Materials Design at Scale with Universal Machine Learning Models and Big Data

Chi Chen1; Yunxing Zuo1; Jian Luo1; Shyue Ping Ong1*
1 Department of NanoEngineering, University of California, San Diego, 9500 Gilman Dr. Mail Code #0448, La Jolla, CA 92093-0448, USA

EXTENDED ABSTRACT: In silico materials design often involves the exploration of vast, diverse chemical spaces. While ab initio methods have been phenomenally successful in materials simulations, their scope of application has always been constrained by their high cost and poor scaling. In this talk, I will highlight the development of graph deep learning models that can revolutionize our ability to explore the entire universe of materials at unprecedented scales and accuracy. I will first showcase the capabilities of such models in the accurate prediction of material properties such as formation energies, band gaps, and elastic constants, and how these capabilities have been applied in a real-world design-predict-synthesize workflow of ultra-incompressible materials. More recently, we have also developed graph deep learning interatomic potentials (IAPs) that can universally work across the periodic table. Our Materials 3-Body Graph Network (M3GNet) models can be considered as “foundational AI models” for materials chemistry, with broad applications in the dynamic simulations and discovery of materials. Finally, I will also provide some perspectives on the remaining challenges and opportunities for materials data generation and ML model development. 

                                                                                          Figure 1. M3GNet architecture

Keywords:machine learning; materials discovery; universal potentials; batteries References

[1] Chen, C.; Ye, W.; Zuo, Y.; Zheng, C.; Ong, S. P. Chem. Mater. 2019, 31 (9), 3564–3572.
[2] Zuo, Y.; Qin, M.; Chen, C.; Ye, W.; Li, X.; Luo, J.; Ong, S. P. Materials Today, 2021, 51, 126–135.
[3] Chen, C.; Ong, S. P. Nat Comput Sci, 2022, 2 (11), 718–728.

Brief Introduction of Speaker
Shyue Ping Ong

Shyue Ping Ong is a Professor of NanoEngineering at the University of California, San Diego. He obtained his PhD from the Massachusetts Institute of Technology in 2011. He leads the Materials Virtual Lab at UCSD, a dynamic group of materials scientists focusing on the interdisciplinary application of materials science, computer science, and data science to accelerate materials design. He is one of the founding developers of the Materials Project, a DOE-funded initiative to make the computed properties of all known materials publicly available for materials innovation. He is also the founder of Python Materials Genomics (pymatgen), an open-source materials analysis library that is used by hundreds of thousands of users worldwide. Dr Ong is a recipient of the US Department of Energy Early Career Research Program and the Office of Naval Research Young Investigator Program awards and a Clarivate Highly Cited Researcher in 2021 and 2022.