Making Molecules Stable: How AI Helps in Relaxations

Making Molecules Stable: How AI Helps in Relaxations

Chemistry

Ashan Balasuntharam

2025

Molecules relax to their most stable, lowest-energy structures through geometry relaxation. While DFT provides accurate results, it is computationally expensive. This work compares machine learning interatomic potentials (MLIPs) to DFT, evaluating how well MLIPs reproduce DFT-level stability at a fraction of the cost.

Computational Chemistry, DFT, Density Functional Theory, MLIP, VASP, Atomic Simulations Environment, CHGNet, MACE, SevenNet