Wiki
Wiki

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

Updated


Statement

Quoted (p. 112): "Finally I shall deal with the following problem of Erdös, Herzog and Piranian [2, Problem 13]. Let zνz_\nu be nn complex numbers which satisfy ∣zμ−zν∣≤2|z_\mu-z_\nu|\le2 (μ,ν=1,⋯ ,n\mu,\nu=1,\cdots,n). Is ∏ν=1∏μ≠ν∣zμ−zν∣\prod_{\nu=1}\prod_{\mu\ne\nu}|z_\mu-z_\nu| maximal if the zνz_\nu are the vertices of a regular nn-gon of diameter 2? We denote the maximum by Δn\Delta_n:

(10)Δn=max⁡z1,⋯,zn∣zμ−zν∣≤2∏ν=1n∏μ≠ν∣zμ−zν∣.(10)\qquad \Delta_n=\max_{\substack{z_1,\cdots,z_n\\|z_\mu-z_\nu|\le2}} \prod_{\nu=1}^n\prod_{\mu\ne\nu}|z_\mu-z_\nu| .

The conjecture implies that Δn=nn\Delta_n=n^n for even nn and Δn=nn(cos⁡π/2n)−n(n−1)\Delta_n=n^n(\cos\pi/2n)^{-n(n-1)} for odd nn. The last quantity is nn(1−π28n2+⋯ )−n(n−1)∼nneπ2/8n^n(1-\frac{\pi^2}{8n^2}+\cdots)^{-n(n-1)}\sim n^ne^{\pi^2/8}."

Theorem 15 (p. 113). "Let KK be a convex continuum of capacity 1. Then, for zν∈Kz_\nu\in K (ν=1,⋯ ,n\nu=1,\cdots,n), ∏ν=1n∏μ≠ν∣zμ−zν∣≤24(n−1)nn\prod_{\nu=1}^n\prod_{\mu\ne\nu}|z_\mu-z_\nu|\le2^{4(n-1)}n^n."

Theorem 16 (p. 114). "If Δn\Delta_n is defined by (10), then Δn≤24(n−1)⋅nn\Delta_n\le2^{4(n-1)}\cdot n^n."

Remark (pp. 114--115, quoted in part): "We can make the following observation in favor of the conjecture of Erdös, Herzog and Piranian about Δn\Delta_n. The convex hull KnK_n of a maximal system {zn(n),⋯ ,zn(n)}\{z_n^{(n)},\cdots,z_n^{(n)}\} [sic] is nearly a disk, for large nn."

The product in (10) is Problem 1045's Δ(z1,…,zn)\Delta(z_1,\ldots,z_n) and the DkD_k of Danzer and Pommerenke (1967); the even-nn value nnn^n the conjecture implies was refuted there for every even n≥4n\ge4 (card), and this paper proves nothing about the regular polygon.

Source. Ch. Pommerenke, On metric properties of complex polynomials, Michigan Math. J. 8 (1961), no. 2, 97--115; the problem and display (10) on printed p. 112, Lemma 6 on pp. 112--113, Theorem 15 on p. 113 with its proof on pp. 113--114, Theorem 16 with its proof and the remark on pp. 114--115 (PDF pp. 16--19 of the publisher's scan), read on the page images (the scan has no text layer). The copy read is identified in the source digest.

Read depth. Claims checked: the problem as recalled, display (10), Lemma 6, Theorems 15 and 16 and the remark were read clause by clause on the page images on 2026-09-22; the proofs of Theorems 15 and 16 and of the remark (half a page together) were read in full and followed, Lemma 6 and Szegő's determinant inequality being taken as printed. The proof of Lemma 6 (p. 113) was read for structure and not checked. Nothing here is independently reviewed.

