Non-intersecting squared Bessel paths and multiple orthogonal polynomials for modified Bessel weights (Q731298)

The aim of this paper is to consider a model of non-intersecting squared Bessel processes as \(n\to \infty\). The squared Bessel process \(\text{BESQ}^d\) is the process \(R^2(t)\), where \(R(t)=\|X(t)\|_2=\sqrt{X^2_1(t)+\dots+X_d^2}\), \(t\geq 0\), under the condition that \(\{X(t), t\geq0\}\) is a \(d\)-dimensional Brownian motion. This model is considered in the confluence case: all paths start at \(t=0\) in the same positive value \(x=a\), remain positive and end at time \(t=T\) in \(x=0\). Such a system of particles can be considered as a process of eigenvalues of a Hermitian matrix valued diffusion process, known as the Laguerre ensemble. Putting \(T=1\) and the time variable \(t\) depending on the number of paths, after some time rescaling so that \(0<t<1\), it is proved that the paths lie in the region of the \(tx\)-plane. This region is described explicitly (see Theorem 2.4). Initially, the paths stay away from the hard edge as \(x=0\), but for a certain critical time \(t^*\) the smallest paths hit the hard edge and from that moment they lie in it. Since, in the confluent case, the biorthogonal ensemble reduces to a multiple orthogonal polynomial ensemble and its correlation kernel is expressed via the \(3\times 3\) matrix valued Riemann-Hilbert problem, it is possible to consider the model in greater detail. The main results are stated in Section~2. Theorem 2.4 leads to an exact description of a simply connected region in the \(tx\)-plane for non-intersecting squared Bessel processes as \(n\to \infty\). The figures of numerical simulation of 50 rescaled non-intersecting \(\text{BESQ}^2\) with \(a=1\) and with \(a=5\) are presented as well. If \(t\neq t^*\), theorems 2.7., 2.8 and 2.9 give the usual scaling limits of the random matrix theory, as \(n\to\infty\), for the corresponding correlation kernel. Here the sine, the Airy and the Bessel kernels serve as limits. The case \(t=t^*\) is left for a separate publication. The proofs of the above mentioned theorems are given in Section~10. They use the steepest descent analysis of the Riemann-Hilbert problem considered in Sections~3-9. The steepest descent analysis itself takes most of the paper. Some statements here are of independent interest. To deal with an unbounded cut along the negative real axis, the construction of the global parametrix is applied in Section~6. The construction of the local parametrices at the hard edge O in Section~7 presents a new technical issue. In the Appendix an alternative approach is presented via equilibrium measures and associated \(g\)-functions. The Introduction of the paper contains the statement of the problem and its relations with previous publications in this field.