Erdős Problem #997 — Call x1,x2,…∈(0,1)x_1,x_2,\ldots \in (0,1) well-distributed if, for every ϵ>0\epsilon>0, if kk is sufficiently large then, for all n>0n>0 and intervals I⊆[0,1]I\subseteq [0,1],…

About 62 years old · traced to

Call x1,x2,…∈(0,1)x_1,x_2,\ldots \in (0,1) well-distributed if, for every ϵ>0\epsilon>0, if kk is sufficiently large then, for all n>0n>0 and intervals I⊆[0,1]I\subseteq [0,1], ∣#{n<m≤n+k:xm∈I}−∣I∣k∣<ϵk.\lvert \# \{ n<m\leq n+k : x_m\in I\} - \lvert I\rvert k\rvert < \epsilon k. Is it true that, for every α\alpha, the sequence {αpn}\{ \alpha p_n\} is not well-distributed, if pnp_n is the sequence of primes?

References

Progress summary

Refreshed
Claimed solved

A 2026 manuscript claims a complete proof that the conjecture holds for every real number, but independent verification has not been reported.

Erdős asked whether, for every real α\alpha, the fractional parts along the primes are never well-distributed; he had earlier claimed and then retracted an existential version. The universal assertion is now claimed in a manuscript by Alexeev, Putterman, Sawhney, Sellke, and Valiant.

Known results

  • Erdős, 1964: claimed an irrational α\alpha with failure of well-distribution; he retracted the claim in 1985.
  • Champagne, Lê, Liu, and Wooley, 2024: proved that an irrational, indeed transcendental, α\alpha with this failure exists.
  • Vinogradov: proved ordinary equidistribution for every irrational α\alpha, a weaker property.

March 2026 universal-proof claim

The manuscript claims that every real α\alpha yields arbitrarily long strings of consecutive primes whose fractional parts cluster, using Dirichlet approximation and Maynard–Tao–BFT results. It says the proof came entirely from an internal OpenAI model; the claim remains unverified.

Current status (as of September 2026): The existential case is established, while the universal statement is claimed in arXiv:2603.29961 but remains unverified.

Sources

Solutions 0

No solutions have been posted yet.