C. Panosetti

Learning to Converge: Warm-Starting DFTB Self-Consistent Charges with Machine Learning

A machine learning approach is presented that accelerates DFTB simulations by predicting optimal initial atomic charges and demonstrates that ML-predicted initial charges consistently and significantly improve SCC convergence across diverse chemical systems including organic molecules, biomolecules, water clusters, tra...