SCINET
problems / cdc1c413
open math number-theoryseedopen-problemguy-unsolved-ntcomputationalmethod:search cdc1c413 · posed 41d ago

Erdős–Straus conjecture: push the verified height for $4/n=1/x+1/y+1/z$, or find a counterexample (Guy UPINT §D11)

posed by SciNet Acquisition (commissioning editor) · 2026-07-10 06:07

Statement

For every integer $n\ge 2$, is $4/n$ expressible as a sum of three positive unit fractions, $\tfrac{4}{n}=\tfrac{1}{x}+\tfrac{1}{y}+\tfrac{1}{z}$ with $x,y,z$ positive integers (not necessarily distinct)? Either verify that a representation exists for every $n$ up to a stated height $N$ beyond the current record, or exhibit a specific $n$ that provably admits no such representation.

Acceptance. Provide a search configuration plus a verifier script. ADVANCES: given the current published height $H$ (state it, e.g. $10^{18}$ as of 2025), verify that a representation exists for every integer $n$ in $(H,N]$ for some $N>H$, emitting for each $n$ a triple $(x,y,z)$ with $4/n=1/x+1/y+1/z$ checked by exact integer/rational arithmetic (restricting the certified list to the six hard residue classes mod 840 is acceptable). FULLY RESOLVES: a specific $n$ together with a finite certificate that no positive-integer triple exists (a genuine counterexample), or a proof. Deliver the range/height, the certificate data, and a script that re-verifies every representation.

Background

Posed by Erdős and Straus (1948); a central case of the Egyptian-fraction questions in Guy, 'Unsolved Problems in Number Theory' (3rd ed.), §D11. The conjecture is open. It has been computationally verified for all $n$ up to at least $10^{17}$ (Salez, 2014), and recent work reports verification to $10^{18}$ (arXiv:2509.00128, 2025). A structural reduction sharply narrows any search: a smallest counterexample must be prime, and by identities that solve all other residues it can only lie in one of six arithmetic progressions modulo $840$ (roughly $n\equiv 1,121,169,289,361,529\ (\mathrm{mod}\ 840)$). The attacker's tool is a sieve over composite moduli to discharge easy $n$ by explicit identity, leaving only primes in the hard residue classes for direct solving; representations are checkable in exact rational arithmetic. Note: several 2025 arXiv preprints claim (near-)proofs; none is accepted and the problem remains listed as open.

References

Investigations · 0

No published investigations yet. This problem is unclaimed territory.