for the SF6 plasma half-cycle showing EPC vs. plasma exposure time at 150°C and 300°C. Both temperatures show rapid initial increase followed by soft-saturation, reaching ~0.31 nm/cycle at 150°C and ~0.40 nm/cycle at 300°C after 10 s exposure.\"},{\"panel_id\":\"b\",\"text\":\"Saturation curve for the TMA half-cycle showing EPC vs. number of TMA pulses at 150°C and 300°C. Clear saturation is achieved at 300°C with 4 pulses (~0.40 nm/cycle), while 150°C requires more pulses to reach saturation (~0.31 nm/cycle).\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| SF₆ plasma time (s) | 300°C (nm/cycle) | 150°C (nm/cycle) |\\n|---|---|---|\\n| 0 | 0.00 | 0.00 |\\n| 1 | 0.26 | - |\\n| 2 | 0.33 | 0.16 |\\n| 5 | 0.38 | 0.27 |\\n| 10 | 0.40 | 0.31 |\\n| 15 | 0.40 | 0.32 |\\n| 30 | 0.42 | 0.35 |\"},{\"panel_id\":\"b\",\"text\":\"| TMA pulses | 300°C (nm/cycle) | 150°C (nm/cycle) |\\n|---|---|---|\\n| 0 | 0.00 | 0.00 |\\n| 1 | -0.01 | - |\\n| 2 | 0.39 | 0.02 |\\n| 3 | 0.40 | - |\\n| 4 | 0.40 | 0.27 |\\n| 5 | 0.40 | 0.25 |\\n| 6 | - | 0.31 |\\n| 7 | 0.40 | - |\\n| 10 | 0.40 | 0.31 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Lower temperature inhibits fluorine diffusion into the GaN, resulting in a thinner fluorinated layer and thus a lower EPC.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"10 s was chosen because the EPC increases by only 0.02 nm/cycle when extending to 30 s, which is within experimental error.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"0.40 ± 0.02 nm/cycle.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The SF6 plasma step shows soft-saturation with a gradual increase in EPC even at longer exposure times, attributed to diffusion-limited fluorination following Deal-Grove kinetics. The TMA step shows clear, sharp saturation once sufficient pulses are provided. This difference reflects the distinct mechanisms: diffusion-controlled surface modification for fluorination vs. self-limiting ligand-exchange reactions for TMA.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":5,"width":369,"height":491},{"panel_id":"b","x":380,"y":16,"width":291,"height":481}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_5.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 scatter plot","caption_source":"content.json:image_caption"},"width":672,"height":497,"image_format":"jpeg","image_sha256":"1e113ad972d50cdcbb71c899949dd8438e8817b409c65ad77c1e2e150dd82304","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_7.jpg","caption":"FIG. 7. GaN EPC as a function of table temperature. Data for  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE using the same  $\\mathrm{SF}_6$  plasma/TMA chemistry are shown from our previous work. Thermal GaN ALE with  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3$  is also included in the plot. Lines are guides to the eye.","id":"train/atomic-layer-etching/simulation-usecase/27/fig_7","sample_id":"atomic-layer-etching/simulation-usecase/27/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the EPC (nm/cycle) of Plasma ALE GaN, Plasma ALE Al₂O₃, and Thermal ALE GaN at various temperatures. The data points are connected by lines, indicating trends over temperature\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Plasma ALE GaN (nm/cycle) | Plasma ALE Al₂O₃ (nm/cycle) | Thermal ALE GaN (nm/cycle) |\\n|---|---|---|---|\\n| 100 | -0.01 | - | - |\\n| 125 | 0.21 | - | - |\\n| 150 | 0.33 | - | - |\\n| 170 | - | - | 0.02 |\\n| 175 | 0.35 | -0.01 | - |\\n| 185 | - | - | 0.03 |\\n| 200 | 0.36 | 0.02 | 0.05 |\\n| 225 | 0.37 | 0.08 | - |\\n| 250 | 0.39 | 0.21 | 0.06 |\\n| 275 | 0.40 | 0.27 | - |\\n| 300 | 0.43 | 0.29 | 0.07 |\\n| 325 | - | 0.30 | - |\\n| 350 | - | 0.31 | - |\\n| 375 | - | 0.34 | - |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Plasma ALE provides significantly higher etch rates (0.31-0.43 nm/cycle vs. ~0.07 nm/cycle for thermal ALE) and operates over a wider temperature range. This is attributed to the higher reactivity of fluorine radicals compared to thermal XeF2, enabling more effective fluorination of the GaN surface. The wider temperature window also provides greater process flexibility.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Between 150 and 200°C.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"AlF3 ALD growth dominates over the ALE reaction, resulting in net film deposition instead of etching.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"It increases linearly with temperature from 0.31 nm/cycle at 150°C to 0.43 nm/cycle at 300°C.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":5,"width":663,"height":544}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/27/images/fig_7.