Wiki
Wiki

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

Updated


Statement

An isosceles set is a set of points among any three of which at most two distances occur (§7.1, p. 46); see Theorem 7.2.2 for the chapter's conventions.

Theorem 7.2.5 (p. 48). Quoted, because the problem pages rest on its wording: "Let XX be an isosceles set in RdR^d, then card⁡(X)≤12(d+1)(d+2)\operatorname{card}(X)\le\frac12(d+1)(d+2). Equality implies that XX is a two-distance set, or a spherical two-distance set together with its center."

In the corpus's words: every isosceles set in Rd\mathbb{R}^d has at most (d+22)\binom{d+2}{2} points, and a set attaining the bound is either a two-distance set or a two-distance set lying on a sphere with the center of that sphere adjoined. The theorem gives no lower bound and does not say for which dd the bound is attained.

Source. A. Blokhuis, Few-distance sets, CWI Tract 7, Centrum voor Wiskunde en Informatica, Amsterdam, 1984; Theorem 7.2.5 on printed p. 48, its proof on p. 49. The edition read is identified in the source digest.

Read depth. Claims checked: the statement was read clause by clause on the page image and its proof read in full and followed. Nothing here is independently reviewed.

Proof pointer

Induction on dd (p. 49). For d=1d=1 an isosceles set has at most 33 points. For d=2d=2 the proof cites Kelly (the tract's reference [K], Amer. Math. Monthly 54 (1947), 227--229) for the maximum 66, attained only by the regular pentagon with its center. For d>2d>2: a two-distance set obeys the bound by Theorem 4.1.1; otherwise Theorem 7.2.2 gives a decomposition (X1,X2)(X_1,X_2). If dim⁡(X1)≠0\dim(X_1)\ne0, Lemma 7.2.1 gives 0<dim⁡(X1)<d0<\dim(X_1)<d and dim⁡(X1)+dim⁡(X2)≤d\dim(X_1)+\dim(X_2)\le d, and the induction hypothesis applied to both parts gives a strict inequality. If dim⁡(X1)=0\dim(X_1)=0, then X1X_1 is a single point and X2X_2 lies on a sphere centered at it; when X2X_2 is not a two-distance set it decomposes again and the first case applies, and otherwise card⁡(X)≤1+12d(d+3)=12(d+1)(d+2)\operatorname{card}(X)\le1+\tfrac12d(d+3)=\tfrac12(d+1)(d+2). The bound 12d(d+3)\tfrac12d(d+3) for a two-distance set on a sphere in Rd\mathbb{R}^d is used at this step without a reference on the page; it is the spherical bound of Delsarte, Goethals and Seidel, whose work the tract's introduction cites (p. 1). Equality thus forces a two-distance set or a centered spherical two-distance set.

Dependencies

Within the tract: Theorem 4.1.1 with s=2s=2, Lemma 7.2.1 (p. 46) and Theorem 7.2.2. Outside it: Kelly's planar result for d=2d=2 and the spherical two-distance bound 12d(d+3)\tfrac12d(d+3).

Bears on

  • Problem 503: an upper bound (d+22)\binom{d+2}{2} on the size of the largest A⊆RdA\subseteq\mathbb{R}^d in which every three points determine an isosceles triangle, the quantity the problem asks for; the theorem does not determine that quantity.
  • Problem 1088: since three points have pairwise distinct distances exactly when they do not form an isosceles triangle, the theorem gives fd(3)≤(d+22)+1f_d(3)\le\binom{d+2}{2}+1 in that problem's notation, as the problem's claim page for this tract records.