Generative Hierarchical Materials Search
Sherry Yang, Simon Batzner, Ruiqi Gao, Muratahan Aykol, Alexander L. Gaunt, Brendan McMorrow, Danilo J. Rezende, Dale Schuurmans, Igor Mordatch, Ekin D. Cubuk
2024-09-12
Summary
This paper talks about Generative Hierarchical Materials Search (GenMS), a new method for generating crystal structures using advanced models that combine language processing and structure generation.
What's the problem?
Creating new crystal structures for materials science can be challenging because it requires a deep understanding of chemistry and physics. Existing methods often lack the ability to effectively translate high-level instructions into specific crystal structures, making it hard to automate the process of discovering new materials.
What's the solution?
To solve this problem, the authors developed GenMS, which uses a combination of a language model and a diffusion model. The language model takes natural language instructions and generates intermediate information about the desired crystal, while the diffusion model creates detailed crystal structures based on that information. Additionally, a graph neural network predicts important properties of these structures. This approach allows for a more controlled generation of crystal structures that meet specific criteria.
Why it matters?
This research is important because it could significantly speed up the discovery of new materials by automating the process of generating crystal structures from simple instructions. By making it easier to create and analyze new materials, GenMS can help advance fields like electronics, pharmaceuticals, and renewable energy.
Abstract
Generative models trained at scale can now produce text, video, and more recently, scientific data such as crystal structures. In applications of generative approaches to materials science, and in particular to crystal structures, the guidance from the domain expert in the form of high-level instructions can be essential for an automated system to output candidate crystals that are viable for downstream research. In this work, we formulate end-to-end language-to-structure generation as a multi-objective optimization problem, and propose Generative Hierarchical Materials Search (GenMS) for controllable generation of crystal structures. GenMS consists of (1) a language model that takes high-level natural language as input and generates intermediate textual information about a crystal (e.g., chemical formulae), and (2) a diffusion model that takes intermediate information as input and generates low-level continuous value crystal structures. GenMS additionally uses a graph neural network to predict properties (e.g., formation energy) from the generated crystal structures. During inference, GenMS leverages all three components to conduct a forward tree search over the space of possible structures. Experiments show that GenMS outperforms other alternatives of directly using language models to generate structures both in satisfying user request and in generating low-energy structures. We confirm that GenMS is able to generate common crystal structures such as double perovskites, or spinels, solely from natural language input, and hence can form the foundation for more complex structure generation in near future.