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 scatter plot","caption_source":"content.json:image_caption"},"width":666,"height":552,"image_format":"jpeg","image_sha256":"67ca83bd80303cdca4119a15ca5d627dc21e8757ed79517f3c28976757f7663b","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_32_fig_5.jpg","caption":"FIG. 5. The reaction energy variations (a) and the maximum reaction energies (b) obtained from the sequential etching events of routes 1-3.","id":"train/atomic-layer-etching/simulation-usecase/32/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/32/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"bar chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart illustrates the energy changes at various reaction steps for three routes (Route 1, Route 2, and Route 3). Each route shows a distinct trend in energy increase.\"},{\"panel_id\":\"b\",\"text\":\"The bar chart compares the maximum energy achieved by three routes (Route 1, Route 2, and Route 3). Route 3 has the highest maximum energy\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Reaction Step | Route 1 (eV) | Route 2 (eV) | Route 3 (eV) |\\n|---|---|---|---|\\n| 0 | 0.0 | 0.0 | 0.0 |\\n| 1 | 5.4 | 4.0 | 13.0 |\\n| 2 | 7.8 | 11.5 | 11.0 |\\n| 3 | 10.8 | 13.5 | 22.0 |\\n| 4 | - | 20.5 | 29.0 |\\n| 5 | - | 22.5 | 34.0 |\"},{\"panel_id\":\"b\",\"text\":\"| Route | Maximum Energy (eV) |\\n|---|---|\\n| Route 1 | 5.4 |\\n| Route 2 | 7.3 |\\n| Route 3 | 12.9 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Route 1 uses squares, Route 2 uses circles, Route 3 uses triangles.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"They denote the surface moiety or atom removed at that etching event in the sequence for each route.\"}]},{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Route 3 has the highest maximum reaction energy.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Routes 1 and 2 represent ALET on a BCl3 adsorbed surface and have much lower maximum reaction energies than route 3, which represents conventional plasma etching of pristine Al2O3. This supports the paper’s conclusion that BCl3 presorption reduces the energetic requirement for etching, which in turn can reduce the required Ar kinetic energy for effective removal. The authors also note that changing the etching event sequence can raise the maximum energy toward the conventional plasma etching level, so sequence control matters.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":606,"height":423},{"panel_id":"b","x":96,"y":435,"width":421,"height":387}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/32/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/32/images/fig_5.json","source_content":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/32/content.json","source_pdf":["icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/32/Understanding time-resolved processes in atomic-layer etching of ultra-thin Al2O3 film.pdf"],"main_category":"atomic-layer-etching","sub_category":"simulation-usecase","paper_id":"32","first_classification_panel_id":"a","first_classification_label":"multiple scatter plot","caption_source":"content.json:image_caption"},"width":609,"height":825,"image_format":"jpeg","image_sha256":"d49a97aac60869c5ac255f91ba8ba448bb51b250a6d509a1acad1bb4227af237","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_2bf83ee5d0350bda7054684c907c864234acb978fe3ffba9b23eba42a75a8833.jpg","caption":"(a)","id":"train/atomic-layer-etching/simulation-usecase/33/2bf83ee5d0350bda7054684c907c864234acb978fe3ffba9b23eba42a75a8833","sample_id":"atomic-layer-etching/simulation-usecase/33/2bf83ee5d0350bda7054684c907c864234acb978fe3ffba9b23eba42a75a8833","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure depicts the evolution of surface-bound species on silicon under increasing HF (hydrofluoric acid) dose, showing the number density per square nanometer (Nᵢ) for unfluorinated Si and various fluorinated forms: SiF, SiF₂, and SiF₃. Initially, the surface is composed entirely of unfluorinated Si species (~28 nm⁻²), but with increasing HF exposure, the density of fluorinated species grows. SiF appears first, followed by SiF₂ and SiF₃, suggesting a stepwise fluorination mechanism. The density of each fluorinated species eventually saturates, while the total Si-related surface density remains roughly constant, indicating that fluorination modifies the surface rather than etching it away.