On September 15^{th} our colleague Maarten Cools-Ceuppens successfully defended his PhD. During his defence Maarten presented his work entitled ‘Incorporating long-range interactions and polarization in machine learning potentials with explicit electrons’. He was supervised by prof. dr. ir. Toon Verstraelen and prof. dr. ir. Joni Dambre.

*Congratulations Maarten!*

__Summary of the PhD in laymen's terms__

With molecular modeling, one can simulate the behavior of all kinds of materials at the level of individual atoms. At this length scale, systems composed of electrons and nuclei adhere to the laws of quantum mechanics, meaning that they can be simulated by solving the Schrödinger equation. In realistic applications, this approach requires an enormous amount of computing power, which remains challenging, even with today's supercomputers. In practice, one must resort to approximate methods instead. A force field is such a popular approximate method, in which electrons are no longer described explicitly, enabling simulations of larger atomistic systems over longer time scales.

For advanced applications, e.g. those involving charge transfer or polarization, conventional force fields do not capture all the relevant physics. Therefore, in this PhD, we have developed a new force field, in which electrons are re-introduced as semi-classical particles. This force field does capture the physics of charge transfer and polarization, yet without the computational burden of quantum-mechanical models. A machine-learning model was trained to incorporate the complex short-range quantum-mechanical interactions between semi-classical electrons. This new model can indeed learn these interactions automatically and truthfully, which was confirmed by accurate predictions of various electronic properties of organic molecules and periodic atomistic models of solids.