Implicit Machine Learning Force Fields Accelerate Molecular Dynamics Simulations

2026-07-31

AI Summary

Traditional machine learning force fields struggle to simultaneously achieve high accuracy and computational efficiency, hindering large-scale molecular dynamics simulations. This work proposes an implicit machine learning force field (I-MLFF), which introduces implicit modeling into force field construction for the first time. By replacing explicit neural network stacking with a self-consistent fixed-point equation, I-MLFF reuses intermediate representations across time steps to accelerate force evaluation. Built upon fixed-point iterations of graph neural networks, the method is compatible with scalar, Cartesian tensor, and SO(3) spherical tensor architectures, thereby combining the low computational overhead of shallow models with the expressive power of deep networks—while preserving atomic resolution and the original integration timestep. Experiments demonstrate 2–5× reductions in both computational and memory costs, significantly enhancing the feasibility of long-timescale, large-system simulations.

Abstract

We introduce implicit machine learning force fields (I-MLFFs), which replace explicit stacks of neural network layers with self-consistent fixed-point equations. In molecular simulations, this formulation enables intermediate representations to be reused across successive timesteps, thereby warm-starting force evaluation. The resulting models effectively combine the computational footprint of a shallow, single-layer MLFF with the representational capacity and accuracy of a deep neural network. Our approach unlocks architecture-agnostic efficiency gains that are inaccessible when force prediction and trajectory integration are considered separately. We demonstrate this across three major classes of graph neural networks: invariant, equivariant Cartesian tensor, and SO(3)-equivariant spherical-tensor architectures. Each yields a two- to five-fold reduction in compute and memory footprint. Crucially, these gains are achieved while retaining full atomistic resolution and the original integration timestep, avoiding spatial or temporal coarse graining. Our contribution therefore advances the scaling frontier of quantum-mechanically faithful molecular simulation, enabling longer trajectories and larger atomistic systems within fixed GPU memory and compute budgets, and thereby opening access to new insights across biomolecular and material systems.

Problem

• machine learning force fields
• molecular dynamics
• computational efficiency
• atomistic simulation
• GPU memory

Innovation

• implicit machine learning force fields
• fixed-point equations
• warm-starting
• equivariant graph neural networks
• molecular dynamics acceleration