Sharp Threshold for Cliques in Random 0/1 Polytope Graphs

TL;DR

This paper investigates the threshold phenomena for clique formation and expansion properties in graphs derived from random 0/1 polytopes, providing new thresholds and combinatorial insights.

Contribution

It establishes precise thresholds for edge density, expansion, and clique formation in random 0/1 polytope graphs, resolving open questions and strengthening previous results.

Findings

Threshold for edge density at p=2^{-n/2}

Strong edge expansion for p ≤ 2^{-n/2 - o(1)}

Threshold for clique formation at p≈2^{- ext{constant} imes n}

Abstract

We study graph-theoretic properties of random $0/1$ polytopes. Specifically, let $Q_p^n \subseteq \{0,1\}^n$ be a random subset where each point is included independently with probability $p$, and consider the graph $G_p$ of the polytope conv$(Q_p^n)$. We provide a short and combinatorial proof that $p = 2^{-n/2}$ is a threshold for the edge density of $G_p$, a result originally due to Kaibel and Remshagen. We next resolve an open question from their paper by showing that for $p \leq 2^{-n/2 - o(1)}$, $G_p$ exhibits strong edge expansion. In particular, we prove that, with high probability, every vertex has degree $(1 - o(1))|Q_p^n|$. Lastly, we determine the threshold for $G_p$ being a clique, strengthening a result of Bondarenko and Brodskiy. We show that with high probability, if $p \geq 2^{-\delta n + o(1)}$, then $G_p$ is not a clique, and if $ p \leq 2^{-\delta n - o(1)}$, then $G_p$ is a clique, where $\delta \approx 0.8295$. Our approach combines a combinatorial characterization of edges in graphs arising from polytopes with the Kim-Vu polynomial concentration inequality.