Wiki
Wiki

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

Updated


Setting

Definitions, pp. 52--53. A rectifiable curve σ\sigma is given by z=Zσ(s)z=Z_\sigma(s) with s∈[0,∣σ∣]s\in[0,|\sigma|] the arc-length parameter and ∣σ∣>0|\sigma|>0; a point z∈σz\in\sigma has multiplicity kσ(z)k_\sigma(z), the number of ss with Zσ(s)=zZ_\sigma(s)=z. A multiple union L=[σ1,…,σm]L=[\sigma_1,\ldots,\sigma_m] of rectifiable curves, which may intersect, has length ∣L∣=∣σ1∣+⋯+∣σm∣|L|=|\sigma_1|+\cdots+|\sigma_m|, counted with multiplicity.

For z1,z2∉σz_1,z_2\notin\sigma,

Ψ(σ;z1,z2)=∫0∣σ∣∣ dArg⁡Zσ(s)−z1Zσ(s)−z2∣,(4)\Psi(\sigma;z_1,z_2)=\int_0^{|\sigma|}\Bigl|\,d\operatorname{Arg} \frac{Z_\sigma(s)-z_1}{Z_\sigma(s)-z_2}\Bigr|, \tag{4}

the variation along σ\sigma of a continuous branch of the argument, where a factor ζ−zj\zeta-z_j is replaced by 11 when zj=∞z_j=\infty. Then Ψ(σ)=sup⁡{Ψ(σ;z1,z2):z1,z2∉σ}\Psi(\sigma)=\sup\{\Psi(\sigma;z_1,z_2):z_1,z_2\notin\sigma\} and Ψ0(σ)=sup⁡{Ψ(σ;z,∞):z∉σ}\Psi_0(\sigma)=\sup\{\Psi(\sigma;z,\infty):z\notin\sigma\} (5), with Ψ0≤Ψ≤2Ψ0\Psi_0\le\Psi\le2\Psi_0, Ψ(σ)≥2π\Psi(\sigma)\ge2\pi always, and equality for circles, lines and non-degenerate segments (p. 52). For a multiple union, Ψ(L;z1,z2)\Psi(L;z_1,z_2) is the sum of the Ψ(σk;z1,z2)\Psi(\sigma_k;z_1,z_2) and Ψ(L)\Psi(L) its supremum as in (5) (p. 53).

The analytic capacity of a compact K⊂CK\subset\mathbb C (p. 53) is the supremum of ∣c−1∣|c_{-1}| over functions f(z)=c−1/z+c−2/z2+⋯f(z)=c_{-1}/z+c_{-2}/z^2+\cdots holomorphic on the component G∞(K)G_\infty(K) of C‾∖K\overline{\mathbb C}\setminus K containing ∞\infty with sup⁡G∞(K)∣f∣≤1\sup_{G_\infty(K)}|f|\le1; and γ(L)=γ(⋃kσk)\gamma(L)=\gamma(\bigcup_k\sigma_k), multiplicity not counted.

Statement

Lemma 1 (p. 53). If L=[σ1,…,σm]L=[\sigma_1,\ldots,\sigma_m] is a multiple union of finitely many piecewise smooth curves and Ψ(L)<∞\Psi(L)<\infty, then

∣L∣=∣σ1∣+⋯+∣σm∣≤Ψ(L) γ(L).(6)|L|=|\sigma_1|+\cdots+|\sigma_m|\le\Psi(L)\,\gamma(L). \tag{6}

For every circle L=σL=\sigma, (6) is an equality (p. 53). The paper says (p. 53) that Lemmas 1 and 4 appeared, in a somewhat different form and without proof, in the author's 1985 note in Dokl. Akad. Nauk SSSR.

Extensions (pp. 55--56). Remark 1 states that Lemma 1 holds for arbitrary rectifiable LL, by inscribing polygonal lines and using the upper semicontinuity of analytic capacity. Lemma 1a (p. 56), which Remark 2 derives from Lemma 1 and whose proof applies Lemma 1 through Remark 1: if σ\sigma is rectifiable with Ψ(σ)<∞\Psi(\sigma)<\infty, EE is a compact subset of C\mathbb C and E={s∈[0,∣σ∣]:Zσ(s)∈E}\mathcal E=\{s\in[0,|\sigma|]:Z_\sigma(s)\in E\}, then ∣E∩σ∣:=mes⁡1E≤Ψ(σ)γ(E∩σ)|E\cap\sigma|:=\operatorname{mes}_1\mathcal E\le\Psi(\sigma)\gamma(E\cap\sigma) (11), the length counted with the multiplicity kσ(z)k_\sigma(z).

Proof pointer

P. 54. Orient each σk\sigma_k so that Re⁡dζ≥0\operatorname{Re}d\zeta\ge0; the multiple projection of LL to the real axis then has length at most ∣∫L+dζ∣|\int_{L^+}d\zeta| (7). The function w0(z)=∫L+dζ/(ζ−z)w_0(z)=\int_{L^+}d\zeta/(\zeta-z) maps G∞(L)G_\infty(L) into a horizontal strip of width at most Ψ(L)\Psi(L); scaling by π/Ψ(L)\pi/\Psi(L), exponentiating and a Möbius map (8) send G∞(L)G_\infty(L) into the unit disc with ∞↦0\infty\mapsto0, and the definition of analytic capacity bounds ∣∫L+dζ∣|\int_{L^+}d\zeta| by 2Ψ(L)γ(L)/π2\Psi(L)\gamma(L)/\pi. The same bound holds for the projection to every line, and Cauchy's formula ∣L∣=12∫0π∣Πφ∣ dφ|L|=\frac12\int_0^\pi|\Pi_\varphi|\,d\varphi gives (6).

Dependencies

The definitions (4)--(5) and of analytic capacity (pp. 52--53); Cauchy's projection formula, cited from Dolzhenko.

Source. V. I. Danchenko, The lengths of lemniscates. Variations of rational functions, Mat. Sb. 198 (2007), no. 8, 51--58 (in Russian); pages are the journal's, as on the source card.

Read depth. Claims checked: the statement, the definitions it uses, Remark 1 and Lemma 1a read on the print; the proof (p. 54) read for its structure, not checked step by step. Nothing here is independently reviewed.

Bears on

  • #114: background only. The lemma is the step of Theorem 1 that turns bounds on Ψ\Psi and γ\gamma of a lemniscate into the length bound 2πn2\pi n at r=1r=1; on its own it says nothing about which polynomial maximises the length.