Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 88: Catalysis and surface reactions III
O 88.3: Vortrag
Donnerstag, 12. März 2026, 15:30–15:45, TRE/MATH
From Processes to Mechanisms: Machine-Learning Based Automated Lattice Mapping as Prerequisite to Microkinetic Modeling — •Aditya Kumar1, Patricia Poths1, King Chun Lai2, Christoph Scheurer1, Sebastian Matera1, and Karsten Reuter1 — 1Fritz-Haber-Institut der MPG, Berlin — 2Max Planck Computing and Data Facility, Garching
Finding the reaction mechanism of a surface reactive system from first principles is a tedious task requiring a lot of human effort and being affected by human bias. To address these points, we have developed the Automatic Process Explorer (APE), which generates a comprehensive library of elementary reaction steps [1]. With easily thousands of steps identified, their manual assembly into a reaction network as a prerequisite to subsequent microkinetic modeling becomes intractable. To this end, we develop an automated framework that infers an effective lattice onto which all APE-discovered elementary steps may be coarse-grained. In this, atomic configurations from APE are analyzed using density-based hierarchical clustering, creating a fine-to-coarse hierarchy of potential lattice sites. All processes are then mapped onto the lattice as discrete occupation changes and, exploiting symmetry, a network of unique elementary reactions on this lattice is generated. We demonstrate the approach for a large set of elementary processes within the initial oxidation of Pd(100) terrace edges, and discuss the feasibility of subsequent efficient lattice-based microkinetic simulations.
[1] K.C. Lai et al., Phys. Rev. Lett. 134, 096201 (2025).
Keywords: Lattice discovery; mechanism mining; restructuring