Tightness of cᵣ upper bound for r ≥ 3

Conjecture on tightness of the upper bound for all r ≥ 3 · arXiv:2105.15195

arXiv Informal medium confidence— first stated 2022-09-22

Authors: David Conlon, Jacob Fox, Huy Tuan Pham

Status open high confidence

Conlon, Fox, and Pham prove in Theorem 1 that $c_r \leq \left(1-\frac{1}{2b_0}\right)\left(1+\frac{1}{2^r b_0 - r}\right)$ for all $r \geq 2$ and establish tightness for $r=2$ (giving $c_2=(2+\sqrt{3})/4$). They conjecture the same formula gives the exact value of $c_r$ for all $r \geq 3$, but acknowledge their methods are insufficient. No follow-up paper resolving or substantially advancing this conjecture for $r \geq 3$ was found in an exhaustive web search covering arXiv through May 2026.

Reviewer notes. No follow-up found. The conjecture is open for all r >= 3. The paper was published in Mathematika (2022). Huy Tuan Pham's publication page and Semantic Scholar citation search returned no citing papers that address the r >= 3 case. The companion paper arXiv:2104.14766 ('Subset sums, completeness and colorings') develops the general techniques but does not resolve the tightness conjecture for r >= 3 either.

Auto-reviewed 2026-05-15 with claude-sonnet-4-6 (web search enabled).

Informal. We suspect that our upper bound $c_r \leq \left(1-\frac{1}{2b_0}\right)\left(1+\frac{1}{2^r b_0 - r}\right)$ is also tight for all $r \geq 3$, where $b_0 > 1$ is the unique root of $b^r - 2rb + r - 1 = 0$.

Context

Theorem 1 gives a general upper bound for $c_r$ for all $r\geq 2$ and proves it is achieved (i.e., tight) for $r=2$, yielding $c_2=(2+\sqrt{3})/4$. The authors believe the same formula gives the exact value for all larger $r$, but remark that their methods are insufficient to prove this, directing the reader to brief concluding remarks for further discussion.

Notes. Stated in prose without a labelled environment: 'We suspect that our upper bound is also tight for all r ≥ 3, but our methods do not seem sufficient for proving this.' PDF extraction may have garbled exponent notation (2^r vs 2r) in the formula.

Source paper

The upper logarithmic density of monochromatic subset sums
David Conlon, Jacob Fox, Huy Tuan Pham · 2022-09-22
https://arxiv.org/abs/2105.15195 PDF source