Experimentally Grounded Machine Learning of PrecursorSurface-Reactant Chemistry in Atomic Layer Deposition
Abstract
Whether a precursor-substrate-reactant combination yields an atomic layer deposition (ALD) process is still settled by trial-and-error experiments. Here we introduce AJIN (ALD Joint Interpretable Network), a machine-learning framework that learns this decision from experimental outcomes. We compiled 560 experimentally observed ALD and area-selective ALD outcomes spanning 76 precursors, 27 substrates, 20 reactants and 25-620 degrees Celsius. Following the ALD cycle, AJIN predicts adsorption from graph-neural-network embeddings of the precursor and substrate with temperature, and deposition by adding the reactant. The best model (CHGNet-MLP) reaches balanced accuracies of 79.2% for adsorption and 84.8% for deposition and transfers partially to unseen precursors. SHAP and ablation analyses show that substrate chemistry governs adsorption, whereas precursor, substrate and reactant all contribute to deposition. Tested against our laboratory's ALD experiments, AJIN correctly predicted 83% of new-substrate cases and 71% of new precursor-substrate adsorption cases, providing an experimentally grounded tool for prioritizing ALD and area-selective ALD chemistries.
Related
- SCO ATLAS: Mapping Spin-Crossover Thermodynamics across Experimentally Grounded Fe(II)N6 Complexes
- Surface chemistry as data for level-set simulation of etching and deposition
- ZnO Nanowires by Chemical Bath Deposition: A Review on Nucleation, Growth, and Green Chemistry
- Chemistry-Informed Transfer and Interaction-Resolved Learning Towards Data-Efficient Materials Exploration
- Physics-grounded machine learning for inverse design of TiO2/CuInGaS2/rGO–CuxS quantum-dot solar cells using surrogate modeling and SCAPS-based validation
BibTeX
@misc{gu2026experimentally,
title = {{Experimentally Grounded Machine Learning of PrecursorSurface-Reactant Chemistry in Atomic Layer Deposition}},
author = {Bonwook Gu and Trinh Ngoc Le and Zunair Masroor and Minhyeok Lee and Han-Bo-Ram Lee},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.10024},
eprint = {2611.10024},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.10024},
}
RIS
TY - UNPB
TI - Experimentally Grounded Machine Learning of PrecursorSurface-Reactant Chemistry in Atomic Layer Deposition
AU - Bonwook Gu
AU - Trinh Ngoc Le
AU - Zunair Masroor
AU - Minhyeok Lee
AU - Han-Bo-Ram Lee
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.10024
UR - https://arxiv.org/abs/2611.10024
AB - Whether a precursor-substrate-reactant combination yields an atomic layer deposition (ALD) process is still settled by trial-and-error experiments. Here we introduce AJIN (ALD Joint Interpretable Network), a machine-learning framework that learns this decision from experimental outcomes. We compiled 560 experimentally observed ALD and area-selective ALD outcomes spanning 76 precursors, 27 substrates, 20 reactants and 25-620 degrees Celsius. Following the ALD cycle, AJIN predicts adsorption from graph-neural-network embeddings of the precursor and substrate with temperature, and deposition by adding the reactant. The best model (CHGNet-MLP) reaches balanced accuracies of 79.2% for adsorption and 84.8% for deposition and transfers partially to unseen precursors. SHAP and ablation analyses show that substrate chemistry governs adsorption, whereas precursor, substrate and reactant all contribute to deposition. Tested against our laboratory's ALD experiments, AJIN correctly predicted 83% of new-substrate cases and 71% of new precursor-substrate adsorption cases, providing an experimentally grounded tool for prioritizing ALD and area-selective ALD chemistries.
ER -