Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Claim. Corollary IV of M. A. Berger, A. Felzenbaum and A. Fraenkel, The Herzog-Schönheim conjecture for finite nilpotent groups, Canad. Math. Bull. 29 (1986), no. 3, 329--333, states: "Any coset partition of a finite nilpotent group into at least two cosets must contain two cosets of the same order." The proof writes the group as the direct product of its Sylow subgroups, so that every coset becomes a product set whose sides have prime-power sizes, and applies the paper's Theorem III: a box of this kind partitioned into two or more such product sets has two parts of equal size. The statements and the proof outline are recorded on the library's source card.

Covers. The case of Problem 274 for finite nilpotent groups, and so for finite abelian groups: no such group has an exact covering by two or more cosets of pairwise different sizes. This answers no to Erdős's abelian question of [Er77c] and [ErGr80] for finite groups. The question for nonnilpotent groups stays open.

Depends on. Nothing in this wiki; the theorem is the paper's own.

Acceptance. Refereed: Canadian Mathematical Bulletin 29 (1986), no. 3, 329--333, doi:10.4153/CMB-1986-050-0, issued 1 September 1986. Not reviewed: the site's commentary does not mention the paper, and the site labels the problem OPEN. Not formalized: see below.

Formalization. The Lean 4 repository Jostamon/erdos274-hs-abelian, linked above at its commit of 2026-07-10, proves the formal-conjectures statement erdos_274.variants.abelian, the finite abelian case, and its README says that the proof follows a simplified form of this paper's argument; it is therefore recorded here as a formalization of the abelian case of this claimant's result. Its files carry Murali Menon's copyright. The formal-conjectures file tags that variant research solved with a formal_proof link to the repository at the same commit, added by pull request 4415 (merged 2026-07-21). This corpus has neither built nor audited the repository, so it gives no formalized evidence, and it does not cover the nilpotent case.