Proof pointer

Lemma 6 (pp. 112--113): for a convex continuum KK of capacity 1 there are monic polynomials fn(z)=zn+⋯f_n(z)=z^n+\cdots, n=1,2,…n=1,2,\ldots, with zeros in KK and max⁡z∈K∣fn(z)∣≤4\max_{z\in K}|f_n(z)|\le4; the zeros are ψ(e2πiν/n)\psi(e^{2\pi i\nu/n}) for the exterior map ψ(w)=w+⋯\psi(w)=w+\cdots of KK, and for fixed z∈Kz\in K the function Φ(w)=e−πi(n+1)/n∏ν(ψ(e2πiν/nw)−z)1/n\Phi(w)=e^{-\pi i(n+1)/n}\prod_\nu(\psi(e^{2\pi i\nu/n}w)-z)^{1/n} is starlike and univalent in ∣w∣>1|w|>1 (convexity of KK makes each factor starlike about 00), so Ψ(w)=Φ(w1/n)n=w+⋯\Psi(w)=\Phi(w^{1/n})^n=w+\cdots is univalent and nonvanishing in ∣w∣>1|w|>1, max⁡∣w∣=1∣Ψ∣≤4\max_{|w|=1}|\Psi|\le4, and ∣fn(z)∣=∣Ψ(1)∣≤4|f_n(z)|=|\Psi(1)|\le4.

Theorem 15 (pp. 113--114): ∏ν∏μ≠ν∣zμ−zν∣\prod_\nu\prod_{\mu\ne\nu}|z_\mu-z_\nu| is the squared modulus of the Vandermonde determinant det⁡(zjk−1)\det(z_j^{k-1}), which equals det⁡(fk−1(zj))\det(f_{k-1}(z_j)) for any monic fkf_k of degree kk (f0=1f_0=1); Hadamard's determinant inequality gives nnmax⁡K∣f1∣2⋯max⁡K∣fn−1∣2n^n\max_K|f_1|^2\cdots\max_K|f_{n-1}|^2 (an inequality the paper attributes to Szegő, citing footnote 7 on p. 236 of its [3]), and Lemma 6 bounds each factor by 1616, giving 42(n−1)nn4^{2(n-1)}n^n.

Theorem 16 (p. 114): for a maximal system with convex hull KK, Theorem 15 applied after scaling KK to capacity 1 gives (12) Δn≤24(n−1)nn(cap⁡K)n(n−1)(12)\ \Delta_n\le2^{4(n-1)}n^n(\operatorname{cap}K)^{n(n-1)}, and 2cap⁡K≤diam⁡K≤22\operatorname{cap}K\le\operatorname{diam}K\le2 gives cap⁡K≤1\operatorname{cap}K\le1. Remark (pp. 114--115): if the hulls KnK_n were not nearly disks, a subsequence would converge to a convex K0K_0 of diameter at most 2 that is not a disk, so cap⁡K0<1\operatorname{cap}K_0<1 and cap⁡Knk≤1−δ<1\operatorname{cap}K_{n_k}\le1-\delta<1, and (12) would give Δnk≤24(nk−1)nknk(1−δ)nk(nk−1)<1\Delta_{n_k}\le2^{4(n_k-1)}n_k^{n_k}(1-\delta)^{n_k(n_k-1)}<1 for large kk, which contradicts Δnk≥nknk\Delta_{n_k}\ge n_k^{n_k}.

Dependencies

Within the paper: Lemma 6. Outside it: Szegő's form of Hadamard's determinant inequality (Fekete, Math. Z. 17 (1923), 228--249, footnote 7 on p. 236, the paper's [3]; not held), the bound max⁡∣w∣=1∣Ψ∣≤4\max_{|w|=1}|\Psi|\le4 for a univalent nonvanishing Ψ(w)=w+⋯\Psi(w)=w+\cdots in ∣w∣>1|w|>1, and 2cap⁡K≤diam⁡K2\operatorname{cap}K\le\operatorname{diam}K.

Bears on

  • Problem 1045: the first general upper bound Δn≤24(n−1)nn\Delta_n\le2^{4(n-1)}n^n on the problem's maximum, exponentially above the conjectured nnn^n scale, later replaced by the nnexp⁡(15n6/7)n^n\exp(15n^{6/7}) of Danzer and Pommerenke (1967); the remark shows the convex hull of a maximal system is nearly a disk for large nn. The paper decides nothing about the regular polygon, and its record of the conjecture's even value nnn^n is the statement the 1967 paper refutes.