Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Girao 2024 monochromatic odd cycles edge coloured complete
lemma_2_1: If an n-vertex graph has no odd cycle of length at most 2k+1 for some k at least log_2 n, deleting at most (log_2 n / k) n vertices leaves a bipartite graph whose components have radius at most k.
lemma_2_2: A non-bipartite graph F has an odd cycle of length at most |V(F) minus V(H')| + (4r+1)m whenever H' has at most m components and lies in a subgraph H of F whose components have radius at most r.
lemma_2_3: Given q pairs of disjoint subsets of [n] with n at least 2^q/2, each pair covering at most a (1-epsilon) fraction of [n] with epsilon greater than 1/q, some set of at least 2^(epsilon q)/2 elements spans no pair a,b with a in A_i and b in B_i.
proposition_4_1: For q at least 1 and delta in (0,1), every q-coloring of the complete graph on (1+delta)2^q vertices has a monochromatic odd cycle of length O(q^2/delta).
theorem_1_2: A q-coloring of K_(2^q+1) forces a monochromatic odd cycle of length at most (2^q+1)/q^(1-epsilon) for every fixed positive epsilon and large q.
António Girão, Zach Hunter, Monochromatic odd cycles in edge-coloured complete graphs. Selected artifact: arXiv:2412.07708v1 (10 December 2024). The arXiv record names arXiv's non-exclusive distribution license (arXiv:2412.07708), every other right reserved.
Erdős and Graham asked for the growth of L(q), the least length such that every q-edge-coloring of the complete graph on 2^q+1 vertices contains a monochromatic odd cycle of length at most L(q); Day and Johnson had shown L(q) is unbounded, growing at least like 2 to the power of order square root of log q. Theorem 1.2 proves that for every epsilon > 0 and all large q, every q-coloring of that complete graph contains a monochromatic odd cycle of length at most (2^q+1)/q^{1-epsilon}, so L(q) = O(2^q/q^{1-o(1)}), to the authors' knowledge the first bound of the form o(2^q) (p. 1). The +1 belongs inside the numerator. The proof (Section 3, p. 3) combines three lemmas of Section 2 (pp. 2--3): Lemma 2.1 shows a graph with no short odd cycle can be made bipartite by deleting a small vertex set leaving components of bounded radius; Lemma 2.2 bounds the shortest odd cycle in a non-bipartite graph in terms of the components of a bounded-radius subgraph; and Lemma 2.3 uses a random choice of one side from each of q pairs of disjoint sets to extract a large set spanning no edge across any pair.
In the concluding remarks (Section 4, pp. 3--4) the paper writes for a bound on the shortest monochromatic odd cycle forced in a -coloring of , and Proposition 4.1 (p. 4) gives for and , a bound that at the exact host gives nothing better than the trivial length .
Read status. Claims checked for Theorem 1.2, Lemmas 2.1--2.3 and Proposition 4.1: their statements and hypotheses were read clause by clause on the page images. No proof is reconstructed or independently certified in this payload.
Source: https://arxiv.org/abs/2412.07708.
Bears on.
- #609: Theorem 1.2 is an upper bound, for each fixed and all large , at the problem's exact host ; it does not determine the growth of . Lemmas 2.1--2.3 enter only as ingredients of its proof. Proposition 4.1 concerns the larger hosts and at gives nothing better than the trivial bound .
Results to transcribe.
- Theorem 1.2 (p. 1): For every epsilon > 0 there is q0 such that every q-coloring of the complete graph on 2^q+1 vertices with q > q0 has a monochromatic odd cycle of length at most (2^q+1)/q^{1-epsilon}.
- Lemma 2.1 (p. 2): If has vertices and no odd cycle of length at most , for some , then some with leaves bipartite with every component of radius at most .
- Lemma 2.2 (p. 2): For a non-bipartite , a subgraph whose components have radius at most , and a subgraph of with at most components, has an odd cycle of length at most .
- Lemma 2.3 (p. 2): Let be an integer and let . Given pairs of disjoint subsets of with for each , there is a set of size at least that contains no together with a , for any .
- Proposition 4.1 (p. 4): For and , .
No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.