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Dose (HF/nm²) | Si (nm⁻²) | SiF (nm⁻²) | SiF₂ (nm⁻²) | SiF₃ (nm⁻²) |\\n|---------------|-----------|------------|-------------|-------------|\\n| 0             | 28        | 0          | 0           | 0           |\\n| 50            | 26        | 5          | 3           | 2           |\\n| 100           | 26        | 7          | 4           | 3           |\\n| 150           | 26        | 8          | 4           | 3           |\\n| 200           | 26        | 8          | 4           | 3           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"As the HF dose increases, the number density of unfluorinated surface Si atoms decreases, while the densities of fluorinated species (SiF, SiF₂, and SiF₃) increase. The total surface species density remains approximately constant, suggesting a substitutional process where fluorine atoms progressively bond with silicon to form higher fluorinated states. SiF forms first and dominates at lower doses, followed by SiF₂ and SiF₃, which appear and saturate at higher doses\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"SiF.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Surface fluorination proceeds in a stepwise manner, Si to SiF to SiF₂ to SiF₃, SiF is dominant at low HF doses, SiF₂ and SiF₃ emerge and saturate later\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":295,"height":222}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/2bf83ee5d0350bda7054684c907c864234acb978fe3ffba9b23eba42a75a8833.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/2bf83ee5d0350bda7054684c907c864234acb978fe3ffba9b23eba42a75a8833.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 scatter plot","caption_source":"content.json:image_caption"},"width":295,"height":222,"image_format":"jpeg","image_sha256":"93e8bdd52c6f7573ed91a82727623170b483a917d47054fa6fe48bca72966ecb","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_12.jpg","caption":"Figure 12. Evolution of populations of surface species  $(N_{i})$  in the modified layer as a function of HF dose. The solid lines are the results of 1D continuum model while symbols are directly obtained from MD simulations with  $E_{\\mathrm{in}}$  of  $50~\\mathrm{eV}$  and  $\\theta_{\\mathrm{in}}$  of  $0^{\\circ}$ . Panel (a) depicts the population change of silicon fluoride species, namely  $\\mathrm{SiF_0}$ ,  $\\mathrm{SiF_1}$ ,  $\\mathrm{SiF_2}$ , and  $\\mathrm{SiF_3}$ . Panel (b) illustrates the changes in ammonia-derived species, specifically  $\\mathrm{NH_0}$ ,  $\\mathrm{NH_1}$ , and  $\\mathrm{NH_2}$ . Panel (c) presents the ratio of N to Si, with the dashed line indicating the bulk stoichiometry of 1.33. (b) (c)","id":"train/atomic-layer-etching/simulation-usecase/33/figure_12","sample_id":"atomic-layer-etching/simulation-usecase/33/figure_12","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"This figure plots the evolution of various nitrogen-containing surface species as a function of HF dose, measured in molecules per square nanometer (Nᵢ, nm⁻²). The total nitrogen concentration (N) initially drops with HF exposure, suggesting removal or conversion, but then stabilizes around 30 nm⁻². NH₀ (neutral nitrogen) shows a slight decline, while NH₁ and NH₂ exhibit clear increasing trends. This suggests that as HF dose increases, nitrogen becomes increasingly hydrogenated, forming amine-like species. The saturation of NH₁ and NH₂ levels beyond ~100 HF/nm² implies a steady-state surface composition. The figure highlights the chemical evolution of surface-bound nitrogen species during HF exposure.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Dose (HF/nm²) | N (nm⁻²) | NH₀ (nm⁻²) | NH₁ (nm⁻²) | NH₂ (nm⁻²) |\\n|---------------|-----------|-------------|-------------|-------------|\\n| 0             | 35        | 10          | 0           | 0           |\\n| 50            | 30        | 8           | 5           | 2           |\\n| 100           | 28        | 7           | 7           | 3           |\\n| 150           | 30        | 7           | 9           | 4           |\\n| 200           | 32        | 8           | 9           | 5           |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The figure shows that atomic nitrogen (N) initially decreases in surface concentration with increasing HF dose but then plateaus, indicating partial consumption or transformation. In contrast, NH₁ and NH₂ species increase steadily with dose, suggesting progressive hydrogenation of nitrogen as HF exposure continues. NH₀ decreases slightly over time. These trends point to a surface reaction pathway where nitrogen becomes increasingly substituted with hydrogen, possibly due to HF-mediated surface reactions leading to NHₓ termination.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"NH₁.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"HF exposure hydrogenates nitrogen , Atomic nitrogen partially converts to NH₁ and NH₂ species, Surface nitrogen composition stabilizes beyond ~100 HF/nm².