Poisson to GOE transition in the distribution of the ratio of consecutive level spacings

Abstract

Probability distribution for the ratio ($r$) of consecutive level spacings of the eigenvalues of a Poisson (generating regular spectra) spectrum and that of a GOE random matrix ensemble are given recently. Going beyond these, for the ensemble generated by the Hamiltonian $H_\lambda = (H_0+\lambda V)/\sqrt{1+\lambda^2}$ interpolating Poisson ($\lambda=0$) and GOE ($\lambda \rightarrow \infty$) we have analyzed the transition curves for $\langle r\rangle$ and $\langle \tilde{r}\rangle$ as $\lambda$ changes from $0$ to $\infty$; $\tilde{r} = min(r,1/r)$. Here, $V$ is a GOE ensemble of real symmetric $d \times d$ matrices and $H_0$ is a diagonal matrix with a Gaussian distribution (with mean equal to zero) for the diagonal matrix elements; spectral variance generated by $H_0$ is assumed to be same as the one generated by $V$. Varying $d$ from 300 to 1000, it is shown that the transition parameter is $\Lambda \sim \lambda^2\,d$, i.e. the $\langle r\rangle$ vs $\lambda$ (similarly for $\langle \tilde{r}\rangle$ vs $\lambda$) curves for different $d$'s merge to a single curve when this is considered as a function of $\Lambda$. Numerically, it is also found that this transition curve generates a mapping to a $3 \times 3$ Poisson to GOE random matrix ensemble. Example for Poisson to GOE transition from a one dimensional interacting spin-1/2 chain is presented.