Wiki
Wiki

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

Updated

Problem 1002

../

claims/: The 2 claim pages of Problem 1002, one per claimant's result; the problem's standing derives from them.


Statement. For any 0<α<10<\alpha<1, let

f(α,n)=1log⁡n∑1≤k≤n(12−{αk}).f(\alpha,n)=\frac{1}{\log n}\sum_{1\leq k\leq n}(\tfrac{1}{2}-\{ \alpha k\}).

Does f(α,n)f(\alpha,n) have an asymptotic distribution function?

In other words, is there a non-decreasing function gg such that g(−∞)=0g(-\infty)=0, g(∞)=1g(\infty)=1, and

lim⁡n→∞∣{α∈(0,1):f(α,n)≤c}∣=g(c)?\lim_{n\to \infty}\lvert \{ \alpha\in (0,1): f(\alpha,n)\leq c\}\rvert=g(c)?

Status. The site labels the problem OPEN (proof-claims thread of 2026-10-07). Two full claims of July 2026, developed independently and both crediting AI systems, assert the answer yes with the same explicit limit, the centered Cauchy law of scale 1/(2π)1/(2\pi): Kwon's manuscript and Wang's preprint with a Lean 4 proof. Neither has been accepted by the site or refereed. This page departs from the site's label and records the problem as proved, because this corpus built and audited a port of Wang's Lean proof, whose two main theorems answer the site's question yes and pin the Cauchy limit (Formalization).

Source. erdosproblems.com/1002, accessed 2026-09-04. Cite as: T. F. Bloom, Erdős Problem #1002, https://www.erdosproblems.com/1002.

References.

  • [Ke60] Kesten, Harry, Uniform distribution mod 1{\rm mod}\,1. Ann. of Math. (2) (1960), 445-471.

Formalization. Statement in formal-conjectures (pinned at the repository's commit of 2026-09-18), tagged research open and citing no formal proof. A port of Wang's Lean development to Lean v4.33.0, in Boris Alexeev's repository, was built and audited by this corpus: the axioms of its two main theorems are exactly the standard three, both match their comparator challenge, and their statements give the site's question answered yes and the explicit Cauchy limit, so Wang's claim is accepted with formalized evidence. Wang's own development and Kwon's Lean development are linked from their claim pages; neither was built here.

Current assessment

The site labels Problem 1002 OPEN (accessed 2026-09-04; proof-claims thread of 2026-10-07). This corpus accepts the answer yes, with the centered Cauchy law of scale 1/(2π)1/(2\pi) as the limit, on Wang's claim: the port of its Lean proof that this corpus built proves both the existence of an asymptotic distribution function in the site's wording and the explicit Cauchy limit, using only the standard axioms, as Wang's claim page records. The acceptance rests on that formal proof alone: neither prose paper is refereed, the site has accepted neither claim, and no outside reviewer is recorded. This page records no literature search beyond the sources below.

Proof claims on the site (thread as of 2026-10-07). The proof-claims tab carries two full claims of the same theorem: Sangyoon Kwon's manuscript, public on GitHub since 2026-07-13 and submitted to the thread on 2026-07-23 at a moderator's invitation, and Shouqiao Wang's preprint, submitted 2026-07-21 with a Lean 4 formalization added on 2026-07-24. Both state that Sn(α)/log⁡nS_n(\alpha)/\log n converges in distribution to the centered Cauchy law of scale 1/(2π)1/(2\pi), so that g(c)=12+π−1arctan⁡(2πc)g(c)=\tfrac12+\pi^{-1}\arctan(2\pi c); the two claimants affirmed on the thread that the proofs were found independently, and an outside report of 2026-08-01 by Millennium Research rebuilt Wang's Lean development, checked its axioms and its statement against the site's wording, and read Kwon's manuscript without finding a fatal error. A Lean formalization of Kwon's proof, co-authored with that report's authors, followed in September 2026. The claim pages record the statements, the links and the limits of this evidence: the site labels the problem OPEN, no curator comment exists and nothing is refereed; this corpus built a port of Wang's Lean development and not Kwon's, which stays a pending full claim.

Progress

Kesten's 1960 paper is the cited historical reference; only DOI metadata is recorded here. Frühwirth and Hauke study temporal limits for fixed rotations, whereas this question asks for a spatial distribution over α\alpha. Neither reference settles that question. The library also holds Wang's 2026 proposed solution, an unrefereed preprint claiming that SN(α)/log⁡NS_N(\alpha)/\log N converges in distribution to a Cauchy law. The result is accepted on this corpus's build of a port of its Lean proof, recorded with Kwon's independent manuscript on the claim pages named under Current assessment.

Known Results

Kesten's 1960 paper [Ke60] is the cited background: it treats the sum with the starting point averaged as a second variable and proves a Cauchy limit law there; its library card records the citation only. No refereed result settles the fixed-start question asked. The two claims recorded above are the answers offered on the site's proof-claims tab: Wang's, accepted here on its formal proof, and Kwon's, pending.

Linked library material

These entries are derived from explicit links on library pages. They are navigation only and do not by themselves record mathematical progress.