Fine-Tuned Language Models Generate Stable Inorganic Materials as Text

TL;DR

This paper demonstrates that fine-tuning large language models on atomistic data can reliably generate stable inorganic materials, outperforming existing models in predicting metastable structures and capturing crystal symmetries.

Contribution

It introduces a novel approach of fine-tuning large language models for material generation, showing high physical constraint adherence and improved metastability prediction.

Findings

90% of generated structures obey physical constraints

49% of generated materials are metastable, nearly double previous models

Language models capture crystal symmetries better with increased scale

Abstract

We propose fine-tuning large language models for generation of stable materials. While unorthodox, fine-tuning large language models on text-encoded atomistic data is simple to implement yet reliable, with around 90% of sampled structures obeying physical constraints on atom positions and charges. Using energy above hull calculations from both learned ML potentials and gold-standard DFT calculations, we show that our strongest model (fine-tuned LLaMA-2 70B) can generate materials predicted to be metastable at about twice the rate (49% vs 28%) of CDVAE, a competing diffusion model. Because of text prompting's inherent flexibility, our models can simultaneously be used for unconditional generation of stable material, infilling of partial structures and text-conditional generation. Finally, we show that language models' ability to capture key symmetries of crystal structures improves with model scale, suggesting that the biases of pretrained LLMs are surprisingly well-suited for atomistic data.