Erdős Problem #986 — For any fixed k≥3k\geq 3, R(k,n)≫nk−1(log⁡n)cR(k,n) \gg \frac{n^{k-1}}{(\log n)^c} for some constant c=c(k)>0c=c(k)>0.

About 79 years old · traced to

For any fixed k≥3k\geq 3, R(k,n)≫nk−1(log⁡n)cR(k,n) \gg \frac{n^{k-1}}{(\log n)^c} for some constant c=c(k)>0c=c(k)>0.

References

Progress summary

Refreshed
Claimed solved

A 2026 preprint claims the conjectured lower bound in every fixed dimension, but the result has not been independently verified.

The conjecture asks whether, for every fixed k≥3k\geq 3, the off-diagonal Ramsey number satisfies the stated near-nk−1n^{k-1} lower bound up to a logarithmic factor. It is attributed to Erdős, apparently from 1947.

Known results

  • Spencer (1977) settled k=3k=3.
  • Mattheus and Verstraete (2023) proved R(4,n)=n3+o(1)R(4,n)=n^{3+o(1)}.
  • Ajtai, Komlós, and Szemerédi proved the corresponding upper bound R(k,n)≪knk−1/(log⁡n)k−2R(k,n)\ll_k n^{k-1}/(\log n)^{k-2}.
  • Earlier general lower bounds were weaker for k≥5k\geq 5.

May 2026 claimed theorem

Bradač’s preprint claims that, for every fixed k≥3k\geq 3, R(k,n)≫nk−1/(log⁡n)2k−4R(k,n)\gg n^{k-1}/(\log n)^{2k-4}, which implies the conjecture. The preprint reports assistance from an unnamed internal OpenAI model; the mathematical claim remains unverified.

Current status (as of September 2026): Bradač’s preprint claims the bound for every fixed k≥3k\geq 3, but independent verification is still absent.

Sources

Solutions 0

No solutions have been posted yet.