Evaluating GRACE and MACE Foundation Models on small Molecules Adsorption in Pure and Cation-Exchanged Siliceous Zeolites.

his repository contains data and resources related to the development and evaluation of machine learning interatomic potentials (MLIPs) for zeolite materials. The contents include: Python script for creating Quantum Espresso input files when the cif/traj/vasp files are known. for placing molecules inside zeolite framework at a specified location for post-processing the outputs from DFT and MLIP calculations. for relaxation of structures using ase package. Geometries of molecules in Si-CHA zeolite and Cation-exchanged zeolite GRACE model MLIP relaxed geometries MACE model MLIP relaxed geometries DFT relaxed geometries starting geometries These files support the reproducibility of our work on evaluation of the performance of pretrained MLIPs for adsorption of small molecules in zeolites.