Entanglement spectrum and block eigenvalue spacing distribution of correlated electron states

TL;DR

This paper investigates the entanglement spectrum of finite correlated electron systems, revealing how it characterizes metal-insulator crossover, deviates from conformal field theory predictions in the crossover regime, and exhibits random matrix behavior.

Contribution

It introduces a modified CFT prediction for finite-size entanglement spectra and links the level spacing distribution to random Toeplitz matrices, advancing understanding of entanglement in correlated systems.

Findings

Entanglement spectrum characterizes insulator-metal crossover.

Deviations from CFT predictions occur in the crossover regime.

Level spacing distribution resembles random Toeplitz matrix ensemble.

Abstract

Entanglement spectrum of finite-size correlated electron systems are investigated using the Gutzwiller projection technique. The product of largest eigenvalue and rank of the block reduced density matrix, which is a measure of distance of the state from the maximally entangled state of the corresponding rank, is seen to characterise the insulator to metal crossover in the state. The fraction of distinct eigenvalues exhibits a `chaotic' behaviour in the crossover region, and it shows a `integrable' behaviour at both insulating and metallic ends. The integrated entanglement spectrum obeys conformal field theory (CFT) prediction at the metal and insulator ends, but shows a noticeable deviation from CFT prediction in the crossover regime, thus it can also track a metal-insulator crossover. A modification of the CFT result for the entanglement spectrum for finite size is proposed which holds in the crossover regime also. The adjacent level spacing distribution of unfolded non-zero eigenvalues for intermediate values of Gutzwiller projection parameter $g$ is the same as that of an ensemble of random matrices obtained by replacing each block of reduced density matrix by a random real symmetric Toeplitz matrix. It is strongly peaked at zero, with an exponential tail proportional to $e^{-(n/R)s}$, where $s$ is the adjacent level spacing, $n$ is number of distinct eigenvalues and $R$ is the rank of the reduced density matrix.