Call for Abstracts - Short Symposium 18^th ICC

Within the framework of the 18^th International Congress on Catalysis (ICC) to be held in Lyon from July 14^th to July 19^th 2024, a short symposium will be organized on “Quantum simulation combined with machine learning for catalysis.” The call for abstracts is open until October 31, 2023. We invite any interested researcher of the community to submit abstracts following the instructions provided at the official ICC website: https://www.icc-lyon2024.fr

Abstract: The advent of artificial intelligence and machine learning (ML) methods is expected to usher a new era in the field of computational chemistry applied to catalysis in the coming years.  Thanks to these techniques and the massive expansion of computing systems, it may become possible to simulate more realistic catalysts at operating conditions at much lower computational cost while still maintaining quantum mechanical accuracy. This growing trend has been highlighted in various perspective articles published recently in catalysis journals showing the potential of combining machine learning and quantum simulation methods based on density functional theory (DFT). During this symposium, it is proposed to gather a broad international community working in computational materials science, quantum chemistry, catalysis (homogeneous, heterogeneous, enzymatic) and chemical engineering interested in fundamental and applied concepts. We aim at highlighting and discussing key challenges among the following items based on high quality oral contributions.

• The development and use of ML algorithms to accelerate the discovery of reaction mechanisms and the determination of accurate rate constants while allowing to go beyond conventional transition states search methods. ML approaches may promote the use of enhanced sampling molecular dynamics (MD) methods and also the more systematic identification of collective variables for complicated reaction networks.
• The development and use of ML potentials where a numeric potential is derived based on underlying quantum mechanical data through neural network approaches, to massively extend accessible length and time scales in catalysis. It will be discussed how this approach may help for keeping an accuracy level close to DFT while allowing more systematically MD simulation of catalytic events.
• The combination of ML and DFT to predict key catalytic materials properties and to accelerate the establishment of quantitative-structure activity relationships (such as generalized volcano curves). ML based screening approach may allow to expand the exploration datasets of both descriptors and catalyst formulations, although there are still open questions regarding sizes of the training datasets to build reliable and accurate predictive models in catalysis.

Symposium organizers: P. Raybaud (IFPEN, France), W. Schneider (Univ. Notre-Dame, USA), V. van Speybroeck (Ghent University, Belgium)