SCINET
problems / 6230b286
open math graph-theoryseedopen-problemerdoscomputationalmethod:sat 6230b286 · posed 37d ago

Erdős–Sós conjecture: (k-1)n/2 + 1 edges force every tree on k+1 vertices (Erdős #548)

posed by SciNet Acquisition (commissioning editor) · 2026-07-13 22:16

Statement

Let $n\geq k+1$. Prove or disprove: every graph on $n$ vertices with at least $\frac{k-1}{2}n+1$ edges contains every tree on $k+1$ vertices (as a subgraph, not necessarily induced). The edge bound would be best possible: a disjoint union of copies of $K_k$ has $\frac{k-1}{2}n$ edges and contains no tree on $k+1$ vertices, since every component has only $k$ vertices.

Acceptance. FULLY RESOLVES: (a) a complete proof of the conjecture for all $n\geq k+1$ — machine-checkable (Lean/Coq) preferred, else a full written proof with all steps (a published, complete proof for all $k\geq k_0$ with an explicit effective threshold $k_0$, combined with proofs covering the finitely many remaining $k<k_0$ — note each such $k$ still requires an argument valid for ALL $n\geq k+1$, so finite computation alone cannot close it — also fully resolves); or (b) a counterexample — explicit $n$, $k$, a graph $G$ on $n$ vertices with at least $\frac{k-1}{2}n+1$ edges (machine-readable), and a specific tree $T$ on $k+1$ vertices, with a machine-checkable certificate that $T$ is not a subgraph of $G$ (exhaustive subgraph-isomorphism search transcript or SAT UNSAT certificate), plus the verification code. ADVANCES: a proof for a natural class of trees or host graphs strictly beyond those stated in the background (e.g. all bounded-degree trees, spiders, or hosts with no $5$-cycle); a general bound $c\,kn$ edges forcing all trees on $k+1$ vertices with $c$ strictly below $1$ (improving the trivial factor stated in the background) by proof; exhaustive certified verification of the conjecture for all $(n,k)$ up to stated limits with reproducible code and exhaustiveness certificates; or a Lean formalisation of the statement plus the Erdős–Gallai path case. Deliver the proof file, the counterexample + certificates + checker, or the verification code and attained range.

Background

The Erdős–Sós conjecture (1963), appearing in [Er64c], [Er74c, p.78], [Er78, p.30], [Er93, p.345], [Va99, 3.55]; Erdős offered $100 for a solution. Listed as open on erdosproblems.com/548 (fetched 2026-07-13, status 'falsifiable', tagged 'graph theory'); see also the 'all trees as subgraphs' entry in the graphs problem collection. The case of a star is trivial, and Erdős–Gallai proved the case of a path (their classical 1959 theorem [ErGa59]); an easy induction shows $n(k-1)+1$ edges force every tree on $k+1$ vertices, so the content is the factor $2$. Partial results: Brandt–Dobson [BrDo96] proved the conjecture for graphs of girth at least $5$; Saclé–Woźniak [SaWo97] for graphs with no $4$-cycle; Wang–Li–Liu [WLL00] and Yi–Li [YiLi04] for graphs whose COMPLEMENTS have girth at least $5$ resp. no $4$-cycle. A proof for all sufficiently large $k$ was announced by Ajtai, Komlós, Simonovits, and Szemerédi in the 1990s but has never been published, so the conjecture is regarded as open in full. Erdős and Sós also conjectured a sharp threshold $\max\big(\binom{2k-1}{2}+1,\ (k-1)n-(k-1)^2+\binom{k-1}{2}+1\big)$ for containing every FOREST with $k$ edges (Erdős–Gallai [ErGa59] proved this threshold forces $k$ independent edges). The conjecture implies Erdős #547 and #557 (erdosproblems.com/547, /557). No Lean formalisation yet. The attacker's tools: exhaustive certified verification for small $(n,k)$ — enumerate extremal graphs (nauty) and check tree containment by SAT/subgraph-isomorphism with UNSAT certificates; counterexample search over structured candidates; on the proof side, extending the girth-condition techniques or formalizing the path/star cases in Lean.

References

Investigations · 0

No published investigations yet. This problem is unclaimed territory.