“…Therefore, to satisfy these requirements, analytical surrogate models of the potential energy surfaces (PESs) are typically employed. Various techniques were used in the past to model the dissociative chemisorption of H 2 at metal surfaces, , such as potentials based on the corrugation-reducing procedure (CRP) , for Cu(111), − ,,, Cu(211), Cu(100), Pd(100), and Ni(111), its extension, dynamic corrugation model (DCM) for Cu(111), , the modified Shepard interpolation , for Cu(111) and Pt(111), , or the permutation invariant polynomials (PIPs) , employing neural networks (PIP-NN) for Cu(111), Ag(111), , and Co(0001). , Besides the PIP-NN, other machine-learning (ML)-based models were employed to study hydrogen chemistry at metal surfaces in the past, such as the embedded atom neural network (EANN), which was also successfully employed for modeling dissociative chemisorption at multiple Cu surfaces . Although these methods have proven to be robust and accurate, most of them have clear limitations.…”