Interpretable machine-learning identification of the crossover from subradiance to superradiance in an atomic array

Abstract

Light–matter interacting quantum systems manifest strong correlations that lead to distinct cooperative spontaneous emissions of subradiance or superradiance. To demonstrate the essence of finite-range correlations in such systems, we consider an atomic array under the resonant dipole–dipole interactions (RDDI) and apply an interpretable machine learning (ML) with the integrated gradients to identify the crossover between the subradiant and superradiant sectors. The machine shows that the next nearest-neighbor (NN) couplings in RDDI play as much as the roles of NN ones in determining the whole eigenspectrum within the training sets. Our results present the advantage of ML approach with explainable ability to reveal the underlying mechanism of correlations in quantum optical systems, which can be potentially applied to investigate many other strongly interacting quantum many-body systems.