ReciNet: Reciprocal Space-Aware Long-Range Modeling for Crystalline Property Prediction

TL;DR

ReciNet is a novel reciprocal space-aware neural network that effectively captures long-range interactions in crystal structures, achieving state-of-the-art accuracy in predicting various crystal properties.

Contribution

The paper introduces ReciNet, a new architecture leveraging reciprocal space Fourier representations to model long-range interactions in crystals, surpassing existing methods.

Findings

ReciNet achieves state-of-the-art accuracy on multiple crystal property benchmarks.

The reciprocal space approach effectively captures long-range interactions.

Multi-property prediction with mixture-of-experts is computationally efficient and beneficial.

Abstract

Predicting properties of crystals from their structures is a fundamental yet challenging task in materials science. Unlike molecules, crystal structures exhibit infinite periodic arrangements of atoms, requiring methods capable of capturing both local and global information effectively. However, current works fall short of capturing long-range interactions within periodic structures. To address this limitation, we leverage \emph{reciprocal space}, the natural domain for periodic crystals, and construct a Fourier series representation from fractional coordinates and reciprocal lattice vectors with learnable filters. Building on this principle, we introduce the reciprocal space-based geometry network (\textbf{ReciNet}), a novel architecture that integrates geometric GNNs and reciprocal blocks to model short-range and long-range interactions, respectively. Experimental results on standard benchmarks JARVIS, Materials Project, and MatBench demonstrate that ReciNet achieves state-of-the-art predictive accuracy across a range of crystal property prediction tasks. Additionally, we explore a model extension to multi-property prediction with the mixture-of-experts, which demonstrates high computational efficiency and reveals positive transfer between correlated properties. These findings highlight the potential of our model as a scalable and accurate solution for crystal property prediction.