For their outstanding doctoral theses, Juliane Heitkämper and Samuel Tovey have been awarded the University Prize of the University of Stuttgart, funded by the Christian and Dorothee Bürkert Foundation. Both completed their doctorates in the environment of the Stuttgart Center for Simulation Science (SimTech): Juliane Heitkämper as a doctoral researcher under the supervision of Johannes Kästner in theoretical chemistry, and Samuel Tovey in the group of Christian Holm at the Institute for Computational Physics.
In her thesis “Computational Investigation of Catalytic Reaction Mechanisms”, Juliane Heitkämper addresses the question of how chemical reactions can be predicted on the computer so accurately that they can be planned and improved in a targeted way. She studies reactions that use catalysts – substances that make reactions faster or more selective. Using advanced quantum-chemical calculations, she investigates which intermediate states occur in these reactions, which pathways the molecules actually follow, and what role factors such as solvents or additives play. One part of her work focuses on hydroboration, a reaction in which a boron compound and hydrogen are added to an organic molecule to transform it into a desired product. By comparing her simulations with experimental data, she can show which reaction pathways are truly relevant and how yield and selectivity can be improved. Her thesis thus contributes to enabling catalytic reactions to be designed more systematically on the computer, rather than relying primarily on time-consuming trial-and-error in the laboratory.
The thesis of Samuel Tovey, entitled “Physics meets Machine Learning: Theory and Application”, builds a bridge between physics and artificial intelligence. On the one hand, he develops methods that use machine learning to efficiently describe physical systems, such as many interacting particles. In this way, simplified models can be derived from highly accurate but very expensive quantum-mechanical calculations and then used in large-scale computer simulations. On the other hand, he applies ideas from physics to better understand learning processes in neural networks: for example, he investigates how to select training data smartly so that a model learns reliably from as little data as possible, and which quantities are suitable to characterize the “state” of a network. His work demonstrates how physical thinking can help to make modern AI methods more transparent and more efficient.
The joint award of the University Prize to Juliane Heitkämper and Samuel Tovey highlights the strength of SimTech at the University of Stuttgart: simulations and data-driven methods are used here across very different fields – from predicting chemical reactions to understanding artificial neural networks.