Talk - Andrew Peterson

Atomistic reaction engineering: New concepts in strained catalysis and the machine-learning acceleration of calculations

by Andrew Peterson, Brown University, USA


Our ability to describe, calculate, and engineer reactions from the atomic scale has grown dramatically in recent years, and is enabling the first-principles design of materials and molecules related to energy and sustainability.

This talk will have two main topics. In the first, I will describe how we can use a simple mechanics-based model to rationalize strain effects in catalysis. Interestingly, this approach provides insights into directional deviations from the scaling relations that are thought to limit the reactivity of most catalytic pathways. I will show that in certain cases exceptions to these scaling relations arise naturally with strain, but also outline how strained systems can be designed that intentionally deviate from these relations --- that is, engineered deviations to the scaling relations.

In the second part of the talk, I will describe recent advances in machine-learning as it applies to atomistic simulations and introduce our software package known as Amp (the Atomistic Machine-learning Package). I will in particular highlight the potential for the routine use of machine learning to accelerate atomistic calculations, potentially allowing for higher-fidelity electronic structure methods, larger systems, and accelerated exploration of the potential energy surface.


Fri 19 Jan 18
10:00 - 11:00


DTU Fysik


Jakob Schiøtz


DTU Lyngby Campus
Building 311
Lounge, 1st floor