SurfaceScience21 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 42: Poster Session III: Poster to Mini-Symposium: Machine learning applications in surface science I
O 42.8: Poster
Dienstag, 2. März 2021, 10:30–12:30, P
IrO2 surface complexions identified through machine-learned interatomic potentials — •Jakob Timmermann1,2, Yonghyuk Lee1,2, Carsten Staacke1,2, Christoph Scheurer1,2, and Karsten Reuter1,2 — 1Fritz-Haber-Institut der MPG — 2Technische Universität München
IrO2 is currently the preferred catalyst for the electrochemical oxygen evolution reaction in proton exchange membrane electrolyzers. Full ab initio molecular dynamics (MD) simulations of the reactive processes at the surface would be highly desirable for mechanistic catalyst improvement, but are computationally not tractable for a foreseeable time. To overcome the limitations regarding system size and propagation time, MDs based on machine-learned interatomic potentials are an appealing alternative. Here, we present a Gaussian Approximation Potential (GAP) approach for IrO2 combining two-body and smooth overlap of atomic positions (SOAP) descriptors to capture the atomic environments. For maximum data efficiency, we pursue an iterative parametrization protocol, in which preliminary GAP potentials based on limited first-principles data are used to generate most meaningful additional structures for retraining. The final GAP potential enables a global geometry optimization of low-index rutile IrO2 facets through simulated annealing. Consecutive ab initio thermodynamics and detailed surface science investigations fully confirm the identified novel (101) and (111) (1x1) terminations as competitive with the most studied (110) facet in reducing environments [1]. [1] J. Timmermann et al., Phys. Rev. Lett. 125, 206101 (2020).