The group in front of Mathildenhöhe (03.2019)
The work we have been pursuing…
Previously, we focused mostly on studying phenomena in magnetic materials using first-principles (DFT, DFT+U, DFT+DMFT) methods, in particular properties driven by spin-orbit coupling.
The main topics are (as illustrated by the sketch below):
- effects of spin-orbit coupling on equilibrium magnetic properties
- spin/charge topological transports (longitudinal, transverse, mesoscopic)
- interplay of spin-orbit coupling and electronic correlations
- interfacial effects
Over the last years, motivated by the confidence from our first-principles calculations, we are working more in the direction of high throughput design of functional materials, mainly (again) magnetic compounds. As there are more good hands around, we are capable of covering:
- ab initio thermodynamics, combining DFT and CALPHAD for phase diagram optimization;
- high throughput screening of (meta-)stable compounds, with systematic evaluation of thermodynamic, mechanical, and dynamic stabilities;
- high throughput characterization of physical properties, such as topological properties;
- multi-scale simulation of magnetic materials.
Yes, you are right, magnetic materials are our beloved. The reason is simply because we still do not understand the magnetism while everything is magnetic.
Although statistics does not guarantee causality, machine learning brings forth a great opportunity for materials science, i.e., it can serve as the fourth paradigm to engineer functional materials. Our work in this direction falls into two categories:
- with data obtained from experiments which can be collected from the literature by natural language processing;
- with data obtained from high throughput DFT calculations, such as interatomic forces.
As you can see, machine learning modeling of materials is data hungry. Nevertheless, the hope is that the real inverse design can be achieved. Stay tuned!