Authors

Abstract

Single-crystal inelastic neutron-scattering (INS) data contain rich information about the structure and dynamics of a material. Yet the challenge of matching sophisticated theoretical models with large data volumesis compounded by computational complexity and the ill-posed nature of the inverse scattering problem. Herewe utilize a novel machine-learning (ML)-assisted framework featuring multiple neural network architecturesto address this via high-dimensional modeling and numerical methods. A comprehensive data set of diffractionand INS measured on the Kitaev material RuCl3 is processed to extract its Hamiltonian. SemiclassicalLandau-Lifshitz dynamics and Monte-Carlo simulations were employed to explore the parameter space ofan extended Kitaev-Heisenberg Hamiltonian. A ML-assisted iterative algorithm was developed to map theuncertainty manifold to match experimental data, a nonlinear autoencoder was used to undertake informationcompression, and radial basis networks were utilized as fast surrogates for diffraction and dynamics simulationsto predict potential spin Hamiltonians with uncertainty. Exact diagonalization calculations were employed toassess the impact of quantum fluctuations on the selected parameters around the best prediction.