Connecting the Random Connection Model

Abstract

Consider the random graph $G({\mathcal P}_{n},r)$ whose vertex set ${\mathcal P}_{n}$ is a Poisson point process of intensity $n$ on $(- \frac{1}{2}, \frac{1}{2}]^d$, $d \geq 2$. Any two vertices $X_i,X_j \in {\mathcal P}_{n}$ are connected by an edge with probability $g\left( \frac{d(X_i,X_j)}{r} \right)$, independently of all other edges, and independent of the other points of ${\mathcal P}_{n}$. $d$ is the toroidal metric, $r > 0$ and $g:[0,\infty) \to [0,1]$ is non-increasing and $\alpha = \int_{\mathbb{R}^d} g(|x|) dx < \infty$. Under suitable conditions on $g$, almost surely, the critical parameter $d_n$ for which $G({\mathcal P}_{n}, \cdot)$ does not have any isolated nodes satisfies $\lim_{n \to \infty} \frac{\alpha n d_n^d}{\log n} = 1$. Let $\beta = \inf\{x > 0: x g\left( \frac{\alpha}{x \theta} \right) > 1 \}$, and $\theta$ be the volume of the unit ball in $\mathbb{R}^d$. Then for all $\gamma > \beta$, $G\left({\mathcal P}_{n}, \left( \frac{\gamma \log n}{\alpha n} \right)^{\frac{1}{d}}\right)$ is connected with probability approaching one as $n \to \infty$. The bound can be seen to be tight for the usual random geometric graph obtained by setting $g = 1_{[0,1]}$. We also prove some useful results on the asymptotic behaviour of the length of the edges and the degree distribution in the {\it connectivity regime}.