[Paper Review] Double scaling limit in the random matrix model: the Riemann-Hilbert approach

Pavel Bleher, Alexander Its|ArXiv.org|Jan 2, 2002

Random Matrices and Applications34 references6 citations

TL;DR

This paper establishes the existence of the double scaling limit in the unitary random matrix model with quartic interaction using the Riemann-Hilbert approach. It proves that correlation functions in this limit are governed by an integrable kernel derived from the psi-function of the Hastings-McLeod solution to the Painlevé II equation, providing a rigorous asymptotic description of eigenvalue statistics near the critical point where the eigenvalue support transitions from one to two intervals.

ABSTRACT

We prove the existence of the double scaling limit in the unitary matrix model with quartic interaction, and we show that the correlation functions in the double scaling limit are expressed in terms of the integrable kernel determined by the psi-function for the Hastings-McLeod solution to the Painlevé II equation. The proof is based on the Riemann-Hilbert approach.

Motivation & Objective

• To rigorously establish the existence of the double scaling limit in the unitary random matrix model with quartic interaction.
• To characterize the asymptotic behavior of correlation functions near the critical point $ t_c = -2\sqrt{g} $, where the eigenvalue support changes from one to two intervals.
• To show that the limiting correlation kernel is expressible in terms of the $ \psi $-function of the Hastings-McLeod solution to the Painlevé II equation.
• To develop a Riemann-Hilbert framework that captures the universal behavior of eigenvalue statistics in the double scaling regime.

Proposed method

• The Riemann-Hilbert problem for orthogonal polynomials is formulated in the double scaling limit, with a focus on the critical scaling near $ t_c $.
• A nonlinear steepest descent method is applied to the Riemann-Hilbert problem, using a Deift-Zhou-type analysis with appropriate contour deformations and parametrix constructions.
• The parametrix is constructed using the solution to the Painlevé II equation, specifically the Hastings-McLeod solution, to model local behavior near the hard edge of the spectrum.
• The error between the exact solution and the parametrix is estimated using $ L^2 \times L^\infty $ norms on the jump contour, yielding $ O(N^{-1/3}) $ and $ O(N^{-1}) $ error terms.
• The asymptotic expansion of the correlation kernel is derived via the Christoffel-Darboux formula and the resulting kernel is shown to be integrable and related to the $ \psi $-function of the Painlevé II equation.
• The method incorporates a change of variable $ \zeta(z) $ that simplifies the structure of the jump matrix and enables a compact representation of the leading-order asymptotics.

Experimental results

Research questions

• RQ1 Does the double scaling limit exist in the unitary random matrix model with quartic potential, and what is its universal structure?
• RQ2 How do the correlation functions behave asymptotically as $ N \to \infty $ and $ t \to t_c $ with appropriate scaling?
• RQ3 Can the limiting kernel be expressed in terms of special functions, and if so, which ones?
• RQ4 What is the role of the Painlevé II equation in describing the universal behavior of eigenvalue statistics at the edge of the spectrum?

Key findings

• The double scaling limit exists for the quartic matrix model with $ V(M) = \frac{t}{2}M^2 + \frac{g}{4}M^4 $, $ g > 0 $, as $ N \to \infty $ and $ t \to t_c = -2\sqrt{g} $.
• The correlation functions in the double scaling limit are governed by an integrable kernel derived from the $ \psi $-function of the Hastings-McLeod solution to the Painlevé II equation.
• The asymptotic expansion of the orthogonal polynomials and the correlation kernel is obtained with an error term of order $ O(N^{-1}) $, improving upon previous $ O(N^{-1/3}) $ estimates.
• The limiting eigenvalue density transitions from a single interval $ [-a,a] $ to a two-interval support $ [-a,-b] \cup [b,a] $ at $ t = t_c $, with $ p(0) = 0 $ in the critical case.
• The Riemann-Hilbert approach allows for a fully rigorous derivation of the universal kernel without relying on prior assumptions about the asymptotic structure of the recurrence coefficients.
• The method provides a systematic framework to analyze the double scaling limit, with the Painlevé II equation emerging naturally from the nonlinear steepest descent analysis.