Wiki
Wiki

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

Updated


Claim. Alexandru Zaharescu, On a conjecture of Graham, J. Number Theory 27 (1987), no. 1, 33--40. The paper proves Graham's conjecture, the statement of Problem 402, for every sufficiently large set. As the zbMATH review (Zbl 0629.10003) states it: let p(n)p(n) be the largest prime below 2n2n and f(n)=2n−p(n)f(n)=2n-p(n); the paper shows that for every chain 0<a1<⋯<an0<a_1<\cdots<a_n of integers some pair has

aigcd⁡(ai,aj)≥n\frac{a_i}{\gcd(a_i,a_j)}\ge n

whenever f(n)<nf(n)<\sqrt n, and then, through Huxley's bound f(n)≤n7/12+εf(n)\le n^{7/12+\varepsilon} for large nn, whenever nn is sufficiently large. Since a/gcd⁡(a,b)≥∣A∣a/\gcd(a,b)\ge|A| is the problem's inequality gcd⁡(a,b)≤a/∣A∣\gcd(a,b)\le a/|A|, this is the problem's claim for every finite set of at least N0N_0 elements. Balasubramanian and Soundararajan describe the argument on p. 1 of their 1996 paper: for a set that violates the conjecture Zaharescu finds α\alpha with rp(α)≥2r_p(\alpha)\ge2, where rp(α)=#{d:αd,(p−α)d∈A}r_p(\alpha)=\#\{d:\alpha d,(p-\alpha)d\in A\} for a prime pp near 2N2N, then many β\beta with rp(β)≥1r_p(\beta)\ge1 coprime to it, a contradiction once pp is close enough to 2N2N; by the nature of the prime-gap input the threshold is of the order e106e^{10^6}. That introduction counts the result as the conjecture "in its weaker form", without the equality case, and the review states only the inequality, while the site's commentary credits Szegedy and Zaharescu with the sharper version that characterizes equality; the page records the disagreement. Szegedy's independent proof is on its own page, and the full resolution on the page of Balasubramanian and Soundararajan.

Covers. The problem's inequality for every finite set AA with ∣A∣≥N0|A|\ge N_0, for some absolute N0N_0. Not covered: the sets of fewer than N0N_0 elements, where N0N_0 is of the order e106e^{10^6}.

Depends on. No page of this wiki.

Acceptance. Refereed: the Journal of Number Theory is a refereed journal, and the paper appeared in its volume 27 (1987). The site's commentary credits Szegedy and Zaharescu with the large-set case, while its PROVED label settles the problem through Balasubramanian and Soundararajan, so no reviewed evidence is listed here. The journal record dates the issue to September 1987, so the page is dated to its first day.