Wiki
Wiki

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

Updated

Erdos 1988 solution problem grunbaum

../

lemma_1: Salamon and Erdős's band bounds: for 0 <= k <= n - 2, n points of which exactly n - k lie on a largest collinear set determine at most k(n-k) + C(k,2) + 1 lines, a sharp bound, and at least the Kelly-Moser bound k(n-k) - C(k,2) + 1.

lemma_2: Salamon and Erdős's lemma that when n >= k(k+1)/2 every integer from the Kelly-Moser bound M_min(k) up to M_max(k) is the number of lines of some configuration in the k-th band, except M_max(k) - 1 and M_max(k) - 3; in particular the Kelly-Moser bound is attained.

lemma_3: Salamon and Erdős's lemma on the upper part of the large bands: for n <= k(k+1)/2 and k <= n - 3 every integer between M_max(k) - 2(n-k) and M_max(k) is taken on except M_max(k) - 1 and M_max(k) - 3, which the paper uses to show that consecutive large bands overlap.

lemma_4: Salamon and Erdős's lemma that for n sufficiently large no configuration in a band beyond k = [sqrt(n+2)] has as few lines as M_max([sqrt(n+2)] - 1), so the large bands stay out of the region of separated bands; the proof uses Beck's theorem.

main_theorem: Salamon and Erdős's answer to Grünbaum's problem for n at least an unspecified n*: the possible numbers of lines determined by n points are the separated bands for k < [sqrt(n+2)], a run of consecutive values whose lower end is given exactly in five cases and which ends at C(n,2) - 4, and the two values C(n,2) - 2 and C(n,2).


P. Salamon, P. Erdős: The solution to a problem of Grünbaum, Canad. Math. Bull. 31 (1988) no. 2, 129--138, DOI 10.4153/CMB-1988-020-2 (MR 89f:52022; Zentralblatt 606.05005). The file prints "© Canadian Mathematical Society 1986" on its first page, the year as printed, every other right reserved.

Salamon and Erdos determine, for all sufficiently large n, every integer that occurs as the number of connecting lines (lines through at least two points) of some set of n points in the plane, solving a problem of Grunbaum. The analysis is organized into bands: the kth band consists of configurations whose largest collinear subset has n-k points, and for k of order at most about sqrt(n) the value ranges of distinct bands do not overlap, so the spectrum can be described band by band. For binom(k,2) <= n-k they show that the Kelly-Moser lower bound M_min(k) = k(n-k) - binom(k,2) + 1 on the line count of the kth band is sharp, and that every value between M_min(k) and the band maximum M_max(k) = k(n-k) + binom(k,2) + 1 is attained except M_max - 1 and M_max - 3. They also give exact formulas for the bottom of the run of consecutive attained values that extends down from binom(n,2) - 4, in particular fixing the best constant c = 1 in Erdos's earlier estimate c n^{3/2} for where that run begins. The paper displays the computed spectra for n = 22 through 28 and remarks on their resemblance to physical spectra, a connection the authors expect to be useful for simulated annealing. This is the source for problem 606 on the possible numbers of lines determined by n points.

Source: https://users.renyi.hu/~p_erdos/1988-36.pdf.

Read status: claims checked for Lemmas 1 to 4 (pp. 131--133), the five cases, the continuum and the mi(n)m_i^{(n)} formulas (pp. 133--136) and the remarks on n∗n^* (p. 137), read clause by clause on the page images of the print; the proofs of Lemmas 1 to 4 followed. Beck's theorem and the Kelly--Moser bound are cited, not proved, in the paper. Nothing here is independently reviewed.

Bears on. #606: the main result (pp. 133--137) determines the possible numbers of lines determined by nn points in the plane for every n≥n∗n\ge n^*, which is the problem's question for all sufficiently large nn; the paper leaves n<n∗n<n^* open and does not compute n∗n^* (p. 137).

Results.

  • Main result (pp. 133--137, unlabelled): for n≥n∗n\ge n^* the possible values are 11, the bands for 1≤k≤[n+2]−11\le k\le[\sqrt{n+2}]-1 less their two top gaps, a continuum whose lower end is given exactly in five cases and which ends at (n2)−4\binom n2-4, and (n2)−2\binom n2-2, (n2)\binom n2; with the best constant c=1c=1 in Erdős's cn3/2cn^{3/2} (p. 134) and explicit formulas for the ordered values mi(n)m_i^{(n)} (pp. 134--136).
  • Lemma 1 (p. 131): for 0≤k≤n−20\le k\le n-2 the kk-th band has at most Mmax⁡(k)=k(n−k)+(k2)+1M_{\max}(k)=k(n-k)+\binom k2+1 lines, attained, and at least the Kelly--Moser bound Mmin⁡(k)=k(n−k)−(k2)+1M_{\min}(k)=k(n-k)-\binom k2+1.
  • Lemma 2 (p. 132): for n≥k(k+1)/2n\ge k(k+1)/2 every value from Mmin⁡(k)M_{\min}(k) to Mmax⁡(k)M_{\max}(k) occurs in the kk-th band except Mmax⁡−1M_{\max}-1 and Mmax⁡−3M_{\max}-3; in particular the Kelly--Moser bound is attained.
  • Lemma 3 (p. 132): for n≤k(k+1)/2n\le k(k+1)/2 and k≤n−3k\le n-3 every value from Mmax⁡(k)−2(n−k)M_{\max}(k)-2(n-k) to Mmax⁡(k)M_{\max}(k) is taken on except Mmax⁡−1M_{\max}-1 and Mmax⁡−3M_{\max}-3, with the paper's qualification for k=n−2,n−3k=n-2,n-3 (p. 133).
  • Lemma 4 (p. 133): for nn sufficiently large every configuration in a band with k>[n+2]k>[\sqrt{n+2}] has more than Mmax⁡([n+2]−1)M_{\max}([\sqrt{n+2}]-1) lines.

No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.