Wiki
Wiki

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

Updated


Statement

Notation (p. 5-01). For a set AA of non-negative integers, D(A)\mathcal D(A) is its ordinary-difference set, the non-negative integers that are differences of two elements of AA.

Theorem 8 (p. 5-07). Let P(x)=amxm+am−1xm−1+⋯+a1x+a0P(x)=a_mx^m+a_{m-1}x^{m-1}+\cdots+a_1x+a_0 be a non-constant polynomial with integer coefficients and ama_m positive, and put K={P(n):n and P(n) positive integers}K=\{P(n): n\text{ and }P(n)\text{ positive integers}\}. Then K∩D(A)≠∅K\cap\mathcal D(A)\ne\emptyset for every set AA of positive upper density if and only if for every integer qq there is an integer mm with q∣P(m)q\mid P(m). (The print reuses the letter mm for the degree and for the integer in the divisibility condition, and says "every integer qq" here; the deduction before the theorem works with every positive integer qq.)

Remarks (pp. 5-07 to 5-08). The result is obvious for polynomials of degree

  1. A monic PP with an integer root has the property, so in particular the kk-th powers do for every positive integer kk. The reducible polynomial (x2−a)(x2−b)(x2−ab)(x^2-a)(x^2-b)(x^2-ab), with aa, bb and abab integers that are not squares, has the property, since it has a linear factor modulo every qq. An irreducible PP of degree at least 2 does not, since by a theorem of Frobenius it has no linear factor modulo some prime qq. By Theorem 3, D(A)\mathcal D(A) may be replaced by D0(A)\mathcal D_0(A) in the statement (p. 5-08).

Proof pointer

P. 5-07, in outline. Necessity is the congruence condition of p. 5-06: if qq divides no value of PP, the multiples of qq form a set of positive density whose difference set misses KK. Sufficiency combines Kamae and Mendès France's criterion (if, for every positive integer qq, the elements of KK divisible by qq, taken in increasing order and multiplied by any irrational θ\theta, are uniformly distributed modulo 1, then KK meets every D(A)\mathcal D(A) with AA of positive upper density) with the uniform distribution modulo 1 of P(n)θP(n)\theta and of P(qn+r)θP(qn+r)\theta, cited from Kuipers and Niederreiter (Theorem 3.2, p. 27, and Theorem 2.1, p. 238); the condition that qq divides some value of PP makes the multiples of qq in KK a non-empty union of such classes.

Read depth

Claims checked: Theorem 8 and the remarks after it were read clause by clause on the page images of the print, and the outline of the deduction on p. 5-07 was followed. The Kamae-Mendès France criterion and the uniform-distribution inputs are cited, not proved, in the survey and were not checked.

Dependencies

Theorem 3 for the D0\mathcal D_0 form. External input: Kamae and Mendès France, Van der Corput's difference theorem (Israel J. Math., then to appear), Example 3 and Theorem 2; Kuipers and Niederreiter, Uniform distribution of sequences (1974), Theorem 3.2 (p. 27) and Theorem 2.1 (p. 238).

Source. Cam L. Stewart, On difference sets of sets of integers, Séminaire Delange-Pisot-Poitou, Théorie des nombres, 19e année (1977/78), Fasc. 1, Exp. No. 5, 8 pp.; pages are cited by the print's own numbering 5-01 to 5-08, as on the source card.

Bears on

No Erdős problem page of the corpus cites this theorem.