Paper Image

Machine learning improves molecular force fields

Paper Authors:

Moritz Thürlemann,

Sereina Riniker

Bullets

Key Details

Combines classical force fields with machine learning corrections

Trained on quantum chemistry data for small systems

Achieves accuracy close to density functional theory

Transfers from small molecules to condensed phases

Shows promise for modeling liquids, crystals, structure prediction

AI generated summary

Machine learning improves molecular force fields

This paper proposes a hybrid machine learning and physics-based model for predicting molecular properties and interactions. It combines classical force fields with machine learning corrections, achieving improved accuracy and transferability from small molecules to condensed phases like liquids and crystals. The model is trained on high-quality quantum chemistry data for small systems, then validated on benchmark datasets. Results show performance on par with density functional theory for diverse molecular datasets, at lower computational cost. The authors demonstrate promising capabilities for modeling challenging condensed phase systems and crystal structure prediction.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up