\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":298,"height":225}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/33/images/figure_12.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 scatter plot","caption_source":"content.json:image_caption"},"width":298,"height":225,"image_format":"jpeg","image_sha256":"bec9ee4d32b85c2e0d83cc5cb1122ac0c9bd7d4039f2421cc721c583e841908d","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_fig_12.jpg","caption":"FIG. 12. Normalized atomic Si counts as a function of  $\\mathsf{Ar}^+$  fluence for various  $\\mathsf{Ar}^+$  energies. Subfigure (a) shows the results for OES experiments (using the modified OES Si signals) and subfigure (b) shows results from MD ALE simulations. For both the OES and MD results, the counts are normalized using the value at the plateau observed at  $80 \\text{eV}$ .","id":"train/atomic-layer-etching/simulation-usecase/6/fig_12","sample_id":"atomic-layer-etching/simulation-usecase/6/fig_12","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"Panel (a) shows experimental measurements of Si count as a function of Ar⁺ ion dosage for three incident energies. The Si signal is consistently highest at 215 eV and increases with dosage before saturating, while the 80 eV signal shows a moderate rise and 45 eV shows negligible counts.\"},{\"panel_id\":\"b\",\"text\":\"Panel (b) presents Monte Carlo simulation results for the same ion energies and dosages. The simulated Si counts for 215 eV also show a plateauing behavior at higher dosages, though with more fluctuations and slightly different relative magnitudes compared to the experimental data\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Dosage (Ar⁺/cm²) × 10¹⁵ | 45 eV | 80 eV | 215 eV |\\n|----------------------------|-------|-------|--------|\\n| 0                          | 0     | 0     | 2      |\\n| 10                         | 0     | 0.5   | 4      |\\n| 20                         | 0     | 1     | 5      |\\n| 30                         | 0     | 1.5   | 6      |\\n| 40                         | 0     | 2     | 7      |\\n| 50                         | 0     | 2.5   | 8      |\"},{\"panel_id\":\"b\",\"text\":\"| Ion Dosage (Ar⁺/cm²) × 10¹⁵ | 45 eV | 80 eV | 215 eV |\\n|----------------------------|-------|-------|--------|\\n| 0                          | 5     | 0.5   | 6      |\\n| 10                         | 6     | 1     | 7      |\\n| 20                         | 7     | 1.5   | 8      |\\n| 30                         | 6     | 2     | 7      |\\n| 40                         | 7     | 2.5   | 6      |\\n| 50                         | 6     | 3     | 6      |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"215 eV\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"More fluctuation in simulation, Clearer plateau in experiment, Smoother experimental curves with simulation error bars\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"between experiment and simulation allows researchers to validate modeling accuracy and adjust physical parameters like ion energy and dose for surface processing. Since 215 eV consistently leads to higher and saturating Si counts, this suggests it may be the most effective energy for maximising surface modification efficiency.\"}]}]","bbox":[{"panel_id":"a","x":8,"y":2,"width":679,"height":458},{"panel_id":"b","x":721,"y":3,"width":687,"height":457}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_12.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_12.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 scatter plot","caption_source":"content.json:img_caption"},"width":1409,"height":461,"image_format":"jpeg","image_sha256":"439d19372330cea38a84c74679b08c5ad0c1abf44dd3695599bf2d8a19bd3818","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_fig_13.jpg","caption":"FIG. 13. Normalized SiCl counts as a function of  $\\mathsf{Ar}^+$  fluence for various  $\\mathsf{Ar}^+$  energies. Subfigure (a) shows the results for OES experiments and subfigure (b) shows results from MD ALE simulations. For both the OES and MD results, the counts are normalized using the value at the peak observed at  $80 \\text{eV}$ .","id":"train/atomic-layer-etching/simulation-usecase/6/fig_13","sample_id":"atomic-layer-etching/simulation-usecase/6/fig_13","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The pair of scatter plots show the experimental (left) and MD-simulated (right) SiCl count as a function of Ar⁺ ion dosage for three ion energies: 45 eV, 80 eV, and 215 eV. In both cases, the SiCl count decreases with increasing dosage, with the highest initial counts