Learning the Language of Crystal Chemistry: Using Concepts from Natural Language to Model Solid State Chemistry

NOMATEN HYBRID-SEMINAR

online: https://meet.goto.com/NCBJmeetings/nomaten-seminar
In-person: NOMATEN seminar room (102)

Tuesday, FEBRUARY 17^th  2026 13:00 PM (CET)

Learning the Language of Crystal Chemistry: Using Concepts from Natural Language to Model Solid State Chemistry

Keith Butler, PhD
Department of Chemistry, University College London

Abstract:

The discovery and design of new materials is critical for advancing carbon-emission reducing technologies such as renewable energy and electric vehicles. Experimental discovery of new materials is typically slow and costly, quantum mechanics (QM) calculations have brought computational materials design within reach. However, QM calculations are often limited to relatively small sets of materials, as their computational costs are too great for large-scale screening, this is the case for calculating properties required for new energy materials. In this talk I will present examples of how we have been adapting concepts from language models to help with building fast and efficient models for materials properties. I will show how we can learn distributed representations of atomic species directly from large databases of crystallographic structures [1, 2]. I will also show how a large language model trained on crystallographic data can help to solve one of the outstanding challenges of solid-state chemistry; the prediction of structure from chemical formula [3].
[1] npj Comput Mater 8, 44 (2022)
[2] APL Machine Learning 2,  (2024)
[3] Nature Commun. 15, 10570 (2024)
 

Bio:

Keith Butler is an Associate Professor in the Department of Chemistry at University College London. He obtained his undergraduate degree in Chemistry from Trinity College Dublin in 2004 and completed a PhD in Computational Chemistry at UCL in 2009 under the supervision of Dewi Lewis, studying the nucleation and growth of zeolites. Following postdoctoral research at the Universities of Sheffield and Bath, working on atomistic simulations of interfaces in photovoltaic materials—including crystalline silicon and hybrid halide perovskites—he developed a strong interest in applying machine learning to materials discovery.
In 2018, Keith moved to the Rutherford Appleton Laboratory, where he helped establish the Scientific Machine Learning (SciML) group. He joined Queen Mary University of London as Senior Lecturer in Green Energy Materials in 2022, before returning to UCL Chemistry in 2023.
Keith’s research combines data-driven methods, such as deep learning and Bayesian statistics, with quantum mechanical simulations to accelerate materials design and experimental characterisation. He was part of the team awarded the Sir George Stokes Award for developments in X-ray diffraction tomography and is a founder member of the Institute of Physics Machine Learning and Neuromorphic Computing Group. He is Deputy Editor of npj Computational Materials and an editorial board member of Machine Learning: Science and Technology.