Digitalization: Artificial intelligence helps with material design
Pascal Friederich, tenure-track professor at the Karlsruhe Institute of Technology (KIT), won the prestigious Heinz Maier-Leibnitz Prize from the German Research Foundation (DFG) this year. We spoke to him about the use of artificial intelligence in material simulation and virtual material design.
Dear Mr. Friederich, before we start with the questions about your research, please tell our readers a few words about yourself.
Thank you for inviting me to the interview. I am currently a junior professor at the Faculty of Computer Science at the Karlsruhe Institute of Technology. I’ve been working there since 2020 on artificial intelligence methods and their use in materials science. I have already completed my doctorate in physics at KIT and then spent several years abroad, including at Harvard University in the USA and the University of Toronto in Canada. There, thanks to a Marie Curie Fellowship from the European Union, I was able to work with world-leading experts in my field of research before returning to Karlsruhe.
You conduct research in the field of materials science – and rely on the use of artificial intelligence (AI) and machine learning. How should we imagine this in practice?
Two things come together here: On the one hand, the great progress made in recent years in the field of machine learning – think of image recognition or translations, for example. Secondly, the efforts of recent years to make scientific data – both from experiments and material simulations – openly accessible worldwide and therefore usable. This opens up new opportunities to adapt the machine learning methods developed for images and language to the issues of materials science and to learn from the data. For example, we can predict material properties on the computer – even before the materials are produced in the laboratory.
Are there certain things in the field of materials science that are made possible by AI?
In simplified terms, many challenges in materials science can be described as optimization or search problems. This means that when developing new solar cells or batteries, for example, there are a large number of free parameters that need to be optimized: the composition of individual materials, the manufacturing conditions, the combination of different materials, and so on. The number of conceivable individual materials alone is huge and has not yet been researched to a large extent. Machine learning methods can make a decisive contribution to speeding up this search by designing materials with the desired properties on the computer and suggesting precisely those experiments that are most informative and therefore most effective for optimizing the material properties.
They recently received the 20,000 euro Heinz Maier-Leibnitz Prize from the German Research Foundation (DFG). The corresponding press release from the Karlsruhe Institute of Technology (KIT) talks about the “increasing demand for high-performance materials”. Do you have a few examples from everyday life for us?
Materials as well as new molecules and molecular materials are needed for many pressing challenges. This ranges from the examples already mentioned, such as solar cells and batteries for the conversion and storage of energy, to the development of new active ingredients and medicines, to the development of new functional materials for quantum computers.
My research here focuses on the development of widely applicable methods for researching new materials, which hopefully can then be used successfully for a variety of different material classes.
What makes the Karlsruhe research environment special for you and how do you experience the cooperation with non-university institutions?
Karlsruhe and KIT in particular offer a wide range of opportunities, not all of which I have certainly exhausted yet. KIT has an excellent reputation not only in the German research landscape but also internationally, which is also reflected in the internationality of the applicants and then also of my research group. The students are highly motivated and very interested in current research, especially in interdisciplinary topics.
The KIT is not only a university, but also a research center of the Helmholtz Association. In addition, there are many research-intensive companies in the Karlsruhe area. This enables us to engage in a wide range of scientific collaborations so that we can put our methods directly into practice.
Photo: KIT, Amadeus Bramsiepe