The extremal number of cycles with all diagonals

TL;DR

This paper determines the maximum number of edges in an n-vertex graph that avoids cycles with all diagonals, proving a tight bound of order n^{3/2} and introducing a new lemma on robust expanders.

Contribution

It resolves a longstanding problem by establishing a tight edge bound for graphs avoiding cycles with all diagonals and introduces a novel lemma on robust expanders.

Findings

Graphs with more than Cn^{3/2} edges contain a cycle with all diagonals.

The bound is tight due to constructions based on finite geometry avoiding four-cycles.

A new lemma on almost-spanning robust expanders is developed.

Abstract

In 1975, Erd\H{o}s asked the following natural question: What is the maximum number of edges that an $n$-vertex graph can have without containing a cycle with all diagonals? Erd\H{o}s observed that the upper bound $O(n^{5/3})$ holds since the complete bipartite graph $K_{3,3}$ can be viewed as a cycle of length six with all diagonals. In this paper, we resolve this old problem. We prove that there exists a constant $C$ such that every $n$-vertex with $Cn^{3/2}$ edges contains a cycle with all diagonals. Since any cycle with all diagonals contains cycles of length four, this bound is best possible using well-known constructions of graphs without a four-cycle based on finite geometry. Among other ideas, our proof involves a novel lemma about finding an `almost-spanning' robust expander which might be of independent interest.