Parameter-Efficient Fine-Tuning of Machine-Learning Interatomic Potentials for Phonon and Thermal Properties

TL;DR

This paper explores efficient fine-tuning methods for machine-learning interatomic potentials, significantly improving phonon and thermal property predictions with minimal data, and introduces the Equitrain framework based on LoRA adaptation.

Contribution

The paper introduces Equitrain, a LoRA-based fine-tuning framework, and demonstrates its superior performance across 53 materials systems for phonon and thermal property predictions.

Findings

Fine-tuning with as few as 10 structures yields significant accuracy gains.

Equitrain outperforms both pretrained and scratch-trained models.

Fine-tuning enables accurate phonon and thermal property predictions.

Abstract

Machine-learning interatomic potentials are widely used as computationally efficient surrogates for density functional theory in atomistic simulations, enabling large-scale, long-time modeling of materials systems. We investigate how different fine-tuning strategies influence the prediction of harmonic phonon band structures, thermal properties, and the potential energy surface along imaginary phonon modes. We achieve substantial accuracy improvements with minimal additional data, with as few as 10 additional training structures already yielding significant gains. In addition to existing approaches, we introduce Equitrain, a finetuning framework that implements LoRA-based adaptation. Across 53 materials systems, we show that fine-tuned models consistently outperform both the underlying pretrained model and models trained from scratch. Equitrain achieves the best overall performance, and our results demonstrate that fine-tuning enables accurate phonon predictions.