Ph.D. Student: Machine-Learning DFT Simulation Package for Electronic Structures under Extreme Conditions 

The Center for Advanced Systems Understanding (CASUS) is a German-Polish research center for data-intensive digital systems research. We combine innovative methods from mathematics, theoretical systems research, simulations, data science, and computer science to provide solutions for a range of disciplines – materials science under ambient and extreme conditions, earth system research, systems biology, and autonomous vehicles.

CASUS was jointly founded in August 2019 by the Helmholtz-Zentrum Dresden-Rossendorf, the Helmholtz Centre for Environmental Research, the Max Planck Institute of Molecular Cell Biology and Genetics, the Technical University of Dresden and the University of Wroclaw. CASUS is located in the heart of Görlitz at the border between Germany and Poland. The CASUS start-up phase is hosted by the Helmholtz-Zentrum Dresden-Rossendorf and is financed by the Federal Ministry of Education and Research and the Saxon State Ministry of Science and Art.

The Department on Matter under Extreme Conditions is looking for a Ph.D. student interested in developing a Machine-Learning DFT Simulation Package for Electronic Structures under Extreme Conditions. Consideration of candidates will begin immediately and will continue until the position is filled.

The Scope of Your Job

Your project will contribute to the ambitious long-term goal of achieving a more accurate and consistent understanding of HED phenomena in the warm dense regime across multiple length and time scales. You are expected to develop a machine-learning (ML) DFT simulation package for calculating energies and atomic forces in configurations of atoms, at a scale and cost unattainable with direct DFT algorithms. You will implement a computational workflow that predicts the local density of states (LDOS) at each grid point in real space as a function of its local environment based on high-fidelity training data generated from DFT-MD. You will also investigate how the accuracy of the ML-DFT methodology varies with respect to the ML methods used (convolutional neural networks, sequence learning, and natural language processing) and the extent of the physics requirements captured.  You will verify the effectiveness of the ML-DFT simulation package with calculations of the equation of state of aluminum and silica as surrogates exhibiting the typical challenges encountered with conventional first-principle methods. You will carry out your research in collaboration with our partners at international research institutions.

Submit your application (including a one-page cover letter, CV, academic degrees, transcripts, etc.) online on the HZDR application portal. Further details can be found there under Job-Id: 76/2020 (997).