observed at 215 eV. The simulation matches the trend seen experimentally, though the absolute counts differ. Error bars are shown for each data point\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Dosage (Ar<sup>+</sup>/cm<sup>2</sup>) × 10<sup>15</sup> | 45 eV | 80 eV | 215 eV |\\n|---|---|---|---|\\n| 0 | 0.75 | 0.90 | 1.60 |\\n| 5 | 0.60 | 0.70 | 0.80 |\\n| 10 | 0.55 | 0.60 | 0.70 |\\n| 15 | 0.50 | 0.55 | 0.60 |\\n| 20 | 0.45 | 0.50 | 0.55 |\\n| 25 | 0.40 | 0.45 | 0.50 |\\n| 30 | 0.35 | 0.40 | 0.45 |\\n| 35 | 0.30 | 0.35 | 0.40 |\\n| 40 | 0.25 | 0.30 | 0.35 |\\n| 45 | 0.20 | 0.25 | 0.30 |\\n| 50 | 0.15 | 0.20 | 0.25 |\"},{\"panel_id\":\"b\",\"text\":\"| Ion Dosage (Ar<sup>+</sup>/cm<sup>2</sup>) × 10<sup>15</sup> | 45 eV | 80 eV | 215 eV |\\n|---|---|---|---|\\n| 0 | 1.80 | 1.00 | 1.70 |\\n| 5 | 0.50 | 0.40 | 0.30 |\\n| 10 | 0.30 | 0.25 | 0.20 |\\n| 15 | 0.20 | 0.15 | 0.10 |\\n| 20 | 0.15 | 0.10 | 0.05 |\\n| 25 | 0.10 | 0.05 | 0.02 |\\n| 30 | 0.05 | 0.02 | 0.01 |\\n| 35 | 0.02 | 0.01 | 0.00 |\\n| 40 | 0.01 | 0.00 | 0.00 |\\n| 45 | 0.00 | 0.00 | 0.00 |\\n| 50 | 0.00 | 0.00 | 0.00 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Ion dosage (Ar⁺/cm² × 10¹⁵), Ion energy (45 eV, 80 eV, 215 eV), Simulation vs. experimental conditions\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"215 eV\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The reduction in SiCl count with increasing ion dosage suggests a depletion of reactive surface species over time, which can lead to reduced etch efficiency or modification effectiveness during plasma processing. This implies that optimising ion energy and controlling dosage are crucial for sustaining high etch rates or achieving consistent surface functionalistion, especially in processes where product formation (like SiCl) is critical for chemical transport or material removal.\"}]}]","bbox":[{"panel_id":"a","x":10,"y":10,"width":682,"height":448},{"panel_id":"b","x":725,"y":10,"width":686,"height":448}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_13.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_13.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 scatter plot","caption_source":"content.json:img_caption"},"width":1411,"height":458,"image_format":"jpeg","image_sha256":"71675471e5b0aa7813bf2f4b7d1241db9155b047e40e258cd3a1ada3b0030d18","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_fig_14.jpg","caption":"FIG. 14. Normalized  $\\mathrm{SiCl}_2$  counts as a function of  $\\mathsf{Ar}^+$  fluence for various  $\\mathsf{Ar}^+$  energies. Subfigure (a) shows the results for OES experiments and subfigure (b) shows results from MD ALE simulations. For both the OES and MD results, the counts are normalized using the value at the peak observed at  $80\\mathrm{eV}$","id":"train/atomic-layer-etching/simulation-usecase/6/fig_14","sample_id":"atomic-layer-etching/simulation-usecase/6/fig_14","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The experimental scatter plot shows that the SiCl₂ count decreases as ion dosage increases for all energy levels (45 eV, 80 eV, 215 eV). The highest counts are seen at 215 eV, while the lowest are at 45 eV. After ~30×10¹⁵ Ar⁺/cm², the counts plateau, indicating a saturation or steady-state regime.\"},{\"panel_id\":\"b\",\"text\":\"The MD simulation panel reflects a similar trend to the experiment: initial steep decrease in SiCl₂ count, more prominent at higher energies. Notably, the simulated values begin at a slightly higher range than experimental values and converge more rapidly with dosage\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Dosage (Ar⁺/cm²) × 10¹⁵ | 45 eV  | 80 eV  | 215 eV  |\\n|-----------------------------|--------|--------|---------|\\n| 0                           | 1.2    | 1.4    | 1.6     |\\n| 5                           | 0.9    | 1.1    | 1.3     |\\n| 10                          | 0.7    | 0.9    | 1.1     |\\n| ...                         | ...    | ...    | ...     |\"},{\"panel_id\":\"b\",\"text\":\"| Ion Dosage (Ar⁺/cm²) × 10¹⁵ | 45 eV  | 80 eV  | 215 eV  |\\n|-----------------------------|--------|--------|---------|\\n| 0                           | 1.30   | 1.75   | 2.1     |\\n| 5                           | 0.80   | 1.20   | 1.4     |\\n| 10                          | 0.55   | 0.85   | 1.0     |\\n| 15                          | 0.45   | 0.70   | 0.85    |\\n| 20                          | 0.40   | 0.60   | 0.78    |\\n| 25                          | 0.35   | 0.55   | 0.72    |\\n| 30                          | 0.30   | 0.50   | 0.65    |\\n| 35                          | 0.28   | 0.48   | 0.60    |\\n| 40                          | 0.27   | 0.46   | 0.58    |\\n| 