Erdős Problem #306 — Let a/b∈Q>0a/b\in \mathbb{Q}_{>0} with bb squarefree.

About 46 years old · traced to

Let a/b∈Q>0a/b\in \mathbb{Q}_{>0} with bb squarefree. Are there integers 1<n1<⋯<nk1<n_1<\cdots<n_k, each the product of two distinct primes, such that ab=1n1+⋯+1nk?\frac{a}{b}=\frac{1}{n_1}+\cdots+\frac{1}{n_k}?

References

Progress summary

Refreshed
Claimed progress

A 2026 paper settles integers and sufficiently large fractions, while the full question remains open; an unverified repository claim says otherwise.

Erdős Problem #306 asks whether every positive rational with squarefree denominator can be expressed as a finite sum of distinct reciprocals of semiprimes. It is the Erdős–Graham problem, also listed as problem D11 in Guy’s Unsolved Problems in Number Theory.

Known results

  • Butler, Erdős and Graham (2015): the analogous statement with denominators having three distinct prime factors holds for positive integers.
  • Johnson (1978): 11 has a representation using 4848 semiprime reciprocals.
  • A later computation found 1717 representations of 11 using 4747 terms, but no shorter example was established.

June 2026 partial proof and conflicting claim

A 2026 paper proves the integer case and proves representability for every squarefree-denominator rational above an explicit threshold; the remaining racabrac{a}{b} below that threshold is reduced to a gap-free-floor conjecture and remains open. A repository claims a theorem proving the full statement, but provides no independently verified proof and conflicts with the paper and a formalization retaining sorry.

Current status (as of June 2026): The integer case and all sufficiently large racabrac{a}{b} are settled, while the small-rational racabrac{a}{b} case remains open; a purported full proof is unverified.

Sources

Solutions 0

No solutions have been posted yet.