Talk at ACHEMA: Universal neural network potential for chemicals and materials discovery

Rudy Coquet, Preferred Computational Chemistry, Inc. (PFCC), will give a talk at ACHEMA, the world’s leading trade show for the process industries in Germany.
ACHEMA official website: https://www.achema.de/en/

---

Event period

June 10 – 14, 2024

Location

Frankfurt am Main, Germany

Booth

Hall 11.0 Booth F77

Presentation date, time and place

June 11 2024, 3:30pm – 4:00pm in Frequenz 2 – Portalhaus

Presentation title

Universal neural network potential for chemicals and materials discovery

Abstract

Atomic-scale simulations are making it possible to do more with complex chemistry than ever before. The rapid advancements of AI technology have brought about revolutionary changes in chemicals and materials discovery.

Such technology will be a vital component of a sustainable future, boosting efficiency for chemists at an individual level and allowing companies to handle increasingly large amounts of data while harnessing the numerous benefits of AI.

This presentation will first introduce a robust universal machine learning interatomic potential that covers much of the periodic table. More than one thousand GPU years were used to generate the ab initio training data guided by active learning. Diverse test simulations have shown this machine learning potential has outstanding performance, with energy error significantly less than the chemical accuracy (43 meV/atom) even for chemically very complex systems. This universal potential, which powers the Matlantis™ atomistic simulator, can run over 10,000 times faster than density functional theory (DFT) when dealing with several thousand atoms, and the latest release allows for more than 10,000 atoms of arbitrary combinations of 72 elements to be simulated together. 

Focusing on use cases ranging from battery materials to lubricants and catalysts, the rest of the presentation will reveal how machine learning-based atomic scale simulations performed with Matlantis™ allow the study of realistic chemical systems, help rationalize experimental results, and foster the discovery of new chemicals. For example, it will be revealed how highly accurate and versatile machine learning models that link molecular structures to Radial Distribution Function (RDF) allowed the screening of tens of thousands of base oil compounds, resulting in the discovery of three compounds that could improve the processability of plastic working lubricants.

This presentation will be especially suitable for computational scientists who face the limitations of current simulation approaches and anyone interested in game-changing atomic scale simulations. No prior knowledge of the subject is required to take this first step towards large-scale chemicals and materials discovery, supported by state-of-the-art machine learning potentials.

Speaker