Accurate, transferable, and verifiable machine-learned interatomic potentials for layered materials

2025-03-19

AI Summary

Accurate atomic-scale modeling of large moiré domains in twisted van der Waals heterostructures remains challenging, hindering reliable prediction of electronic and optical properties. Method: We propose a novel “intra-layer–inter-layer decoupled” machine learning interatomic potential (MLIP) framework. A physics-driven stacking configuration distribution metric is introduced for rigorous validation; remarkably, we discover that one-dimensional moiré systems serve as efficient, model-agnostic, layer-transferable surrogates for accuracy assessment. The MLIP integrates physics-informed data curation, DFT-based benchmarking, and multi-layer relaxation optimization. Contribution/Results: Prediction accuracy for energy and atomic forces improves tenfold. Applied to the HfS₂/GaS heterostructure, the framework enables end-to-end, quantitatively reliable prediction from atomic structure to electronic properties. It further generalizes to complex multi-layer moiré systems, establishing a rigorous, efficient, and verifiable structural modeling paradigm for twistronics.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Humans and AI: Human-in-the-loop Machine Learning
• Multiagent Systems: Multiagent Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Twisted layered van-der-Waals materials often exhibit unique electronic and optical properties absent in their non-twisted counterparts. Unfortunately, predicting such properties is hindered by the difficulty in determining the atomic structure in materials displaying large moir'e domains. Here, we introduce a split machine-learned interatomic potential and dataset curation approach that separates intralayer and interlayer interactions and significantly improves model accuracy -- with a tenfold increase in energy and force prediction accuracy relative to conventional models. We further demonstrate that traditional MLIP validation metrics -- force and energy errors -- are inadequate for moir'e structures and develop a more holistic, physically-motivated metric based on the distribution of stacking configurations. This metric effectively compares the entirety of large-scale moir'e domains between two structures instead of relying on conventional measures evaluated on smaller commensurate cells. Finally, we establish that one-dimensional instead of two-dimensional moir'e structures can serve as efficient surrogate systems for validating MLIPs, allowing for a practical model validation protocol against explicit DFT calculations. Applying our framework to HfS2/GaS bilayers reveals that accurate structural predictions directly translate into reliable electronic properties. Our model-agnostic approach integrates seamlessly with various intralayer and interlayer interaction models, enabling computationally tractable relaxation of moir'e materials, from bilayer to complex multilayers, with rigorously validated accuracy.

Problem

• Predicting electronic and optical properties in twisted layered materials.
• Improving accuracy of machine-learned interatomic potentials for moiré structures.
• Developing a holistic validation metric for large-scale moiré domains.

Innovation

• Split machine-learned interatomic potential improves accuracy
• Holistic metric for moiré structure validation developed
• One-dimensional moiré structures used for efficient validation