Potential energy surface for AlF-AlF collisions

This work presents a new approach to generating diatom-diatom potential energy surfaces (PES) through machine learning techniques. In particular, after using some energies at given geometries calculated at CCSD(T) (coupled cluster with single and double and perturbative triple excitations) level of theory, we employ a Gaussian process regression method based on a particular set of molecular features valid for all range of distances, i.e., a single model works for the long-range and short-range region of the PES. Finally, as an example, we calculate the PES for AlF-AlF and the density of states and lifetime of intermediate complexes via the developed machine learning approach.