The International Center for Advanced Energy Conversion at Surfaces supported the Workshop “Developing High-Dimensional Potential Energy Surfaces: From the Gas Phase to Materials” that took place in Göttingen from the 24th to 26th of April in 2019. The Workshop was jointly organized by Jörg Behler (Georg-August-Universität Göttingen, Germany), Joel Bowman (Emory University, USA), Cábor Csányi (Cambridge University, UK), and Alexander Kandratsenka (MPI for Biophysical Chemistry, Göttingen, Germany).

The rapid progress in modern machine learning (ML) techniques has important consequences for almost all fields of science including chemical sciences and closely related fields like materials science and condensed matter physics. Among the most rapidly evolving applications of ML methods is their use in computer simulations with the aim to understand complex chemical reactions and to quantitatively predict properties of new materials. For this purpose, substantial progress has been made in the development of a new generation of accurate, ab initio based atomistic ML potentials, which provide a direct relation between the atomic configuration and the potential energy.

The workshop brought together leading researchers who develop and apply Machine Learning techniques with the common goal of determining the fundamental properties of “small” molecules, biomolecules, and materials.  These properties include high-dimensional potential energy surfaces, atomic densities, and molecular properties, such as dipole moments and polarizabilities.  Communities in the areas of materials, biomolecules, gas-phase molecules and complexes have formed over the past ten or so years and a major objective of the workshop was to bring these communities together to hear and learn from each other’s experience.