Explosion in weighted hyperbolic random graphs and geometric inhomogeneous random graphs (Q2301480)

This paper derives distributional results on weighted distances in the so-called geometric inhomogeneous random graphs. This is a model that was introduced recently by \textit{K. Bringmann} et al. [``Average distance in a general class of scale-free networks with underlying geometry'', Preprint, \url{arXiv:1602.05712}]. There, a random graph is formed on a set of \(n\) points that are distributed according to some measure on a metric space. Furthermore, these points have weights which i.i.d. random variables. The probability that two of them are joined by an edge is a function of their distance in the underlying space as well as of their weights. In particular, it is proportional to the product of their weights. Also, the collection of the corresponding events for all pairs of point forms an independent family. In this paper, the weights are assumed to follow a power law distribution. In the usual set-up of first passage percolation, the edges are assumed to have i.i.d. weights sampled from some distribution \(L\) and the weighted distances (or \(L\)-distances) between vertices are considered. Specifically, the random variable that is considered is the explosion time. Roughly speaking, this is the smallest distance \(t\) around a certain vertex where the ball of weighted distance at most \(t\) becomes infinite. The authors show that if the vertex weights have finite second moment, then explosion does not occur. If the power law has exponent between 2 and 3 (that is, infinite second moment), they give a condition on \(L\) which characterises the occurrence of explosion. Furthermore, if explosion occurs, they show that the \(L\)-distance of two randomly selected vertices converges in distribution to the sum of two i.i.d. random variables distributed as the explosion time around a random vertex conditioned on being finite. Analogous results are shown in the context of scale-free percolation on \(\mathbb{Z}^d\) in which the underlying graph is formed by joining non-nearest neighbours of \(\mathbb{Z}^d\) with probability that depends on their distance in \(\mathbb{Z}^d\) and the product of their weights. Again, the vertices are assumed to have i.i.d. weight that follow a power law and edges have i.i.d. weights sampled from a distribution \(L\). A distributional result as above is derived when the condition that yields explosion holds.