Physchem.cz

Use of ML-accelerated (and quantum chemistry-supported) simulations for bottom-up design of metal/metal-oxide cluster-based supported nanocatalysts for sustainable chemistry applications.The grand challenge in the field of supported nanocatalysts is to design materials with the maximum catalytic activity/selectivity, using the least catalytic material, while maintaining stability against deactivation. Single atom and sub-nanometre metal particles on oxide supports are a promising class of system, but there are significant gaps in knowledge at the atomic scale. Costly electronic structure calculations do not allow for operando modelling of these systems, their evolution, or the full reaction landscape.We wish to know:· How do the clusters interact with the surface?· How can we control their motion and growth?· What effect do particle size and shape have on the particular catalytic applications?· How can we utilise the effects of alloying to generate better catalysts?· What is the best system for a particular green chemistry application?Systems of interest include late transition metal and coinage metal clusters (Determination of the influence of the cluster composition, size and oxidation state, in addition to the interaction with (realistic models of) support, on nanocatalyst stability and activity will be a primary goal of the project, which will leverage ML to go far beyond the capacities of DFT and post-HF methods.The project will employ a variety of computational techniques, including ML-driven (surrogate model) unbiased global structure optimization, in combination with ML-enhanced atomistic dynamical simulations, for development of model systems. Discrete path sampling techniques for characterisation of the underlying energy landscape. Grand canonical Monte Carlo simulations will be deployed for the generation of metal/metal-oxo phase diagrams, to explain the complex oxidation behaviour at the sub-nanoscale. ML-assisted reaction path sampling techniques (OPES/variational autoencoders/umbrella sampling) will be employed to determine catalytic reaction mechanisms and transfer learning schemes will be used for the smooth transition between metals and the improvement of sampling efficiency across related reactive processes.Computational structural characterisation (alongside catalyst evolution) will be conducted alongside experimental verification by international collaborators, (e.g. LEED IV, STM, TEM, GISAXS, XPS and RAIRS), while reactivity calculations will support measurements of catalytic activity (e.g. TOF-MS).

Relevant Publications from the group:

1.  Nanoscale, 2024,10.1039/D4NR00017J
2. Angewandte Chemie, 2022, https://doi.org/10.1002/anie.202213361
3. Catal. Sci. Technol., 2022,10.1039/D1CY02270A
4. Angewandte Chemie, 2020,https://doi.org/10.1002/anie.202015138
5. ACS Catalysis, 2020, 10.1021/acscatal.0c01344
6. J. Phys. Chem. C, 2017, 10.1021/acs.jpcc.6b12072
7. ACS Catalysis, 2016, 10.1021/acscatal.5b02708

Ongoing collaborations relevant to the project

• D.J. Wales (Cambridge)
• J.C. Schoen (MPI-Stuttgart)
• G. Parkinson (TU Wien)
• A. Fortunelli (Pisa)
• S. Vajda/F. Loi (Heyrovsky)
• M. Mazur / J. Cejka (CU)