Random matrix central limit theorems for nonintersecting random walks

From MaRDI portal



Abstract: We consider nonintersecting random walks satisfying the condition that the increments have a finite moment generating function. We prove that in a certain limiting regime where the number of walks and the number of time steps grow to infinity, several limiting distributions of the walks at the mid-time behave as the eigenvalues of random Hermitian matrices as the dimension of the matrices grows to infinity.


Some stochastic models interpreted in terms of nonintersecting walks are recalled, their limit laws coincide with those known in random matrix theory. Then the walk \(S(t)=(S_0(t),\dots, S_k(t))\), \(0\leq t\leq 2\), of \(k+1\) particles in \(\mathbb{R}^{k+1}\), \(k\in \mathbb{Z}_+\), is studied; the time intervals between the steps equal \(2/N_k\), \(N_k \in \mathbb{N}\). It is conditioned on the events, in the trajectory space, of \(\{S_0(t) < \ldots < S_k(t)\), \(0\leq t\leq 2\}\) and the return of \(S(2)\) to the \(h_k\)-neighborhood of \(S(0)\). Assuming specific \(N_k\) growth and \(h_k\) decrease, as \(k\to \infty\), two main theorems show that the particles locations at \(t=1\) in two aspects behave statistically, after scaling, like the eigenvalues of a large Hermitian matrix from the Gaussian unitary ensemble. The proofs employ the Komlos-Major-Tusnady method to approximate \(S(t)\), with finite moment generating function for increments, by a nonintersecting Brownian bridge having the same start and end points. Next the arising limiting distributions for approximations are computed and identified.



Cites work


Cited in
(24)








This page was built for publication: Random matrix central limit theorems for nonintersecting random walks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2456030)