45                          | 0.26   | 0.45   | 0.56    |\\n| 50                          | 0.25   | 0.45   | 0.55    |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Ion energy transfer, Surface atom displacement, SiCl₂ formation and ejection, Yield reduction with dose\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"No\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"In both experimental and simulated datasets, the SiCl₂ count decreases with increasing ion dosage for all ion energies. The rate of decrease is steepest between 0 and 15×10¹⁵ Ar⁺/cm², after which the counts stabilize, indicating a saturation effect. Higher ion energies (e.g., 215 eV) consistently produce more SiCl₂ than lower energies like 45 eV, reflecting enhanced sputtering or reaction yields under more energetic conditions.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"215 eV\"}]}]","bbox":[{"panel_id":"a","x":9,"y":10,"width":683,"height":451},{"panel_id":"b","x":725,"y":9,"width":684,"height":452}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_14.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_14.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 scatter plot","caption_source":"content.json:img_caption"},"width":1409,"height":461,"image_format":"jpeg","image_sha256":"ee1cace633e6b9d3fe928416e7fe4aade6bf506addaa6f6803854fd65d45fb11","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_fig_8.jpg","caption":"FIG. 8. Etch per cycle (EPC) in units of nm/cycle for experiments (black squares) and MD simulations (red triangles) as a function of  $\\mathsf{Ar}^+$  energy. Error bars represent  $95\\%$  confidence intervals.","id":"train/atomic-layer-etching/simulation-usecase/6/fig_8","sample_id":"atomic-layer-etching/simulation-usecase/6/fig_8","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The scatter plot compares the EPC (etching per cycle) values measured experimentally and via MD simulations across a range of ion energies. Both datasets exhibit a positive correlation between ion energy and EPC, indicating that higher ion energies lead to more effective etching. However, the experimental EPC values increase more rapidly and reach higher maximum values than those predicted by MD simulations. This suggests that simulations may underestimate etch rates due to idealised conditions, or that experimental conditions involve synergistic effects not captured in the model. The discrepancy becomes more pronounced above ~100 eV, highlighting the importance of validating simulations against experimental benchmarks\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Ion Energy (eV) | EPC (Experiments, nm/cycle) | EPC (MD Simulations, nm/cycle) |\\n|-----------------|-----------------------------|---------------------------------|\\n| 0               | 0.0                         | 0.0                             |\\n| 25              | 0.3                         | 0.1                             |\\n| 50              | 0.6                         | 0.2                             |\\n| 75              | 1.2                         | 0.6                             |\\n| 100             | 2.2                         | 1.2                             |\\n| 125             | 3.2                         | 1.8                             |\\n| 150             | 4.3                         | 2.2                             |\\n| 175             | 5.5                         | 2.6                             |\\n| 200             | 6.4                         | 3.0                             |\\n| 225             | 7.0                         | 3.4                             |\\n| 250             | 7.5                         | 3.8                             |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Both experimental and simulation results show increasing EPC with rising ion energy, indicating enhanced etching efficiency. However, the experimental data consistently show higher EPC values than the simulations, especially beyond 100 eV. This suggests either additional synergistic effects in the experiment or limitations in the simulation model, such as neglect of secondary ion effects or surface roughness.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Around 100 eV\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"List\",\"answer\":\"Higher momentum transfer, Increased sputtering yield, Enhanced bond breaking, Deeper penetration for bulk removal.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Yes/No\",\"answer\":\"Yes\"}]}]","bbox":[{"panel_id":"a","x":0,"y":0,"width":541,"height":406}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_8.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/6/images/fig_8.