On the diameter of random uniform hypergraphs in dense regime
Kartick Adhikari, Asrafunnesa Khatun
TL;DR
This work extends the classical diameter results for Erdős–Rényi graphs to dense random t-uniform hypergraphs, showing a sharp two-point concentration of the diameter at d and d+1 under a specific scaling of p with n. The authors deploy the Stein-Chen Poisson-approximation method, coupled with Harris–FKG inequalities and exchangeable sequences, to control the count of remote vertex pairs whose distance exceeds d. Central to the argument are precise asymptotics for the single and paired remote events, yielding a Poisson limit for the number of remote pairs and thereby a two-point diameter distribution in the dense regime. The approach also provides a versatile framework that could tackle diameter-type questions in other complex networks, including higher-order structures like Linial–Meshulam-type complexes.
Abstract
For a fixed natural number $t\geq 2$, we consider $t$-uniform random hypergraphs $\mathscr{H} (n,t,p)$ on $n$ vertices $[n]=\{1,\ldots, n\}$, where each $t$-subset of $[n]$ is included as a hyperedge with probability $p$ and independently. We show that the diameter of $\mathscr{H} (n,t,p)$ is concentrated only at two points in the dense regime. More precisely, suppose $diam(\mathcal H)$ denotes the diameter of a hypergraph $\mathcal H$ on $n$ vertices. We show that, for fixed $t,c,d$ constants, if $n$ and $p$ (depends on $t,c,d,n$) satisfy $$ \frac{ (t-1)^ {d} N^{d} p^{d}} {n}= \log \left( \frac{n^2}{c} \right), \mbox{ where } N={n-1\choose t-1}, $$ $c$ is a positive constant and $d\geq2$ is a natural number, then $$ \lim_{n \to \infty} \mathbb{P} \left( diam( \mathcal{H}) = d \right) = e^{- \frac{c}{2}} \text{ and } \lim_{n \to \infty} \mathbb{P} \left( diam(\mathcal{H}) = d+1 \right) = 1- e^{- \frac{c}{2}}. $$ In particular, the case where $t = 2$ corresponds to the diameter of the Erdős-Rényi graph, as established by Bollobás in \cite[Theorem~6]{bollobas1981diameter}. Bollob\' as's result was proven using the moments method, which is challenging to apply in our context due to the complexity of the model. In this paper, we utilize the Stein-Chen method along with coupling techniques to prove our result. This approach can potentially be used to solve various problems, in particular diameter problems, in more complex networks.