Session

DescriptionMachine learned potentials are an important tool in materials science and chemistry since they provide a highly accurate and computationally efficient approximation of the potential energy surface. Most of the currently used methods however, are based on a local description of atomic environments and are thus unable to describe effects which take place over long distances beyond the local cutoff, such as charge transfer or a change in the total charge of a system. The established methods can therefore not be applied to systems where such long range effects play an important role i.e. aromatic organic molecules, sp2 hybridized carbon systems or metal clusters adsorbed on doped substrates. This inherent shortcoming has recently gained attention which caused the development of a new generation of methods. In this minisymposium, we will discuss, with prominent experts in the field, the current challenges of incorporating long range interactions into machine learned potentials.