Hauser Group: Development of a neural network for metal cluster simulations
We are looking for a postdoctoral researcher (12 months with optional extension) with a solid theoretical background in quantum mechanics, numerical methods and excellent programming skills. The aim of this project is the development of multi-layer feed-forward neural network for the computation of properties of mixed-metallic nanoclusters (equilibrium structures, adsorption energies, optical and chemical properties). This type of network design has recently been suggested by Behler and Parrinello, and determines the total energy of the system by a summation over single atomic energy contributions evaluated for their given local environment. The network will be developed in Python (from scratch) and trained with ab initio data generated with standard program packages such as VASP, Q-Chem and others. The researcher will be tightly collaborating with Ralf Meyer, who is doing his PhD thesis on the subject. They will be supported by two master students; see below for details. The FWF project starts in December 2016.