Theoretical materials chemistry develops and applies electronic structure and scale bridging methods.

Electronic structure methods

The basis of computational materials discovery is the continuous development of electronic structure methods. We develop new density functional theory methods which are distributed in WIEN2k, opens an external URL in a new window and GPAW, opens an external URL in a new window codes.

Machine learning

The availability of large datasets combined with new algorithms and vastly improved computing power has driven surge of in machine learning. While uses such as beating the best go players and self-driving cars have reached mainstream media, it is also revolutionizing materials chemistry. We apply both supervised and unsupervised methods as well as generative models to speed up atomistic simulations and discover new materials. We distribute the NeuralIL, opens an external URL in a new window neural network based force field and take part in the TACO SFB, opens an external URL in a new window.

Materials discovery

New materials for battery, thermoelectric, photovoltaic applications as well as catalysts are key to the solution of the grand challenges facing the world today. Using ab-initio methods we use the calculated thermochemistry and properties to discover and understand new materials.

Heat management poses a grand challenge that can make the whole difference between a working device and an abandoned design. With the mean free path of heat carrying phonons being in the micro-meter range, extending the ab-initio modelling of thermal transport to the mesoscale level will represent a tremendous advancement. With special focus on power electronics and thermoelectric materials we are developing a scale-bridging framework that opens the door to predictive modelling of materials of industrial interest. We distribute the BoltzTraP2 codes and are partners in the ALMAbte, opens an external URL in a new window consortium.