Coloring random subgraphs

Status partial high confidence

Bukh's 2008 conjecture that $\mathbb{E}(\chi(G_{1/2})) \geq c\,\chi(G)/\log\chi(G)$ for all graphs $G$ remains open. Partial results confirm it for graphs with small independence number $\alpha(G) \leq O(n/k)$ (Shinkar 2016) and for the fractional chromatic number analogue (Mohar–Wu 2018), and a 2023 paper by the original proposer with collaborators establishes tail-probability bounds for $\chi(G_{1/2})$ and studies the degeneracy of $G_{1/2}$, but the conjecture in its original integral form is unresolved.

Cited literature (5)

  • Igor Shinkar · arXiv preprint · arXiv:1612.04319

    Proves Bukh's conjecture for the special case of graphs with $\chi(G)=k$ and independence number $\alpha(G) \leq O(n/k)$, and gives exponential tail bounds showing $\Pr[\chi(G_{1/2}) \leq d] \leq \exp(-\Omega(k(k-d^3)/d^3))$ for all $d \leq k^{1/3}$.

  • Bojan Mohar, Hehui Wu · Journal of Graph Theory · arXiv:1807.06285

    Proves the fractional-chromatic-number analogue of Bukh's conjecture: for every graph with fractional chromatic number $\geq n$, a random subgraph $G_{1/2}$ has fractional chromatic number $\geq n/(8\log_2(4n))$ with probability $> 1 - 1/(2n)$.

  • Ross Berkowitz, Pat Devlin, Catherine Lee, Henry Reichard, David Townley · arXiv preprint · arXiv:1811.02018

    Establishes a new spectral lower bound on $\mathbb{E}[\chi(G_p)]$ as progress toward Bukh's conjecture, and confirms the conjecture for planar graphs and graphs with chromatic number less than 4.

  • Boris Bukh, Michael Krivelevich, Bhargav Narayanan · Combinatorics, Probability and Computing · arXiv:2312.08340 · doi:10.1017/S0963548325000069

    Proves tail bounds for $\chi(G_{1/2})$ across several regimes, and shows there exist infinitely many $k$-regular graphs whose degeneracy (colouring number) of $G_{1/2}$ is at most $k/3+o(k)$ w.h.p.; explicitly lists the conjecture $\mathbb{E}[\chi(G_{1/2})] = \Omega(k/\log k)$ as an open problem (Problem 1.1).

  • Bojan Mohar, Hehui Wu · arXiv preprint · arXiv:1807.06285

    Proves an affirmative answer to a strengthening of Bukh's question for the fractional chromatic number: for every graph $G$ with $\chi_f(G) \geq t$, the fractional chromatic number of $G_p$ is at least $t/(8\log_{1/p}(et)+4)$ with probability at least $1-1/(2t)$.

Reviewer notes. The Wiley journal page for Mohar–Wu (doi:10.1002/jgt.22571) returned HTTP 403; the DOI is inferred from the Wiley URL appearing in search results but was not directly fetch-verified, so it is left null. The 2023/2025 Bukh–Krivelevich–Narayanan paper was verified via both the arXiv abstract page and the Cambridge Core journal page (published CPC vol. 34 iss. 4, 2025). No resolution of the original integer chromatic number conjecture was found through 2026.

Auto-reviewed 2026-05-08 with claude-sonnet-4-6 (web search enabled) · 142s.

Problem. Does there exist a constant $ c $ so that $ {\mathbb E}(\chi(G_{1/2})) > c \frac{\chi(G)}{\log \chi(G)} $ ?
Keywords: coloring · random graph

Discussion

It is a classical result that the above problem has a positive answer when $ G $ is the complete graph. More generally, the lower bound $ {\mathbb E}(\chi(G_{1/2})) \ge c \frac{\chi(G)}{\log |V(G)|} $ is known. It is easy to obtain the bound $ {\mathbb E}(\chi(G_{1/2})) \ge (\chi(G))^{1/2} $ , since we may imagine forming two random subgraphs $ H,H' $ of $ G $ by putting each edge of $ G $ in either $ H $ or $ H' $ independently with probability $ 1/2 $ . Then $ \chi(H) \chi(H') \ge \chi(G) $ and this gives the desired bound. A similar argument with three subgraphs shows that $ {\mathbb E}(\chi(G_{1/3})) \ge (\chi(G))^{1/3} $ , however these arguments all seem to require integer multiples, so the best known lower bound on $ {\mathbb E}(\chi(G_{49/100})) $ of this form is $ (\chi(G))^{1/3} $ .

Bibliography