Multimodal Learning for Materials

2023-11-30

AI Summary

Existing machine learning approaches in materials science predominantly focus on unimodal property modeling, failing to capture implicit cross-modal correlations among structural, electronic, and thermodynamic properties. To address this, we propose MultiMat—the first multimodal self-supervised learning framework for materials—built upon a hybrid Transformer–graph neural network architecture that jointly encodes crystal structures and diverse physical properties. MultiMat incorporates contrastive learning and a multi-task decoupled head to enable cross-property knowledge transfer and discover interpretable latent features. After large-scale joint pretraining on Materials Project, the model achieves state-of-the-art performance across multiple property prediction benchmarks. Furthermore, it supports zero-shot generative design of thermodynamically stable novel materials and uncovers emergent physical features consistent with valence-bonding principles.

Abstract

Artificial intelligence is transforming computational materials science, improving the prediction of material properties, and accelerating the discovery of novel materials. Recently, publicly available material data repositories have grown rapidly. This growth encompasses not only more materials, but also a greater variety and quantity of their associated properties. Existing machine learning efforts in materials science focus primarily on single-modality tasks, i.e., relationships between materials and a single physical property, thus not taking advantage of the rich and multimodal set of material properties. Here, we introduce Multimodal Learning for Materials (MultiMat), which enables self-supervised multi-modality training of foundation models for materials. We demonstrate our framework's potential using data from the Materials Project database on multiple axes: (i) MultiMat achieves state-of-the-art performance for challenging material property prediction tasks; (ii) MultiMat enables novel and accurate material discovery via latent space similarity, enabling screening for stable materials with desired properties; and (iii) MultiMat encodes interpretable emergent features that may provide novel scientific insights.

Problem

• Predict material properties using multimodal data
• Accelerate discovery of novel materials
• Enable interpretable features for scientific insights

Innovation

• Multimodal Learning for Materials (MultiMat)
• Self-supervised multi-modality training
• State-of-the-art material property prediction