Supplementary data for "Generalized Machine Learning Potentials for Water Adsorption in Al-Based Metal–Organic Frameworks Considering Local Flexibility"

This Zenodo includes four parts, i.e., four compressed folders below. DFT-Dataset: Training dataset for MACE-AlMOFs-68 model (~43000 configurations for 68 MOFs) Training dataset for MACE-AlMOFs-v1 model (~160000 configurations for 416 MOFs). Test datasets including 420 MOFs (~100000 configurations) Datasets to evaluate energy offsets for 352 MOFs and H2O. (~40000 configurations) Scripts: CP2K calculation script MACE training script models: MACE-AlMOFs-68 model MACE-AlMOFs-v1 model Models fine-tuned on the same training dataset to MACE-AlMOFs-68 model, but with a different ratio between energy weight and force weight in the final stage (10:1, 1:1, 1:10, 1:100, 1:1000). We investigate the influence of energy weight on energy offsets. Results: AlMOFs dataset, including 420 unique AlMOFs with 439 ligands Optimized structure from MACE-AlMOFs-v1 model The accessible volume and Qst of Al-MOFs for screening studies. (v1.3 update)