Nonintersecting Brownian motions on the unit circle (Q282506): Difference between revisions

The paper under review shows that the Pearcey and (symmetric) tacnode processes arise as limits of the simplest model of nonintersecting Brownian motions on a circle and analyzes the total winding number of the particles on the circle, that has no counterpart in the models on the real line. The nonintersecting Brownian motions on the circle with \(n\) particles are functions \(x: [0, T]\to S^1 \times \cdots \times S^1\) such that \(x(t)= (x_1(t), \dots, x_n(t))\) with \(x_k (t) = e^{i B_k(t)/\sqrt{n}}\). The model NIBM\(_{0\to T}\) refers to \(x_k(0)=x_k(T)=1\) for all \(1\leq k \leq n\). The transition probability density function \(P(A_n, B_n, t) =\det (P(a_i, b_j, t; \varepsilon (n)))_{1\leq i, j\leq n}\) for \(A_n=\{a_1, \dots, a_n\}\) and \(B_n=\{b_1, \dots, b_n\}\) follows from the Karlin-McGregor formula for odd \(n\) and a more careful study for even \(n\). The model NIBM\(_{0\to T}\) splits into the subcritical case (\(T< \pi^2\)), the critical case (\(T=\pi^2\)) related to the Painlevé II equation and Lax pairs, and the supercritical case (\(T>\pi^2\)) related to elliptical integrals. Theorem 1.2 summarizes the asymptotic behavior of \(P(\text{total\;winding\;number\;equal}\;0)\) as \(n\to\infty\); for the subcritical case, it is exponentially close 1; for the critical case, it is polynomial \(n^{-1/3}\) close to 1; for the supercritical case, \(P(\text{total\;winding\;number\;equals}\;w)= q^{w^2}\sqrt{\frac{\pi}{2\tilde{K}} }+o(n^{-1})\), where \(q=\exp (- \frac{\pi K(\sqrt {q-\tilde{k}^3}}{K(\tilde{k})}))\). Theorem 1.3 states that the asymptotic limit of the correlation kernel \(K_{t_i, t_j}(x, y)|\frac{dy}{d\eta}|\) equals \(K^{\mathrm{Pearcey}}_{-\tau_j, \tau_i}(\eta, \xi)\) for the supercritical case and \(K^{\mathrm{tac}}_{-\tau_j, \tau_i}(\eta, \xi; \sigma)\) for the critical case, and the limit of the multi-time correlation function is given in Theorem 1.4 for the supercritical case. Section 2 derives the transition probability density of NIBM and the joint correlation function and the correlation kernel of NIBM\(_{0\to T}\). By applying results of \textit{S. Karlin} and \textit{J. McGregor} [Pac. J. Math. 9, 1141--1164 (1959; Zbl 0092.34503)] and \textit{M. Fulmek} [Sémin. Lothar. Comb. 52, B52b, 16 p. (2004; Zbl 1064.05013)], the authors derive the \(\tau\)-deformed versions. Section 3 characterizes the asymptotic results for the discrete Gaussian orthogonal polynomials which are used in the later Sections 4 and 5. The key ingredient in the Riemann-Hilbert analysis of orthogonal polynomials is the equilibrium measure associated with the weighted function. For the subcritical and critical cases, the equilibrium measure for the discrete Gaussian orthogonal polynomials is given by the Wigner semicircle law of \textit{P. Deift} [Orthogonal polynomials and random matrices: a Riemann-Hilbert approach. New York, NY: Courant Institute of Mathematical Sciences; Providence, RI: American Mathematical Society (AMS) (2000; Zbl 0997.47033)], as is proved in Proposition 3.1; for the supercritical case, the density function of the equilibrium measure is given in Proposition 3.3. The role of the \(g\)-function defined by \(\int \log (z-x) \rho_T(x) \,dx\) in the asymptotic limits of the paper is crucial, and its derivative relations are explicitly given in Proposition 3.4 for all three cases. Proposition 3.6 gives the asymptotic limit for the discrete Gaussian orthogonal polynomials for the supercritical case, and Proposition 3.7 for the critical case. Section 4 is devoted to the proof of Theorem 1.2. by deriving the differential equation for the Hankel determinant in Proposition 4.1, and by the probability related to the ratio of the Hankel determinant with an integral. The asymptotical limit of the \(\tau\)-deformed correlation kernel is first related to the contour integral formula of the \(\tau\)-deformed correlation kernel and with the Christoffel-Darboux formula in Section 5.1. Limiting Pearcey processes and limiting tacnode processes are discussed in Section 5.2 and 5.3, respectively. The proof of Theorem 1.4 is given in Section 5.4 by comparing Fourier coefficients of the limit of 2-correlation functions. Section 6 describes the interpolation problem and the Riemann-Hilbert problem associated to discrete Gaussian orthogonal polynomials with the help of Legendre's relation, and includes proofs of previous Lemmas and Propositions. Appendix A constructs the steepest descent contours, and Appendix B provides the proof of Proposition 1.5. Reviewer's remarks. The paper is dry and technical. It would be nice to have more insights on how to reach the main results and how to construct the asymptotic limits. There are some typos and errors, for instance, formula (1) should have \(k=1, 2, \dots, n\). For the proof of Theorem 1.2, the key step in formula (184) is given without proof, and the authors simply state that it follows from the formulas (80) and (81), where both (80) and (81) lack proof or reasoning.