Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Bennett 2022 erdos gyarfas function so gyarfas was
Patrick Bennett, Ryan Cushman, Andrzej Dudek, Paweł Prałat, The Erdős-Gyárfás function -- so Gyárfás was right. arXiv:2207.02920 (2022); published in J. Combin. Theory Ser. B 169 (2024), 253--297, DOI 10.1016/j.jctb.2024.07.001 (Crossref record read 2026-10-07). The arXiv record (https://arxiv.org/abs/2207.02920, read 2026-10-02) names the Creative Commons Attribution 4.0 license.
A (4,5)-coloring of K_n is an edge-coloring in which every 4-clique receives at least five distinct colors, and the Erdos-Gyarfas function f(n,4,5) is the least number of colors admitting one. The paper shows there exist (4,5)-colorings of K_n with (5/6)n + o(n) colors, matching the known lower bound and so settling f(n,4,5) = (5/6)n + o(n). This resolves a disagreement between Erdos and Gyarfas recorded in their 1997 paper on f(n,p,q), where Erdos expected the coefficient of n to be 1 and Gyarfas expected it to be closer to 5/6. The construction is a two-phase randomized process: phase one colors almost all edges via a random triangle-removal-style process (following Bollobas-Erdos and Bohman-Frieze-Lubetzky) analyzed by the differential equation method to get dynamic concentration, using (5/6)n + epsilon n/2 colors; phase two finishes with a fresh epsilon n/2 colors using the Lovasz Local Lemma. Chernoff and Freedman inequalities (Lemmas 1 and 2) supply the concentration tools. This settles problem 136 on the growth of f(n,4,5).
Source: https://arxiv.org/abs/2207.02920. The held PDF is arXiv:2207.02920v1 (6 July 2022, 35 pages), the only arXiv version; the labels and pages below are the preprint's, and the journal version was not compared with it.
Bears on. #136
Results to transcribe.
- Theorem 1 (p. 2): f(n,4,5) = (5/6)n + o(n). The new part is the upper bound, (4,5)-colorings of K_n with (5/6)n + o(n) colors; the lower bound f(n,4,5) >= (5/6)(n-1) was proved by Erdos and Gyarfas, after Erdos, Elekes and Furedi had stated it (p. 2), and is restated with their proof as Theorem 2 (p. 4).
- Two-phase process: Phase one uses a randomized triangle-removal coloring analyzed by the differential equation method; phase two (Section 12) completes the coloring via the Lovász Local Lemma with a fresh set of epsilon n/2 colors, for an arbitrary fixed epsilon > 0.
- Context (Erdős-Gyárfás thresholds): Recalls f(n,p,2) between Omega(log n/log log n) and O(log n), f(n,4,3) = n^{o(1)}, f(n,4,4) = n^{1/2+o(1)}, and f(n,p,p) >= n^{1/(p-2)}.