Julien Steffen
(Universität Erlangen-Nürnberg)

Abstract:
In recent years, novel approaches for chemical catalysts have been developed, two of these are catalytically active liquid metal solutions (SCALMS) ^[1] and confined spaces like metal organic frameworks or graphene-coated metal surfaces ^[2]. These are flexible and adaptive systems and need to be simulated by means of molecular dynamics samplings. To enable those samplings with high accuracy, needed for rate constant calculations, machine-learned potentials (MLP) are well suited. In this talk, the setup and application of MLPs to SCALMS surfaces ^{[3, 4]} and hydrogen storage in confined spaces, a conceptual precursor to confined-space catalysis, are outlined. Different strategies for efficient MLP parametrization are explained, their performance and stability are critically evaluated, and the treatment of nuclear quantum effects with ring-polymer molecular dynamics, important for processes involving hydrogen at low temperatures ^{[5, 6]}, is elucidated.
 
References:
^[1] N. Taccardi et al., Nat. Chem. 9, 862-867 (2017).
^[2] M. Prieto et al., J. Am. Chem. Soc. 143, 8780-8790 (2021). [3] M. Moritz et al., ACS Catal. 14, 6440-6450 (2024).
^[4] A. Søgaard et al., ChemPhysChem e202400651 (2025).
^[5] J. Steffen, J. Chem. Theory Comput. 19, 5334-5355 (2023).
^[6] J. Steffen, A. Alibakhshi, J. Chem. Phys. 161, 184116 (2024).