Evaluation of the catalytic reaction constants of alcohols dehydration by a combined machine learning-quantum chemistry approach.

The advent of Machine Learning techniques (ML) is stimulating modern computational chemistry. This methodology may open avenues to address challenging questions in catalytic reactions involved in biomass derived molecules transformation. In particular, free energies of activation and rate constants determination may become accessible by Ab Initio Molecular Dynamics (AIMD) combined with ML.
We intend to apply AIMD-ML to a relevant case study in heterogeneous catalysis: the dehydration of bio-based alcohols catalyzed by γ-alumina surfaces. We expect that ML will help to overcome two limitations of AIMD (biased or unbiased): the identification of proper collective variables (CV), summarizing key information of a reaction step, and the construction of faster empirical Hamiltonians (including the use of neural networks), thus allowing more statistically sound AIMD trajectories. Finally, we will use AIMD-ML in conjunction to Adaptative Multilevel Splitting (AMS) algorithms, for the direct determination of transition probabilities.
We propose to benchmark AIMD-ML with previous calculations performed with conventional approaches based on Transition State Theory (TST) and harmonic approximation. Ultimately, we target to discriminate among various mechanisms and various key intermediates involved in the dehydration of alcohols to olefins as well as to identify chemical trends as a function of the alcohol molecule and alumina active sites. 
This project offers the candidate a unique opportunity to learn the advanced simulation techniques both in ML and computational chemistry, and the opportunity to join a growing community of scientists working in these fields. He will work within the framework of a multi-partner project where chemists (IFPEN) and mathematicians (IFPEN and Ecole des Ponts – INRIA) will collaborate.   

Keywords: Enhanced sampling, Machine Learning, Ab Initio, Molecular Dynamics, Catalysis

• IFPEN supervisors    Dr. Pascal Raybaud : https://www.ifpenergiesnouvelles.com/page/pascal-raybaud and Dr. Manuel Corral Valero : https://orcid.org/0000-0002-4457-3914 
• Academic supervisor    Pr. Tony Lelièvre, Ecole des Ponts ParisTech, tony.lelievre@enpc.fr 
• Doctoral School    ED 206 Chimie de Lyon
• Employer    IFPEN
• PhD location    IFPEN (Lyon site) and Ecole des Ponts (Champs sur Marne), France  
• Duration and start date    3 years, starting on October or November 2023
• Academic requirements    Master degree in Theoretical Chemistry or Applied Mathematics
• Language requirements    Fluency in French or English, willing to learn French
• Other requirements    Desired knowledge of python computer language