Thomas Purcell*

Degrees and Appointments 

• Bachelor of Science: Chemistry, NYU, 2013
• Ph.D.: Physical Chemistry, Northwestern University, 2018
• Alexander von Humboldt Fellow: NOMAD Lab, Fritz-Haber-Institut der MPG, 2018-2022
• Group Leader: NOMAD Lab, Fritz-Haber-Institut der MPG, 2022-2023

Specialties: Materials Discovery, Artificial Intelligence, Electronic Structure Calculations

RESEARCH GROUP WEB PAGE

Research Description

Our main objective is to create AI-guided workflows for materials discovery. We are especially interested in the low-data limit, where the incomplete sampling of chemical space obscures our physical understanding of the target properties. With newly implemented active learning frameworks, that combine traditional high-throughput computational chemistry with machine learning, we efficiently search for new, optimal materials for renewable energy applications. We then probe the resulting models to gain insights into the physical mechanisms that control the properties of interest and create heuristics for them.

To achieve these goals, my group focuses on both developing new AI algorithms and designing/executing computational workflows. The new algorithms will enable us to create more descriptive models by better representing the molecules and materials comprising the datasets, while the high-throughput calculations provide us with consistent data. More importantly, incorporating the learning steps within the workflows themselves allows us to sample chemical space more efficiently and focus our calculations on the most promising candidates. Exposing students to both sides of the group creates an environment where new ideas can be quickly prototyped and exemplified on both toy and real-world systems. Students also gain invaluable experience with data science, high-performance computing, and applying these techniques to solve pressing chemical problems.