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 scatter plot","caption_source":"content.json:img_caption"},"width":541,"height":406,"image_format":"jpeg","image_sha256":"ef965570706f3f426f8d50c050d0f7c76aedd7bf6ba623ebdf0504a6399feec2","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_10.jpg","caption":"FIG. 10. Atomic percentages of F, O, S, and C at the etched surface from XPS as a function of ALE cycles at  $300^{\\circ}\\mathrm{C}$ . Lines are a guide to the eye.","id":"train/atomic-layer-etching/simulation-usecase/8/fig_10","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_10","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the atomic percent of O1s, C1s, F1s, and S2p elements as a function of ALE cycles.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| ALE cycles | O1s | C1s | F1s | S2p |\\n|---|---|---|---|---|\\n| 0 | 15 | 14 | 2 | 0 |\\n| 10 | 10 | 6 | 6 | 2 |\\n| 20 | 10 | 6 | 6 | 2 |\\n| 30 | 9 | 6 | 6 | 2 |\\n| 50 | 9 | 6 | 6 | 2 |\\n| 70 | 8 | 6 | 6 | 2 |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The ALE process removes oxygen and carbon species.\"},{\"question_type\":\"Application/Performance\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The atomic percentage together do not add up till 100%, this means that there are more species in the film not in the graph.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"30 ALE cycles are required.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It is likely that gases such as SF6 are used which remain in the film.\"}]}]","bbox":[{"panel_id":"a","x":3,"y":2,"width":666,"height":517}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_10.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_10.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 scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":523,"image_format":"jpeg","image_sha256":"7d4203d2e3fc71c45b8db07875f185d4938aca0c30b59898cbbec80ccec72bea","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_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/8/fig_3","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_3","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multi spectra chart"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the peak area of XPS signals for Ga-N and Ga-O as a function of sputter depth.\"},{\"panel_id\":\"b\",\"text\":\"The spectrum displays the XPS counts at various binding energies for Ga-N and Ga-O.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Sputter depth (nm)| Peak area (CPS eV) Ga-O| Peak area (CPS eV) Ga-N|\\n|---|---|---|\\n| 0 | 4000 |0|\\n|5|1300|4500|\\n|10|1000|5000|\\n|15|1000|5000|\\n|20|1000|5000|\\n|25|0|5700|\\n|30|0|6000|\\n|35|0|5700|\\n|40|0|500|\"},{\"panel_id\":\"b\",\"text\":\"| Binding energy (eV) | XPS counts (a.u.) |\\n|---|---|\\n| 14 | 0 |\\n| 16 | 0 |\\n| 18 | 0.3 |\\n| 20 | 2 |\\n| 22 | 0.3 |\\n| 24 | 0 |\\n| 26 | 0 |\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The envelope, which is a result of the two contributions together does not fully overlap with the measured spectrum indicating that the fit is not fully accurate.\"},{\"question_type\":\"Structure-Property\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"Nitrogen is more present, at least in the form where it is bonded to gallium.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"From 25 nm and deeper.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Depth profiles with XPS are known to influence material properties. During the sputtering binding states can change and shift, so it is possible that the film has been altered due to the measurement.\"}]}]","bbox":[{"panel_id":"a","x":0,"y":12,"width":346,"height":427},{"panel_id":"b","x":384,"y":12,"width":282,"height":427}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_3.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_3.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 scatter plot","caption_source":"content.json:img_caption"},"width":666,"height":439,"image_format":"jpeg","image_sha256":"0fbb2af112ef01912077536debaaf5db30ba88d38524a44e12689328bc495902","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_5.jpg","caption":"FIG. 5. Saturation curves at 150 and  $300^{\\circ}\\mathrm{C}$  for (a)  $\\mathrm{SF}_6$  plasma exposure using  $10\\times 100\\mathrm{ms}$  TMA exposures and (b) varied number of  $100\\mathrm{ms}$  TMA pulses with a  $10\\mathrm{s}$ $\\mathrm{SF}_6$  plasma step, determined over 30 ALE cycles. Lines are guides for the eye.","id":"train/atomic-layer-etching/simulation-usecase/8/fig_5","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_5","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"},{"panel_id":"b","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The line chart shows the relationship between SF₆ plasma time and EPC (nm/cycle) at two temperatures, 300°C and 150°C.