Berlin 2018 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

MM: Fachverband Metall- und Materialphysik

MM 47: Poster Session II

MM 47.27: Poster

Wednesday, March 14, 2018, 18:30–19:45, Poster C

Reduction of Redundant Quantum Mechanical Computations Using Machine Learning Methods for Nanocatalysts — •Eiaki Morooka, Adam Foster, and Marc Jäger — Aalto University, Helsinki, Finland

Platinum Group Metals (PGMs) are used for fuel cells, batteries and for automobile filters called autocatalysts, and Europe dominates the platinum consumption. Nevertheless, there is no primary PGM production in the EU, and recycling remains limited, while PGMs are increasingly adopted in emerging technologies for green energy conversion devices. PGMs should be substituted by inexpensive, earth-abundant catalysts, such as bimetallic transition metals with non-metallic elements. These clusters must be rationally designed and fine-tuned by using combinations of several chemical elements, different structures and sizes, which is unfeasible for both experiments and traditional quantum mechanical computations. We are developing machine learning tools to screen nanoclusters using a state-of-the-art chemical descriptor called smooth overlap of atomic positions (SOAP) to drastically reduce quantum mechanical computations. Specifically, by scanning through the similarities of local chemical environments of surface hydrogens and eliminating the surface hydrogens with similar local chemical environments.