Total: 1
We formulate a physics-informed data-driven method for modeling anisotropic pairwise interactions in the presence of long-ranged electrostatics. The method separates the total interaction into a long-ranged electrostatic interaction that is approximated using a multipole expansion truncated at the dipole level and a short-ranged residual interaction that is approximated using multivariate Chebyshev polynomials fit to measurements from a limited number of configurations. We assess the approach on a sequence of aromatic molecules (benzene, benzonitrile, and phenoxide), finding that it produces satisfactory results using a modest cutoff distance for the short-ranged interaction. This method has applications for modeling complex interactions for, and conducting dynamic simulations of, synthetic and biological materials with charge.