Unraveling subnanocluster stability and fluxionality with MLIPs

R. Laplaza

The number of plausible configurations of metallic subnanoclusters (SNCs) increases exponentially with the number of involved atoms. In catalytic conditions

SNC-based catalysts may also undergo rapid changes due to external potential or magnetic fields

as well as interactions with the support or the adsorbates. Thus

the relative stability and fluxionality of SNCs in different conditions is of paramount importance for the realistic modelling of heterogeneous catalytic processes and

ultimately

for detailed mechanistic understanding and principled catalyst design.

We introduce MLIP-based approaches that leverage the increasing accuracy and availability of transferable foundation models to automatically deal with these challenging global optimization problems towards the realistic modelling of heterogeneous catalytic processes. We are able to thoroughly explore increasingly complex potential energy surfaces at unprecedented scales

and find unexpected results in terms of surface coverage

active site availability

and overall configurational complexity.