SwarmThinkers: Learning Physically Consistent Atomic KMC Transitions at Scale

2025-05-26

AI Summary

This study addresses the longstanding challenge of simultaneously ensuring physical consistency, interpretability, and cross-scale scalability in atomistic dynamical simulations. We propose the first reinforcement learning framework that models diffusion as a thermodynamically constrained, decentralized decision-making process among atomic agents. Methodologically: (1) we adopt a centralized-training-with-decentralized-execution (CTDE) architecture with a shared policy network, treating atoms as agents governed by free-energy constraints; (2) we introduce a transition-state reweighting mechanism that jointly incorporates learning preferences and physically grounded hopping rates; (3) the framework achieves zero-shot generalization across system size, solute concentration, and temperature. Evaluated on irradiation-induced precipitation in Fe–Cu alloys, our method attains supercomputer-level accuracy on a single A100 GPU—achieving an average 3,185× speedup (up to 4,963×) and 485× memory reduction versus conventional molecular dynamics.

Technology Category

• Machine Learning: Reinforcement Learning
• Search and Optimization: Sampling/Simulation-based Search
• Multiagent Systems: Multiagent Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Can a scientific simulation system be physically consistent, interpretable by design, and scalable across regimes--all at once? Despite decades of progress, this trifecta remains elusive. Classical methods like Kinetic Monte Carlo ensure thermodynamic accuracy but scale poorly; learning-based methods offer efficiency but often sacrifice physical consistency and interpretability. We present SwarmThinkers, a reinforcement learning framework that recasts atomic-scale simulation as a physically grounded swarm intelligence system. Each diffusing particle is modeled as a local decision-making agent that selects transitions via a shared policy network trained under thermodynamic constraints. A reweighting mechanism fuses learned preferences with transition rates, preserving statistical fidelity while enabling interpretable, step-wise decision making. Training follows a centralized-training, decentralized-execution paradigm, allowing the policy to generalize across system sizes, concentrations, and temperatures without retraining. On a benchmark simulating radiation-induced Fe-Cu alloy precipitation, SwarmThinkers is the first system to achieve full-scale, physically consistent simulation on a single A100 GPU, previously attainable only via OpenKMC on a supercomputer. It delivers up to 4963x (3185x on average) faster computation with 485x lower memory usage. By treating particles as decision-makers, not passive samplers, SwarmThinkers marks a paradigm shift in scientific simulation--one that unifies physical consistency, interpretability, and scalability through agent-driven intelligence.

Problem

• Achieving physically consistent and scalable atomic simulations
• Balancing thermodynamic accuracy with computational efficiency
• Enabling interpretable particle-level decision-making in KMC methods

Innovation

• Reinforcement learning for atomic-scale simulation
• Shared policy network with thermodynamic constraints
• Centralized-training decentralized-execution paradigm