Solutions of random-phase approximation equation for positive-semidefinite stability matrix

Abstract

It is mathematically proven that, if the stability matrix $\mathsf{S}$ is positive-semidefinite, solutions of the random-phase approximation (RPA) equation are all physical or belong to Nambu-Goldstone (NG) modes, and the NG-mode solutions may form Jordan blocks of $\mathsf{N}\mathsf{S}$ ($\mathsf{N}$ is the norm matrix) but their dimension is not more than two. This guarantees that the NG modes in the RPA can be separated out via canonically conjugate variables.