Non-intersecting squared Bessel paths: Critical time and double scaling limit (Q647382)

The paper deals with the squared Bessel process with parameter \(\alpha > -1\) and transition density \[ p^{\alpha}_t(x,y)=\frac{1}{2t}\left(\frac{y}{x}\right)^{\frac{\alpha}{2}}e^{-\frac{x+y}{2t}}\mathbf{1}_{\alpha}\left(\frac{\sqrt{xy}}{t}\right), \quad x>0, \quad y\geq0, \] \[ p^{\alpha}_t(0,y)=\frac{y^{\alpha}}{(2t)^{\alpha+1}\Gamma(\alpha+1)}e^{-\frac{y}{2t}}, \quad y>0, \] where \(\mathbf{1}_{\alpha}\) denotes the modified Bessel function of the first kind of order \(\alpha\). The double scaling limit is considered for a model of \(n\) non-intersecting squared Bessel processes in the confluent case: all paths start at time \(t=0\) at the positive value \(x=a\), remain positive, and are conditioned to end at time \(t=1\) at \(x=0\). After certain rescaling, the paths fill a region in the \(tx\)-plane as \(n\rightarrow\infty\), that intersects the hard edge at \(x=0\) at a critical time \(t=t^*\). The critical time \(t=t^*\) is of interest in this paper. It is considered the limit as \(n\rightarrow\infty\) of the correlation kernel at critical time \(t^*\) and in the double scaling regime. The transition form the Airy kernel to the Bessel kernel is described by a new one-parameter family of limiting kernels that arise as limiting kernels around the critical time. The time rescaling \(t\mapsto\frac{t}{2n}\); \(T\mapsto\frac{1}{2n}\) is introduced and analysis of the rescaled kernel \(K_n(x,y,t)\) is based on the study of \(3\times3\) matrix valued Riemann-Hilbert problem by the Deift-Zhou steepest descent method. This RH problem has a unique solution in terms of multiple orthogonal polynomials for the modified Bessel weights. In this rescaling, the critical time is \(t^*=\frac{a}{a+1}\). Theorem 1.2 is the main result of the paper. Let \(K_n\) be the correlation kernel for the positions of the rescaled non-intersecting squared Bessel paths with parameter \(\alpha>-1\), starting at \(a>0\) at time \(t=0\) and ending at zero at time 1. Let \(t^*=\frac{a}{a+1}\); \(c^*=\frac{a}{(a+1)^2}\). Then, for every fixed \(\tau\in\mathbb{R}\) and \(x,y>0\), \[ \lim\limits_{n\rightarrow\infty}\frac{c^*}{n^{\frac{3}{2}}}K_n\left(\frac{c^* x}{n^{\frac{3}{2}}},\frac{c^* y}{n^{\frac{3}{2}}}; t^*-\frac{c^*\tau}{\sqrt{n}}\right)=K^{crit}_{\alpha}(x,y;\tau), \] where the critical kernel is defined by a certain contour integral with two certain loops as contours of integration. In proving Theorem 1.2, the steepest descent analysis needs five transformations of the RH problem that lead to an RH problem for the result of the final transformation, normalized at infinity and with jump matrices that are close to the identity matrix if \(n\) is large. This is done in Sections 2, 3, 6 and 7. At a certain stage of the analysis a local parametrix at the origin is needed. In the critical case, considered here, coalescing of the soft edge with the hard edge leads to the construction of a new local parametrix. The global and local parametrices are considered in Section 4 and 5. Section 8 presents an expression for the critical kernel and the proof of Theorem 1.2.