SKM 2023 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 36: Data Driven Materials Science: Big Data and Work Flows – Microstructure-Property-Relationships (joint session MM/CPP)
MM 36.3: Talk
Thursday, March 30, 2023, 10:45–11:00, SCH A 251
A Machine-Learning Framework to Identify Equivalent Atoms at Real Crystalline Surfaces — •King Chun Lai, Sebastian Matera, Christoph Scheurer, and Karsten Reuter — Fritz Haber Institute of the Max Planck Society, Berlin, Germany
Functional surfaces and interfaces even of crystalline materials are characterized by breaks of symmetry and long-range order. Yet, even though such a crystalline surface may for instance exhibit numerous vacancies, adatoms, steps, kinks or islands, there are generally still many equivalent atoms, where equivalence refers to an identical or near-identical local environment. There are many equivalent terrace atoms, adatoms, step or kink atoms. In atomic-scale modeling and simulation, identifying these groups of equivalent atoms is a routine task, not least because one would e.g. restrict demanding first-principles calculations like the determination of an adsorption configuration and concomitant adsorption energy to only one site of each equivalence group. Aiming to automatize this routine task, we here present a machine-learning framework to identify all groups of equivalent atoms for any surface or nanoparticle geometry. The initial classification rests on the representation of the local atomic environment through a high-dimensional smooth overlap of atomic positions (SOAP) vector. We then achieve a fuzzy classification by mean-shift clustering within a low-dimensional embedded representation of the SOAP points as obtained by multidimensional scaling (MDS). The performance of this classification framework will be demonstrated with examples of Pd surfaces.