Wiki
Wiki

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

Updated


Statement

With the notation of printed p. 105 (ϱ(P,Q)\varrho(P,Q) the distance between points of the plane, ∥x∥\|x\| the distance from the real xx to the nearest integer, and N(X,δ)N(X,\delta) the maximum number of points P1,…,PmP_1,\ldots,P_m in the circle of radius XX with ∥ϱ(Pi,Pj)∥≥δ\|\varrho(P_i,P_j)\|\ge\delta for 1≤i<j≤m1\le i<j\le m), as printed on pp. 106--107:

Theorem 1. "Let

0<δ≤16⋅84(5)0<\delta\le\frac1{6\cdot8^4} \tag{5}

and XX be sufficiently large depending on δ\delta. Then

N(X,δ)>X1/2−δ1/7.(6)N(X,\delta)>X^{1/2-\delta^{1/7}}. \tag{6}

The theorem implies obviously the

Corollary. For ε>0\varepsilon>0 arbitrary small, there exists δ0=δ0(ε)>0\delta_0=\delta_0(\varepsilon)>0 such that for 0<δ<δ0(ε)0<\delta<\delta_0(\varepsilon), N(X,δ)>X1/2−εN(X,\delta)>X^{1/2-\varepsilon} whenever XX is large enough (depending on ε\varepsilon and δ\delta)."

Since the exponent 1/2−δ1/71/2-\delta^{1/7} is positive in the range (5), N(X,δ)→∞N(X,\delta)\to\infty as X→∞X\to\infty for every such δ\delta.

Source. A. Sárközy, On distances near integers, II, Studia Sci. Math. Hungar. 11 (1976), 105--111; Theorem 1 on printed p. 106 and the Corollary on p. 107 (PDF pp. 2--3 of the extract of the volume scan; volume physical pp. 112--113), read on the rendered page images (the OCR text layer garbles the formulas). The edition read is identified in the source digest.

Read depth. Claims checked: the notation, the Lemma, Theorem 1 and the Corollary were read clause by clause on the page images. The proof (pp. 107--110) was read for its structure and not checked.

Proof pointer

Pp. 107--110. Choose the integer kk with (7) 1/(6k4)≥δ>1/(6(k+1)4)1/(6k^4)\ge\delta>1/(6(k+1)^4), so that k≥8k\ge8 and (9) k>δ−1/7k>\delta^{-1/7}, and the integer tt with (10) 2k2t+3≤X<2k2(t+1)+32k^{2t+3}\le X<2k^{2(t+1)+3}. Take all points (13) P(u)=(x(u),y(u))P^{(u)}=(x^{(u)},y^{(u)}) with x(u)=∑i=0tεi(u)k2i+2x^{(u)}=\sum_{i=0}^t\varepsilon_i^{(u)}k^{2i+2} and y(u)=∑i=0tεi(u)kiy^{(u)}=\sum_{i=0}^t\varepsilon_i^{(u)}k^i, the digits satisfying (14) 0≤εi(u)≤k−20\le\varepsilon_i^{(u)}\le k-2; there are (15) m=(k−1)t+1>X1/2−δ1/7m=(k-1)^{t+1}>X^{1/2-\delta^{1/7}} of them and they lie in the circle of radius XX. For u≠vu\ne v the Lemma of p. 106 (if aa is a positive integer and 3δ<b2/a<2(1−δ)3\delta<b^2/a<2(1-\delta) then ∥a2+b2∥>δ\|\sqrt{a^2+b^2}\|>\delta) is applied to a=x(u)−x(v)a=x^{(u)}-x^{(v)} and b=y(u)−y(v)b=y^{(u)}-y^{(v)}, whose sizes are controlled through the highest differing digit ((19)--(28)), giving ∥ϱ(P(u),P(v))∥≥δ\|\varrho(P^{(u)},P^{(v)})\|\ge\delta (16). Not reconstructed here.

Dependencies

The Lemma of p. 106 (proved on the same page from a+δ<a2+b2<a+(1−δ)a+\delta<\sqrt{a^2+b^2}<a+(1-\delta)); otherwise self-contained.

Bears on

  • Problem 466: the problem asks for some δ>0\delta>0 with N(X,δ)→∞N(X,\delta)\to\infty; Theorem 1 gives N(X,δ)>X1/2−δ1/7→∞N(X,\delta)>X^{1/2-\delta^{1/7}}\to\infty for every δ≤1/(6⋅84)\delta\le1/(6\cdot8^4), the site's "N(X,δ)>X1/2−δ1/7N(X,\delta)>X^{1/2-\delta^{1/7}}" for all sufficiently small δ\delta; the paper reports Graham's earlier N(X,1/10)>110log⁡XN(X,1/10)>\frac1{10}\log X on pp. 105--106.
  • Problem 465: with Konyagin's N(X,δ)<C(δ)X1/2N(X,\delta)<C(\delta)X^{1/2} the exponent 1/21/2 is the truth for small δ\delta up to the δ1/7\delta^{1/7} and the constant.