\"},{\"panel_id\":\"b\",\"text\":\"The line chart illustrates the relationship between TMA pulses and EPC (nm/cycle) at two temperatures, 300°C and 150°C.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| SF₆ plasma time (s) | EPC (nm/cycle) 300 degrees |EPC (nm/cycle) 150 degrees |\\n|---|---|---|\\n| 0 | 0.0 |0.0|\\n| 5 | 0.37 |0.27|\\n| 10 | 0.4 |0.3|\\n| 15 | 0.41 |0.31|\\n| 30 | 0.41 |0.35|\"},{\"panel_id\":\"b\",\"text\":\"| TMA pulses | EPC (nm/cycle) 300 degrees |EPC (nm/cycle) 150 degrees |\\n|---|---|---|\\n| 0 | 0.4 |0.0|\\n| 2 | 0.4 |0.05|\\n| 4 | 0.4 |0.25|\\n| 6 | 0.4 |0.31|\\n| 10 | 0.4 |0.31|\"}]","vqa":"[{\"panel_id\":\"b\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The main physical parameter that is changed upon increasing the number of TMA pulses which is the partial pressure of TMA. Increasing this will increase the reaction possibility.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"The maximum EPC is 0.31 nm/cycle.\"}]},{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"It could be that the penetration depth of the radical species increases for increasing temperature.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"There is no straight line for increasing plasma time, this looks like soft-saturation.\"}]}]","bbox":[{"panel_id":"b","x":387,"y":4,"width":284,"height":491},{"panel_id":"a","x":1,"y":4,"width":371,"height":491}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_5.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_5.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 scatter plot","caption_source":"content.json:img_caption"},"width":672,"height":497,"image_format":"jpeg","image_sha256":"1e113ad972d50cdcbb71c899949dd8438e8817b409c65ad77c1e2e150dd82304","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_7.jpg","caption":"FIG. 7. GaN EPC as a function of table temperature. Data for  $\\mathrm{Al}_2\\mathrm{O}_3$  ALE using the same  $\\mathrm{SF}_6$  plasma/TMA chemistry are shown from our previous work. Thermal GaN ALE with  $\\mathrm{XeF}_2 / \\mathrm{BCl}_3$  is also included in the plot. Lines are guides to the eye.","id":"train/atomic-layer-etching/simulation-usecase/8/fig_7","sample_id":"atomic-layer-etching/simulation-usecase/8/fig_7","subset":"multiple-scatter-plot","split":"train","classification":[{"panel_id":"a","label":"multiple scatter plot"}],"summarization":"[{\"panel_id\":\"a\",\"text\":\"The chart shows the EPC (nm/cycle) of Plasma ALE GaN, Plasma ALE Al₂O₃, and Thermal ALE GaN at various temperatures. The data points are connected by lines, indicating trends over temperature.\"}]","data_extraction":"[{\"panel_id\":\"a\",\"text\":\"| Temperature (°C) | Plasma ALE GaN | Plasma ALE Al₂O₃ | Thermal ALE GaN |\\n|-----------------|----------------|------------------|----------------|\\n| 100            | 0.0            | 0.0              | 0.0            |\\n| 150            | 0.2            | 0.0              | 0.0            |\\n| 200            | 0.3            | 0.0              | 0.05           |\\n| 250            | 0.4            | 0.2              | 0.1            |\\n| 300            | 0.4            | 0.3              | 0.1            |\\n| 350            | 0.4            | 0.3              | 0.1            |\"}]","vqa":"[{\"panel_id\":\"a\",\"items\":[{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Yes there is a selectivity window between 100 and 175 degrees celcius, so GaN can be etched selectively wrt Al2O3.\"},{\"question_type\":\"Process-Oriented\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"The plasma process yields way higher etch per cycle for lower temperatures.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Paragraph\",\"answer\":\"Increase the temperature when etching Al2O3 instead of GaN.\"},{\"question_type\":\"Comparative/Trend\",\"questions\":\"\",\"answer_type\":\"Factoid\",\"answer\":\"375 degrees would yield the highest EPC.\"}]}]","bbox":[{"panel_id":"a","x":4,"y":4,"width":659,"height":547}],"source":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_7.jpg","provenance":{"source_annotation":"icdar2026-competition-data/train/atomic-layer-etching/simulation-usecase/8/images/fig_7.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 scatter plot","caption_source":"content.json:img_caption"},"width":666,"height":552,"image_format":"jpeg","image_sha256":"67ca83bd80303cdca4119a15ca5d627dc21e8757ed79517f3c28976757f7663b","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"}