Teaching Machines Quantum Physics: Machine-Learned Functionals in Density Functional Theory

Density Functional Theory (DFT) is one of the most widely used tools in physics, chemistry, and materials science. It enables researchers to predict the properties of molecules and materials – from catalysts to semiconductors – at a fraction of the cost of laboratory experiments. The accuracy of DFT, however, depends on the choice of the exchange – correlation functional, an approximation that captures how electrons interact. Despite decades of effort, finding more accurate and broadly applicable functionals remains a central challenge.

In this talk, I will give a broad introduction to DFT and explain how new methods from deep learning can help tackle this problem. Using differentiable DFT, we can train neural networks directly within quantum simulations, offering a systematic way to construct functionals with improved accuracy. The talk will not assume prior knowledge and will highlight how quantum physics and machine learning are coming together to advance one of today’s most important computational tools.

I am a DPhil in Atomic and Laser Physics working on modelling the behaviour of electrons in gases and materials. During my undergraduate degree at RWTH Aachen University in Germany, I worked for the Fraunhofer-Institute for Lasertechnology (ILT) on a project aiming to build the world’s first nuclear clock (“Thorium Nuclear Clock”). Afterwards I worked in Switzerland on quantum cascade lasers for CO2 sensors. Outside of my work I love doing sports and going to a wide range of talks (on almost any subject really).