EquiFiLM enhances AI force fields for electronic state changes

Researchers have developed EquiFiLM, a novel extension for foundation machine learning force fields (MLFFs) that enables them to handle externally induced changes to electronic states. This method uses a lightweight, per-layer Feature-wise Linear Modulation (FiLM) block to continuously condition MLFFs, allowing them to learn from minimal training data about effects like charging or applied fields. When applied to charged liquid water using the MACE-MatPES backbone, the resulting E-MACE model showed significant reductions in force and energy root-mean-square error compared to a baseline without EquiFiLM, while maintaining indistinguishable inference costs. AI

IMPACT Enhances the capability of AI force fields to model complex chemical processes involving electronic state changes, potentially accelerating materials science and drug discovery.

RANK_REASON The cluster contains a research paper detailing a new method for enhancing machine learning force fields. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• DFT
• EquiFiLM
• MACE-MatPES
• MACE-MP-0
• Samuel Sahel-Schackis

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →