The impact of first principles calculations on discoveries made for industrially relevant systems is growing year after year. The present talk will illustrate the various stages of the catalyst understanding and performance prediction where computational catalysis has a crucial role to play, with a focus on the simulation of complex proton-exchanged zeolite catalysts of industrial relevance.[1,2] Most calculations discussed were performed in the framework of the periodic density functional theory (DFT). Comparison with experimental spectroscopic or catalytic feature is key to get advance knowledge and to validate the prediction procedures, making use of micro-kinetic models based on DFT calculations.

In the case of the bifunctional hydroconversion of n-heptane in large pores zeolites, performing ab initio molecular dynamics (AIMD) appeared to be necessary to capture the right reaction intermediates. This ensemble of reactions, involving alkene isomerization and cracking in zeolites, is highly relevant in the field of refining, biomass conversion and plastic recycling. Accounting for the dynamic behavior along key-reaction steps, the molecular origins of the respective rates of type B versus type A isomerization steps (affecting and not affecting the branching degree, respectively) was elucidated,[3] as well as the nature of the kinetically relevant intermediates for isomerization steps involving secondary carbenium ions.[4] The dynamic nature of the type B1 cracking transition state (from secondary to tertiary carbenium ions) makes it more stable with respect to the one of type B2 cracking (from tertiary to secondary carbenium ion, revealed to deviate from the one of a beta-scission). [5] The corresponding rate constants were used to integrate a single event microkinetic models, with a satisfactory prediction ability, comparing with experimental kinetic features measured thanks to a high-throughput experiment setup.[6] However, a deviation close to 15 kJ/mol for the difference between isomerization and relevant cracking steps was observed, the compensation of which appeared to be necessary for good selectivity predictions. We recently showed that this difference is entirely explained by the level of theory. A much higher level (Random Phase Approximation, RPA) than DFT appeared to be necessary, in combination with AIMD. Due to the computational cost of this method, combining RPA with AIMD would require about 103 years if calculations, but thanks to a newly developed Machine Learning Perturbation Theory, these much accurate evaluations could be done within a few months.[7]

We recently undertook to extend these methodologies to the simulation of the transformation of bio-sourced molecules. Our first works devoted to the understanding of the mechanisms of isobutanol dehydration into linear alkenes, catalyzed by proton-exchanged zeolites, will be presented.[8]

However, considering the reactivity of the bulk sites only may not be representative enough of the real catalyst, for which efforts need to be dedicated in terms of simulation of defects, of external surfaces, and of their dealumination. Recent achievements in this direction will also be illustrated.[9]

[1] C. Chizallet, ACS Catal., 2020, 10, 5579

[2] C. Chizallet, C. Bouchy, K. Larmier, G. Pirngruber, Chem. Rev. 2023, 123, 6107

[3] J. Rey, A. Gomez, P. Raybaud, C. Chizallet, T. Bučko, J. Catal., 2019, 373, 361-373.

[4] J. Rey, P. Raybaud, C. Chizallet, T. Bučko, ACS Catal., 2019, 9, 9813-9828.

[5] J. Rey, C; Bignaud, P. Raybaud, T. Bučko, C. Chizallet, Ang. Chem. Int. Ed., 2020, 59, 18938

[6] J.-M. Schweitzer et al., ACS Catal., 2022, 12, 1068-1081., ACS Catal., 2022, 12, 1068

[7] J. Rey et al., Ang. Chem. Int. Ed., 2024, 63, e202312392, Catal. Sci. Technol., in press, https://doi.org/10.1039/D4CY00548A

[8] M. Gesvandtnerova et al., J. Catal., 2022, 413, 786, ACS Catal., 2024, 14, 7478

[9] J. Rey et al., ChemCatChem, 2017, 9, 2176 ; L. Treps et al., ACS Catal., 2020, 10, 3297 ; J. Phys. Chem. C, 2021, 125, 2163 ; T. Jarrin et al., ChemCatChem, 2023, 15, e202201302, T. Jarrin et al., ACS Catalysis, 2024, 14, 1639.