Components, large and small, are as they should be I: supercritical percolation on regular graphs of growing degree

TL;DR

This paper establishes conditions under which supercritical percolation on regular graphs with growing degree exhibits a phase transition similar to Erdős–Rényi graphs, unifying several well-known graph models.

Contribution

It provides a unified framework with sufficient conditions for phase transition in supercritical percolation on regular graphs of growing degree, covering multiple classical graph families.

Findings

Presence of a giant component of size proportional to $y( ext{epsilon})n$

Other components are typically of size $O(rac{ ext{log} n}{ ext{epsilon}^2})$

Results are tight under weaker expansion conditions

Abstract

We provide sufficient conditions for a regular graph $G$ of growing degree $d$, guaranteeing a phase transition in its random subgraph $G_p$ similar to that of $G(n,p)$ when $p\cdot d\approx 1$. These conditions capture several well-studied graphs, such as (percolation on) the complete graph $K_n$, the binary hypercube $Q^d$, $d$-regular expanders, and random $d$-regular graphs. In particular, this serves as a unified proof for these (and other) cases. Suppose that $G$ is a $d$-regular graph on $n$ vertices, with $d=\omega(1)$. Let $\epsilon>0$ be a small constant, and let $p=\frac{1+\epsilon}{d}$. Let $y(\epsilon)$ be the survival probability of a Galton-Watson tree with offspring distribution Po$(1+\epsilon)$. We show that if $G$ satisfies a (very) mild edge expansion requirement, and if one has fairly good control on the expansion of small sets in $G$, then typically the percolated random subgraph $G_p$ contains a unique giant component of asymptotic order $y(\epsilon)n$, and all the other components in $G_p$ are of order $O(\log n/\epsilon^2)$. We also show that this result is tight, in the sense that if one asks for a slightly weaker control on the expansion of small sets in $G$, then there are $d$-regular graphs $G$ on $n$ vertices, where typically the second largest component is of order $\Omega(d\log (n/d))=\omega(\log n)$. This is the first of a two-part sequence of papers. In the subsequent work, we consider supercritical percolation on regular graphs of constant degree, and establish similar sufficient (and essentially tight) conditions in that setting.