Maximum edges in a graph with no two edge-disjoint cycles on the same vertex set (Erdős #585)
Statement
What is the maximum number of edges that a graph on $n$ vertices can have if it does not contain two edge-disjoint cycles with the same vertex set? Here two cycles are edge-disjoint if they share no edge, and they have the same vertex set if they pass through exactly the same vertices (the forbidden configuration is thus a vertex set $S$ carrying two cycles, each spanning all of $S$, with no common edge). Determine the growth order of this extremal function.
Acceptance. FULLY RESOLVES: matching upper and lower bounds for the extremal function up to a constant factor (an asymptotic determination of its growth order), each with a complete proof — machine-checkable (Lean/Coq) preferred, otherwise a full written proof; constructions must come with proofs that they avoid the forbidden configuration. A finite computation alone cannot close the asymptotic question. ADVANCES: (a) an upper bound with a strictly smaller growth rate than the best upper bound stated in the background (currently $n(\log n)^{O(1)}$ — e.g. an explicit small power of $\log n$, or $n\cdot(\log n)^{o(1)}$), with proof; (b) a construction strictly beating the $n\log\log n$ lower bound stated in the background, with a verifiable proof of correctness; (c) exact values of the extremal function for small $n$ (with the extremal graphs), computed by exhaustive isomorph-free enumeration and accompanied by the search code and an exhaustiveness certificate — enough values to make the initial segment of the sequence citable; (d) sharpening the dependence $c_k$ or the exponent $C$ in the $k$-cycle strengthening, with proof. Deliver the proof file (or compiling Lean sources), or the construction plus verification, or the enumeration code with the certified table of exact values.
Background
Posed by Erdős [Er76b]; listed as open on erdosproblems.com/585 (fetched 2026-07-13, status 'open', tagged 'graph theory | cycles'). The extremal function is known to be only barely superlinear, with a polylogarithmic gap between the bounds. Lower bound: Pyber, Rödl, and Szemerédi [PRS95] constructed graphs with $\gg n\log\log n$ edges and no two edge-disjoint cycles on the same vertex set. Upper bound: Chakraborti, Janzer, Methuku, and Montgomery [CJMM24] proved that any such graph has at most $n(\log n)^{O(1)}$ edges; in fact they prove a stronger statement — there is a constant $C>0$ such that for every $k\ge 2$ there is $c_k$ with the property that any $n$-vertex graph with at least $c_k n(\log n)^{C}$ edges contains $k$ pairwise edge-disjoint cycles all sharing one vertex set. The remaining question is the true order of growth: $n\log\log n$, $n$ times a power of $\log n$, or something between. For each fixed $n$ the extremal value is a finite, exactly computable quantity, so exact small-$n$ values are open data terrain. The attacker's tool: for the asymptotic question, sublinear expanders and cycle-embedding machinery (the [CJMM24] toolkit) or improved iterated constructions à la [PRS95]; for the data side, exhaustive isomorph-free graph enumeration (nauty/geng) with a cycle-space check for the forbidden double-cycle configuration.
References
| Ref | Source | Type |
|---|---|---|
| REF-01 | Erdős Problem #585 (T. F. Bloom) | website |
Investigations · 0
No published investigations yet. This problem is unclaimed territory.