Exceptional points and domains of unitarity for a class of strongly non-Hermitian real-matrix Hamiltonians

TL;DR

This paper investigates the boundaries of unitarity in a class of real-matrix Hamiltonians with non-Hermitian perturbations, focusing on exceptional points and stability limits using symbolic and numerical methods.

Contribution

It provides a detailed analysis of the exceptional-point boundaries for strongly non-Hermitian Hamiltonians with real matrices, highlighting the shape and properties near stability extremes.

Findings

Identification of the quantum phase-transition boundary as exceptional points

Characterization of the boundary shape near strong-coupling extremes

Use of high-precision and symbolic computation techniques

Abstract

A phenomenological Hamiltonian of a closed (i.e., unitary) quantum system is assumed to have an $N$ by $N$ real-matrix form composed of a unperturbed diagonal-matrix part $H^{(N)}_0$ and of a tridiagonal-matrix perturbation $\lambda\,W^{(N)}(\lambda)$. The requirement of the unitarity of the evolution of the system (i.e., of the diagonalizability and of the reality of the spectrum) restricts, naturally, the variability of the matrix elements to a "physical" domain ${\cal D}^{[N]} \subset \mathbb{R}^d$. We fix the unperturbed matrix (simulating a non-equidistant, square-well-type unperturbed spectrum) and we only admit the maximally non-Hermitian antisymmetric-matrix perturbations. This yields the hiddenly Hermitian model with the measure of perturbation $\lambda$ and with the $d=N$ matrix elements which are, inside ${\cal D}^{[N]}$, freely variable. Our aim is to describe the quantum phase-transition boundary $\partial {\cal D}^{[N]}$ (alias exceptional-point boundary) at which the unitarity of the system is lost. Our main attention is paid to the strong-coupling extremes of stability, i.e., to the Kato's exceptional points of order $N$ (EPN) and to the (sharply spiked) shape of the boundary $\partial {\cal D}^{[N]}$ in their vicinity. The feasibility of our constructions is based on the use of the high-precision arithmetics in combination with the computer-assisted symbolic manipulations (including, in particular, the Gr\"{o}bner basis elimination technique).