Wiki
Wiki

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

Updated


Statement

With Y(x)Y(x) as in Definition 1 (the largest yy such that residue classes ap mod pa_p\bmod p, one for each prime p≤xp\le x, cover [y]={1,…,⌊y⌋}[y]=\{1,\ldots,\lfloor y\rfloor\}) and P(x)P(x) the product of the primes less than or equal to xx:

Lemma 1.1 (p. 3). G(P(x)+Y(x)+x)≥Y(x)G(P(x)+Y(x)+x)\ge Y(x), where G(X)G(X) is the largest gap between consecutive primes less than XX.

Display (1.3) (p. 4). If nn is a positive integer, Jacobsthal's function j(n)j(n) is "the maximal gap between integers coprime to nn. In particular j(P(x))j(P(x)) is the maximal gap between numbers free of prime factors ≤x\le x, or equivalently 11 plus the longest string of consecutive integers, each divisible by some prime p≤xp\le x." The construction in the proof of Lemma 1.1 "in fact proves that"

Y(x)=j(P(x))−1.Y(x)=j(P(x))-1.

Source. K. Ford, B. Green, S. Konyagin, J. Maynard and T. Tao, Long gaps between primes, arXiv:1412.5029v3 (14 July 2016, 40 pp.); Lemma 1.1 with its proof on p. 3 and (1.3) on p. 4, read on the page images and in the text layer. Published in J. Amer. Math. Soc. 31 (2018), no. 1, 65--105, DOI 10.1090/jams/876; the journal text was not compared.

Read depth. Claims checked for both statements; the proof of Lemma 1.1 was read in full (below) and its steps were followed here. Display (1.3) carries no separate proof in the paper; the two inequalities it asserts were checked here from the definitions: a covering of [y][y] yields, through the mm of the proof, the yy consecutive integers m+1,…,m+ym+1,\ldots,m+y each divisible by a prime p≤xp\le x, so j(P(x))≥y+1j(P(x))\ge y+1; conversely a run of j(P(x))−1j(P(x))-1 consecutive integers m+1,…,m+j−1m+1,\ldots,m+j-1 each sharing a factor with P(x)P(x) gives the covering ap:=−m mod pa_p:=-m\bmod p of [j−1][j-1], so Y(x)≥j(P(x))−1Y(x)\ge j(P(x))-1. This check is an authored remark, not the paper's text.

Proof pointer

The paper's proof (p. 3) is a Chinese remainder construction. Fix a covering of [y][y], y=Y(x)y=Y(x), by one class ap mod pa_p\bmod p for each prime p≤xp\le x, and take m∈(x,x+P(x)]m\in(x,x+P(x)] with m≡−ap(modp)m\equiv-a_p\pmod p for every such pp. Each t∈[y]t\in[y] lies in some class ap mod pa_p\bmod p, so pp divides m+tm+t, and m+t>x≥pm+t>x\ge p makes m+tm+t composite. The yy consecutive integers m+1,…,m+ym+1,\ldots,m+y are therefore all composite, and as m+y≤P(x)+Y(x)+xm+y\le P(x)+Y(x)+x this gives G(P(x)+Y(x)+x)≥Y(x)G(P(x)+Y(x)+x)\ge Y(x).

Dependencies

The Chinese remainder theorem only.

Bears on

  • Problem 687: the transfer from Y(x)Y(x) to prime gaps and the identity with Jacobsthal's function that the site's commentary alludes to ("associated with Jacobsthal").
  • Problem 970: (1.3) is the identity through which the paper's bound (1.2) becomes a lower bound for the Jacobsthal function of a number with π(x)\pi(x) prime factors.
  • Problem 929: the same Chinese remainder construction shows that S(k)S(k) is the least xx with Y(x)≥kY(x)\ge k (made explicit on the problem page).