Jörg Neugebauer
Max Planck Institute, DE
Biography
Prof. Dr. Jörg Neugebauer is Director and Executive at the Max-Planck-Institut für Eisenforschung GmbH in Düsseldorf, Germany. His research areas include electronic structure methods, statistical physics and thermodynamics, computational materials science, theoretical solid-state
physics, defect chemistry, optoelectronics, electrochemistry, numerical algorithms, machine learning and workflows. He received his diploma and PhD degree from the Humboldt University Berlin and his Habilitation (venia legendi) from the Technical University Berlin. He worked as a visiting scientist at the Fritz-Haber Institute (FHI) in Berlin and the XEROX Palo Alto Research Center (CA/USA), before becoming head of an independent Max-Planck-Research group (C3) at
the FHI in 1998. In 2004 he was appointed Professor (C4) at the University of Paderborn and shortly afterwards head of Computational Materials Design at the Max-Planck-Institut für
Eisenforschung. He is a scientific member of the Max-Planck-Society, elected member of the
Academy of Sciences and Arts of North Rhine-Westphalia, a professor at the University of
Paderborn, an honorary professor at the Ruhr-Universität Bochum and director of the Advanced Study Group “Modelling” at ICAMS. He has served as elected member of the DFG-Review Boards “Chemistry” and “Materials Science”, elected Chair of the German Physical Society (DPG) Division “Metals and Materials” and is since 2019 member of the Supervisory Board of the Karlsruhe Institute of Technology (KIT). He is member of the Management Board of the Consortium MaterialDigital, of the NFDI-MatWerk-Consortium and speaker of the International Max-Planck- Research School for Sustainable Metallurgy (IMPRS SusMet). He received an ERC Advanced Grant for the Project „SMARTMET“ and the Ernst-Mach-Medal of the Czech Academy of Science.
He is author or co-authors of 25 books, book chapters and conference proceedings and over 400 scientific papers in academic journals.
Title
Maximizing High-Throughput Discovery and Machine Learning Efficiency Through Computational Workflows
Assisted by Dr. Sarath Menon
Abstract
The advent of high-throughput computation and discovery combined with machine learning
is revolutionizing the field of computational materials science. It enables the simulation of
large systems and complex material properties with ab initio accuracy. However, the
development of these data-driven activities is often a computationally complex and
intensive task, requiring the combination and orchestration of multiple and often
incompatible simulation codes. Automated, reliable, and robust computational workflows
are required to design and execute the underlying complex simulation protocols. Using the pyiron framework (pyiron.org), the tutorial provides a hands-on introduction to all aspects
of workflow design, testing, and execution, with a strong focus on materials science and
atomistic simulations.
Outline:
1 | Introduction
- FAIR data and workflows
- Introduction into pyiron
- Prototype examples of workflows in materials science
- Generating atomic structure datasets
- Computing static material properties
- Running MD simulations
- Analysis and refinement of DFT datasets
- Fitting ML potentials
- Validation of the potentials
- Defects of materials
- Material properties at finite temperature
- Thermodynamic phase diagrams