Data and code for: Mapping Adsorption Selectivity Patterns via Thermodynamics-Aware Machine Learning.

This repository contains the data and computational workflow accompanying the manuscript "Mapping Adsorption Selectivity Patterns via Thermodynamics-Aware Machine Learning" (Soto and Peña, J. Chem. Theory Comput. 2026). Included are: a Google Colab notebook implementing the complete pipeline (structure generation, MACE-MP-0 geometry optimization and NVT molecular dynamics, SOAP descriptor computation, UMAP/HDBSCAN archetype discovery, and a zentropy-inspired neural network with XGBoost baseline); the base Mo-doped graphene structure and a 72-configuration design of experiments for NH₃ adsorption; intermediate data products (optimized structures, MD trajectories, ML predictions, entropy tables) in CSV, JSON, NumPy, and VASP formats; and publication-quality figures.