Application of the extended $q$-discrete Toda equation to computing eigenvalues of Hessenberg totally nonnegative matrices

TL;DR

This paper extends the q-discrete Toda equation and applies it to compute eigenvalues of Hessenberg totally nonnegative matrices, demonstrating convergence and numerical verification of the method.

Contribution

It introduces a new extension of the q-discrete Toda equation and applies it to Hessenberg TN matrices, linking it with shifted LR transformations and convergence analysis.

Findings

The extended q-discrete Toda equation effectively computes Hessenberg TN eigenvalues.

Numerical examples confirm convergence to the true eigenvalues.

The approach connects integrable systems with matrix eigenvalue problems.

Abstract

The Toda equation is one of the most famous integrable systems, and its time-discretization is simply the recursion formula of the quotient-difference (qd) algorithm for computing eigenvalues of tridiagonal matrices. An extension of the Toda equation is the q-Toda equation, which is derived by replacing standard derivatives with the so-called q-derivatives involving a parameter q such that 0 < q < 1. In our previous paper, we showed that a discretization of the q-Toda equation is shown to be also applicable to computing tridiagonal eigenvalues. In this paper, we consider another extension of the q-discrete Toda equation and find an application to computing eigenvalues of Hessenberg totally nonnegative (TN) matrices, which are matrices where all minors are nonnegative. There are two key components to our approach. First, we consider the extended q-discrete equation from the perspective of shifted LR transformations, similarly to the discrete Toda and its q-analogue cases. Second, we clarify asymptotic convergence as discrete-time goes to infinity in the qdiscrete Toda equation by focusing on TN properties. We also present two examples to numerically verify convergence to Hessenberg TN eigenvalues numerically.