Curated dataset of thermal stability measurements for atomic layer deposition precursor compounds

# ALD Precursor Thermal Stability Dataset# Version 1.0 ## DescriptionA curated dataset of thermal stability measurements for atomic layer deposition(ALD) precursor compounds. This dataset contains 1,117 records for ALD precursorcompounds, of which 719 (64.4%) include thermal stability measurements, covering939 unique molecular structures and 52 metal and metalloid elements. ## Files- ALD_precursor_raw.csv : Complete raw dataset (1,117 records x 19 columns)- ALD_precursor_cleaned.csv : Cleaned dataset after curation (566 records x 14 columns)- dataset_schema.csv : Column definitions with data types- column_dictionary.csv : Allowed values for categorical columns- source_reference_table.csv : Source provenance lookup table- README.txt : This file ## Data Sources- Literature (peer-reviewed TGA/DSC): 644 records- Safety data sheets (SDS/SADT): 75 records- Public databases (no thermal data): 398 records ## Validation Summary- SMILES validity rate: 939/1,117 (84.1%)- Duplicate SMILES groups: 21 total (7 with decomposition data from multiple sources) - Of the 7 groups with decomp data: range <= 150 C in 5 groups, >= 336 C in 2 groups (flagged as conflicts)- SDS vs. literature decomposition temperature offset: +153 C- Records with T_onset > T_decomp: 109/234 (46.6%)- Conflict threshold (200 C) sensitivity: 150/200/250 C all flag same 2 groups ## Usage Notes1. Temperature type distinction: Onset temperature (t_onset_c) and decomposition temperature (t_decomp_c) represent different physical quantities and should not be combined without careful consideration. 2. SADT filtering: Records with sadt_flag = True report self-accelerating decomposition temperatures measured under fundamentally different conditions from laboratory TGA or DSC. Exclude these for homogeneous analyses. 3. Compound naming: common_name provides the most recognizable non-IUPAC identifier (abbreviation or trade name). iupac_name provides the IUPAC systematic name where available (97% of entries). 4. Source type: "literature" = records with TGA/DSC thermal stability data from peer-reviewed articles; "SDS" = safety data sheet SADT values; "database" = entries from public databases with physical properties but no original thermal stability measurements. 5. Duplicate awareness: duplicate_group_id links records for the same molecule from different sources. Records with conflict_flag = True have temperature ranges exceeding 200 degrees C within the group. 6. Cleaned vs. raw: The cleaned dataset applies conservative quality filters. Users may start from the raw dataset and apply custom criteria using the quality flag columns. 7. Measurement environment: The measurement_environment column is sparsely populated. Measurement atmosphere can influence results by 20-50 degrees C. 8. Metal coverage: Silicon-based precursors dominate (~53% of entries with SMILES). Verify sample sizes for non-silicon metal analyses. 9. Appropriate use cases: Comparative screening, candidate identification, cross-referencing via SMILES, ML with uncertainty quantification. Caution: Absolute temperature prediction is limited by 50-100 C inter-lab variability; SDS values are +153 C offset from literature; subgroups with N < 20 have limited statistical power. ## CitationIf you use this dataset, please cite:[Full citation to be added upon publication] ## LicenseCreative Commons Attribution 4.0 International (CC BY 4.0) ## RepositoryZenodo (10.5281/zenodo.19369338) ## ContactCorresponding author: Kyungjun Kang ([e-mail]) ## Version History- v1.0 (2026): Initial release