Non-intersecting, simple, symmetric random walks and the extended Hahn kernel. (Q2574202): Difference between revisions

Fixing three positive integers \(a,b,c\) with \(c\geq b\), one may generate uniformly at random a rhombus tiling of an \(abc\)-hexagon, i.e. a hexagon with side lengths \(a,b,c,a,b,c\), by considering \(a\) simple, symmetric random walks on \(\mathbb{Z}\) with respective starting points \(0,2,\ldots,(2a-2)\) and endpoints \((c-b), (c-b+2),\ldots, (c-b+2a-2)\) at time \(b+c\), conditioned not to intersect during the whole (discrete) time interval \([0;b+c]\); such processes may also be used to generate other combinatorial objects of interest, e.g. boxed planar partitions. The main purpose of the present paper is to establish that this process is a determinantal point process, and to compute the corresponding correlation kernel in terms of associated Hahn polynomials. This is accomplished by using the general framework of \textit{K. Johansson} [Commun. Math. Phys. 242, 277--329 (2003; Zbl 1031.60084)] and a variant of the orthogonal polynomial method. First, a continuous scaling limit is considered, where \(a\) remains fixed whereas \(b=c\nearrow\infty\): one thus obtains non-intersecting Brownian motions all started at the origin and conditioned to end at the origin at time \(T\), a model which can also be viewed as a transformation of Dyson's Hermitian Brownian motion, and the author indicates how to compute the correlation kernel in such a context, using Hermite polynomials and the orthogonal polynomial method. The discrete model is then treated in a similar way, using associated Hahn polynomials and a suitable modification of the orthogonal polynomial method. Finally, some remarks and a conjecture are made regarding the asymptotics of the obtained kernel and the continuous scaling limit where \(a=b=c\nearrow\infty\).