Tridiagonal Models for Dyson Brownian Motion

Publication date 10 July 2017

Abstract: In this paper, we consider tridiagonal matrices the eigenvalues of which evolve according to

-Dyson Brownian motion. This is the stochastic gradient flow on

m a t h b b R^{n}

given by, for all

1 l e q i l e q n,

[ dlambda_{i,t} = sqrt{frac{2}{�eta}}dZ_{i,t} - �iggl( frac{V'(lambda_i)}{2} - sum_{j: j

eq i} frac{1}{lambda_i - lambda_j} �iggr),dt ] where $V$ is a constraining potential and $l e f t {Z_{i, t} i g h t}_{1}^{n}$ are independent standard Brownian motions. This flow is stationary with respect to the distribution [ ho^{�eta}_N(lambda) = frac{1}{Z^{�eta}_N} e^{-frac{�eta}{2} left( -sum_{1 leq i

eq j leq N} log|lambda_i - lambda_j| + sum_{i=1}^N V(lambda_i) ight) }. ] The particular choice of

V (t) = 2 t^{2}

leads to an eigenvalue distribution constrained to lie roughly in

(- s q r t n, s q r t n) .

We study evolution of the entries of one choice of tridiagonal flow for this

V

in the

n o i n f t y

limit. On the way to describing the evolution of the tridiagonal matrices we give the derivative of the Lanczos tridiagonalization algorithm under perturbation.

This page was built for publication: Tridiagonal Models for Dyson Brownian Motion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6288811)