An SO(3)-equivariant reciprocal-space neural potential for long-range interactions

1. 1Developing SO(3)-equivariant neural networks to capture rotational symmetries
2. 2Implementing reciprocal-space formulations to model long-range interactions
3. 3Integrating the proposed method with existing machine-learning interatomic potentials
4. 4Evaluating the performance of the new potential on benchmark molecular and condensed-phase systems

This research benefits materials scientists and computational chemists on a team, as it improves the accuracy of machine-learning interatomic potentials for long-range interactions

💡 The proposed method can accurately represent anisotropic, slowly decaying multipolar correlations in realistic materials

💡 New SO(3)-equivariant reciprocal-space neural potential for long-range interactions in molecular systems!