Approaching Thouless Energy and Griffiths Regime in Random Spin Systems By Singular Value Decomposition
We employ singular value decomposition (SVD) to study the eigenvalue spectra of random spin systems. By SVD, eigenvalue spectrum is decomposed into orthonormal modes $W_k$ with weight $λ_k$. We show that the scree plot ($λ_k$ with respect to $k$) in the ergodic phase contains two branches that both follow power-law $λ_k\sim k^{-α}$ but with different exponents $α$. By evaluating $W_k$, it's verified the part of $λ_k $ with $k>k_{\text{Th}}$ is universal that follows random matrix theory, where $k_{\text{Th}}$ is related to the Thouless energy. We further demonstrate that $α$ corresponds only to the exponential part of the level spacing distribution while being insensitive to the level repulsion, or equivalently the system's symmetry. Consequently, $α$ gives an underestimation for the many-body localization transition point, which suggests a non-ergodic behavior that may be attributed to the Griffiths regime.