Title:Approximating Excited States using Neural Networks

Abstract:We propose a new neural network method to solve the eigenstate problem in quantum mechanics. We build a neural network representation and combine it with variational Monte Carlo. Our method can approximate both the ground states and excited states, while the previous variational based methods only approximated the ground states. The successful approximation of excited states is realized by adding orthogonal constraints to energy expectation. We discuss both one-dimensiona land two-dimensional conditions, and the results are in good agreement with theoretical values.