Dual View Random Solved Random Open
PROVED (LEAN) This has been solved in the affirmative and the proof verified in Lean.
Does every finite colouring of the integers have a monochromatic solution to $1=\sum \frac{1}{n_i}$ with $2\leq n_1<\cdots <n_k$?
The answer is yes, as proved by Croot [Cr03] - indeed, there are infinitely many disjoint such monochromatic solutions.

In [ErGr80] they also ask for a monochromatic representation of any $\frac{a}{b}>0$. This follows from the case of $1$ - indeed, consider the induced colouring of $\{\tfrac{n}{b}: b\mid n\}$. By the above there are $a$ solutions to\[1=\sum_i \frac{1}{n_i/b},\]and hence $a$ solutions to $\frac{1}{b}=\sum_i \frac{1}{n_i}$, where all $n_i$ are distinct (across the $a$ many solutions). Summing across all variables then yields $\frac{a}{b}=\sum_j \frac{1}{m_j}$ where all $m_j$ are distinct and the same colour, as required.

See also [298].
Additional thanks to: Euro Vidal Sampaio and Desmond Weisenberg
Proof expositions (0)
If you would like to contribute an exposition of a proof related to this problem, please message a moderator or leave your exposition as a comment.

No proof expositions yet.
Comments (1) Proof claims (0)
More information and links
This page was last edited 07 April 2026. (View history) (View the LaTeX source)

When referring to this problem, please use the original sources of Erdős. If you wish to acknowledge this website, the recommended citation format is:

T. F. Bloom, Erdős Problem #46, https://www.erdosproblems.com/46, accessed 2026-09-01

From the external database. (You can help update this.)
Formalised statement? Yes
Reactions
Likes holyterror
Open to collaboration None
Currently working on None
Looks difficult None
Looks tractable None
Could be formalisable None
Working on formalising None