In 2020, BAM launched its junior research group programme. Outstanding young scientists from Germany and abroad can set up their own independent junior research group and establish themselves in the community. Their research should be oriented towards BAM's thematic fields - energy, infrastructure, environment, materials, and analytical sciences. Dr. Janine George from the Materials Chemistry Department is currently in the process of setting up such a junior research group on the topic of "computational materials design". She is particularly interested in the data-driven search for new materials and the fundamental chemical understanding of their structure and properties.
You studied and researched at RWTH Aachen University, the Université catholique de Louvain and briefly at the University of Oxford. What brought you to BAM in Berlin?
The junior research group funding at BAM immediately interested me. Firstly, because I was allowed to develop the research topic independently within the framework of AI-based materials research and with reference to BAM's tasks, and there are also many opportunities for collaboration within BAM. And secondly, by having such secure financial means at the beginning of the junior research group career. To get a permanent position and such group funding at the same time at this point of career is very unusual in Germany. Typically, comparable positions are limited to 3 to 5 years and there is often less funding. BAM's location in Berlin also helped in the decision but was only one of many criteria.
You started with the lecture "Data-driven materials discovery and chemical understanding" - what fascinates you about this topic?
I think it very fascinating how well material properties of known and unknown materials can nowadays be predicted without any preconditions. This makes it possible to search for new materials on the computer for applications in the field of batteries or for solar cells. We can calculate huge databases of materials properties without having to synthesize and experimentally characterize all these materials. We can then search these databases or use them to apply machine learning methods. This is often not possible based on experimental data alone, as these are often obtained under different and sometimes unknown conditions, and we have far too little experimental data in many cases. Overall, this can help to find new materials or even to gain new chemical/physical understanding in this field.
What do you want to achieve with your junior group?
The aim is to develop new chemical rules that make it possible to search for materials with interesting applications more quickly and with more chemical understanding. Some materials properties can be calculated, but it takes a long time and eats up a lot of computing time. It would be nice to find more shortcuts to arrive at these properties. In addition, it currently takes a long time to find completely new materials that can be synthesised. Here, too, I would like to develop shortcuts based on chemical/physical rules. For example, the periodic table of the elements and the relationships between the elements that can be derived from it offer such possibilities for more than 150 years. There are many similar rules/heuristics in chemistry. As part of this research, we will develop some automations for calculations and publish our own databases of material properties.
In an article published in Trends in Chemistry in 2021, you write about "Chemist versus Machine" - will there be more chemists or more programmers in your group?
The junior research group will work on computational materials science. This means that we will do research somewhere between chemistry, physics, and the materials sciences. One focus for us is the simulation of the electronic structure of materials and thus quantum mechanics. In addition, data analysis and machine learning will also play a role. The group will consist of materials scientists, physicists and chemists who already have experience in the field of simulation. Nowadays, there are also degree programmes at universities in Germany, for example, that are dealing with such simulations. Graduates in materials informatics could also be part of the group. So far, however, there are hardly any degree programmes in this field. Pure computer scientists would probably have a hard time because they usually do not have training in quantum mechanics and do not have the chemical and physical understanding for the simulations.
BAM has been researching safety in technology and chemistry for 150 years - how does this relate to AI-based materials research for you?
The data-driven material search allows us, for example, to look for materials consisting of non-toxic elements, which could certainly facilitate their use as solar cells, for example. We are also looking specifically at materials properties that could be important for the safety of components. Overheating of components can become a problem and thus, for example, thermal conductivity is an important property of a material in this context.