Speaker: Karsten Reuter, Fritz-Haber-Institur der Max-Planck-Gesellschaft

Machine learning (ML) promises a significant enhancement of multi-scale modeling capabilities in the context of energy conversion and storage (ECS). In particular, ML interatomic potentials (MLIPs) trained with first-principles data already offer orders of magnitude speed-ups in the computation of predictive-quality energies and forces in atomic-scale simulations. This new efficiency finally allows to heads-on tackle the highly dynamic evolution of working interfaces in ECS systems, where the targeted functionality like catalytic activity or ion mobility both inherently drives and results from ongoing substantial structural, compositional and morphological changes. Unable to fully capture such operando evolution, direct first-principles based multiscale modeling focused hitherto on model (single-)crystalline surfaces or interfaces, where the system dynamics was typically restricted to select reacting or diffusing species that were considered central for a targeted primary function. The MLIP-enabled enhanced sampling capabilities instead allow to assess the thermodynamic stability of complex, possibly amorphous configurations and thereby establish reliable structural models for the working interfaces. Automated process exploration in turn provides more systematic access to the elementary steps that drive the operando evolution, paving the way for microkinetic simulations that analyze the entanglement of this evolution with the primary function.