Flow Matching for Accelerated Simulation of Atomic Transport in Materials

2024-10-02

arXiv.org

AI Summary

Molecular dynamics (MD) simulations of lithium-ion diffusion in crystalline materials—particularly solid-state electrolytes—are computationally prohibitive due to high first-principles costs. Method: This work introduces LiFlow, the first framework to apply flow matching to atomic transport modeling. It employs a propagator–corrector dual-module architecture to jointly enforce geometric and physical constraints, and incorporates a temperature- and composition-adaptive Maxwell–Boltzmann prior to enable cross-system generalization. Contribution/Results: LiFlow achieves Spearman correlation coefficients of 0.7–0.8 for predicting lithium diffusion mean-square displacement (MSD) across 4,186 materials. It generalizes to larger supercells and longer time scales, accelerating diffusion property prediction by up to 6×10⁵× relative to ab initio MD—thereby dramatically enhancing simulation efficiency and scalability for solid-state ionics research.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Matrix & Tensor Methods
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

We introduce LiFlow, a generative framework to accelerate molecular dynamics (MD) simulations for crystalline materials that formulates the task as conditional generation of atomic displacements. The model uses flow matching, with a Propagator submodel to generate atomic displacements and a Corrector to locally correct unphysical geometries, and incorporates an adaptive prior based on the Maxwell-Boltzmann distribution to account for chemical and thermal conditions. We benchmark LiFlow on a dataset comprising 25-ps trajectories of lithium diffusion across 4,186 solid-state electrolyte (SSE) candidates at four temperatures. The model obtains a consistent Spearman rank correlation of 0.7-0.8 for lithium mean squared displacement (MSD) predictions on unseen compositions. Furthermore, LiFlow generalizes from short training trajectories to larger supercells and longer simulations while maintaining high accuracy. With speed-ups of up to 600,000$ imes$ compared to first-principles methods, LiFlow enables scalable simulations at significantly larger length and time scales.

Problem

• Accelerates atomic transport simulations
• Predicts lithium diffusion in materials
• Enables large-scale molecular dynamics

Innovation

• Flow matching for atomic transport
• Adaptive prior based on Maxwell-Boltzmann
• Generative framework for MD simulations