Eigenvalue attraction

DOI10.1007/S10955-015-1424-5MaRDI QIDQ267728zbMATH OpenOpenAlexFDO

Publication date 11 April 2016

Published in Journal of Statistical Physics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1404.4113

Brownian motion circulant matrices eigenvalue perturbation theory Hatano-Nelson model Hermitian matrices normal matrices random matrix theory symmetric matrices

Mathematics Subject Classification ID

Eigenvalues, singular values, and eigenvectors (15A18) Random matrices (algebraic aspects) (15B52) Hermitian, skew-Hermitian, and related matrices (15B57) Random matrices (probabilistic aspects) (60B20) Brownian motion (60J65)

Abstract: We prove that the complex conjugate (c.c.) eigenvalues of a smoothly varying real matrix attract (Eq. 15). We offer a dynamical perspective on the motion and interaction of the eigenvalues in the complex plane, derive their governing equations and discuss applications. C.c. pairs closest to the real axis, or those that are ill-conditioned, attract most strongly and can collide to become exactly real. As an application we consider random perturbations of a fixed matrix

M

. If

M

is Normal, the total expected force on any eigenvalue is shown to be only the attraction of its c.c. (Eq. 24) and when

M

is circulant the strength of interaction can be related to the power spectrum of white noise. We extend this by calculating the expected force (Eq. 41) for real stochastic processes with zero-mean and independent intervals. To quantify the dominance of the c.c. attraction, we calculate the variance of other forces. We apply the results to the Hatano-Nelson model and provide other numerical illustrations. It is our hope that the simple dynamical perspective herein might help better understanding of the aggregation and low density of the eigenvalues of real random matrices on and near the real line respectively. In the appendix we provide a Matlab code for plotting the trajectories of the eigenvalues.

Review text

The conventional wisdom from random matrix theory is that eigenvalues repel one another. This stems from the fact that the degeneracy of an eigenvalue is a codimension two phenomenon for real symmetric matrices and a codimension-three phenomenon for complex Hermitian matrices. For reasonable distributions on the entries of the matrix, the probability that two eigenvalues are close is small for real symmetric matrices and yet smaller for complex Hermitian matrices. Something particularly relevant to this paper is that if the entries of the matrix \(M_0\) evolve under symmetric (or Hermitian) Brownian motion, \(t \mapsto M(t)\), \(M(0) = M_0\) a.s., then the eigenvalues evolve according to Dyson Brownian motion -- with each eigenvalue exerting a repulsive force on the others. A fact discussed in this paper is that this repulsion is a consequence of the symmetry conditions, not Brownian motion. The main result of the paper is that for a smooth flow on real matrices \(t \mapsto M(t)\), with no symmetry conditions, complex conjugate pairs of eigenvalues (non-zero imaginary part) exert a (possibly zero) attractive force on one another. The force is inversely proportional to the magnitude of the imaginary part of the eigenvalue. Thus, in the absence of forces from other eigenvalues, the acceleration vector of the eigenvalues in a complex conjugate pair will be directed toward the real axis (assuming \(\ddot M(t) = 0\)). As evidenced by the simple matrix \[ M(t) = \begin{pmatrix} 1 & 1 \\ -1 & 0 \end{pmatrix} + t \begin{pmatrix} 0 & 1 \\ -1 & 0\end{pmatrix}, \] the attraction property does not mean the eigenvalues will collide as \(t \to \infty\). Assume \(\lambda(t)\) is a complex eigenvalue, \(\lambda(0) > 0\) with \(\dot \lambda(0) = v_0 > 0\). Generally, if \(\ddot \lambda(t) = -i a(t)\) where \(a(t) \geq 0\) is integrable and if \(v_0\) is sufficiently large, then \(\dot \lambda(t) > 0\) for all \(t\) -- there is no collision on the real axis. Nonetheless, with appropriately unbiased random dynamics it is conceivable that some eigenvalues will experience forces that drive them to collide on the real axis. This paper clearly drives this intuition. The paper specializes the results to normal and circulant matrices. The results are further specialized to some examples, including the Hatano-Nelson model. \texttt{MATLAB(R)} code is included.

Recommendations

Cites work

Cited in

(7)

Describes a project that uses

Uses Software

Matlab

This page was built for publication: Eigenvalue attraction

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q267728)