betaFeedback

WorksMathematics

A hyperbolic area proof of the Kahn-Lyubich Covering Lemma and its generalization

Abstract

The Covering Lemma of J. Kahn and M. Lyubich is a fundamental analytic tool for controlling conformal moduli of annuli under branched coverings of finite degree. In the last several decades, it has played a crucial role in establishing rigidity of multicritical polynomial and rational maps. The classical proof uses the Near-Degenerate Regime as the main mechanism. In this short note, we prove this lemma using hyperbolic area and the Dirichlet energy. We also establish a version of the Covering Lemma for holomorphic maps of infinite degree. Our argument is inspired by recent applications of the hyperbolic area method to the study of wandering domains in rational and transcendental dynamics.

Introduction

The Covering Lemma of J. Kahn and M. Lyubich [8] has been essential in establishing local connectivity and rigidity results for complex polynomials of higher degree [KL09a, KvS09, DY26] and, through puzzle renormalization via complex box mappings [9 CDKvS22], for rational maps, most notably Newton maps [4, 16, 13]. It has also been used to prove local connectivity of boundaries of bounded polynomial Fatou components [12, 17]. Furthermore, the Covering Lemma has played an important role in recent advances towards one of the central conjectures in the field – the MLC conjecture (“the Mandelbrot set is locally connected”), and in the development of the Near-Degenerate Regime mechanism; we refer the reader to a very nice survey [5].

The original proof of the Covering Lemma [8] relies on the near-degenerate philosophy and a careful analysis of conformal rectangles and their moduli (extremal widths). Inspired by recent applications of the hyperbolic area method to wandering domains in rational and transcendental dynamics [DPSU26, PRW26, Ye26a, Ye26b], in this note we present an alternative (somewhat elementary) proof using hyperbolic geometry, hyperbolic area, and its connection with conformal moduli and extremal widths through Dirichlet energy. Furthermore, we extend the Covering Lemma to maps of infinite degree.

We consider non-constant holomorphic maps f:U→Vf: U \to V between topological disks with finitely many singular values. Thus, there is a finite set Sf⊂VS_f \subset V such that f:U∖f−1(Sf)→V∖Sff: U \setminus f^{-1}(S_f) \to V \setminus S_f is a covering map; we choose SfS_f minimal with this property. The elements of SfS_f are critical or asymptotic values of ff. If ff is proper, then it has finite degree and its singular values are precisely its critical values.

This work was partially supported by Agencia Estatal de Investigación grants PID2023-147252NB-I00, CNS2025-166633, and by the Agència de Gestió d’Ajuts Universitaris i de Recerca of the Generalitat de Catalunya grant number 2025 SGR 00322.

Theorem 1 (The Covering Lemma). Let Λ⋐Λ′⋐U\Lambda\Subset\Lambda' \Subset U and B⋐B′⋐VB \Subset B' \Subset V be two nests of open Jordan disks in C^\widehat{\mathbb{C}}. Let f:U→Vf: U \to V be a non-constant holomorphic map with at most D≥1D \ge1 singular values. Assume that the restrictions f:Λ′→B′f: \Lambda' \to B' and f:Λ→Bf: \Lambda\to B are proper, and that the degree deg⁡(Λ′→B′)=d≤D\deg(\Lambda' \to B') = d \le D.

Then for every η∈(0,1]\eta\in(0,1] there exists an explicit ε=ε(η,D)\varepsilon= \varepsilon(\eta,D) with the following property: If the annulus U∖Λ‾U \setminus\overline{\Lambda} is ε\varepsilon-degenerate, i.e.,

mod⁡(U∖Λ‾)<ε,\operatorname{mod}(U \setminus\overline{\Lambda}) < \varepsilon,

then one of the following alternatives holds:

(1) either

mod⁡(U∖Λ‾)≥1ηmod⁡(B′∖B‾),\operatorname{mod}(U \setminus\overline{\Lambda}) \ge\frac{1}{\eta}\operatorname{mod}(B' \setminus\overline{B}),

(2) or the collar assumption mod⁡(B′∖B‾)≥ηmod⁡(U∖Λ‾)\operatorname{mod}(B' \setminus\overline{B}) \ge\eta\operatorname{mod}(U \setminus\overline{\Lambda}) is satisfied and

mod⁡(U∖Λ‾)≥η32d2mod⁡(V∖B‾).\operatorname{mod}(U \setminus\overline{\Lambda}) \ge\frac{\eta}{32d^{2}}\operatorname{mod}(V \setminus\overline{B}).

The Covering Lemma shows that if the annuli are sufficiently degenerate (see Figure 1), then their pullback is essentially controlled by the local degree dd, which may be much smaller than the number of sheets over V∖SfV \setminus S_f; the latter may be even infinite. This theorem recovers the statement of [8] when f:U→Vf: U \to V has finite degree (up to a universal multiplicative constant). Let us give some further remarks.

The domains in the Covering Lemma

Figure 1. The domains in the Covering Lemma.

Remark 1 (Value of ε\varepsilon). It will be clear from the proof that we can choose

ε=η32πD3.\varepsilon= \frac{\eta}{32\pi D^{3}}.

Remark 2. As in [9 CDKvS22], the Covering Lemma can be stated as

mod⁡(U∖Λ‾)≥min⁡{ε,1ηmod⁡(B′∖B‾),η32d2mod⁡(V∖B‾)}.\operatorname{mod}(U \setminus\overline{\Lambda}) \ge\min\left\{\varepsilon,\frac{1}{\eta}\operatorname{mod}(B' \setminus\overline{B}),\frac{\eta}{32d^{2}}\operatorname{mod}(V \setminus\overline{B})\right\}.

Remark 3 (Finite degree Covering Lemma). If f:U→Vf: U \to V is proper of degree DD, then our approach gives that, provided

mod⁡(U∖Λ‾)<132πD3=:ε,\operatorname{mod}(U \setminus\overline{\Lambda}) < \frac{1}{32\pi D^3} =: \varepsilon,

we have the following estimate:

mod⁡(V∖B‾)mod⁡(B′∖B‾)<32d2mod⁡(U∖Λ‾)2.\operatorname{mod}(V \setminus\overline{B})\operatorname{mod}(B' \setminus\overline{B}) < 32d^2 \operatorname{mod}(U \setminus\overline{\Lambda})^2.

This gives the desired bound under the collar assumption mod⁡(B′∖B‾)≥ηmod⁡(U∖Λ‾)\operatorname{mod}(B' \setminus\overline{B}) \ge\eta\operatorname{mod}(U \setminus\overline{\Lambda}). In particular, the degeneration parameter ε\varepsilon does not depend on the collar parameter η\eta.

Remark 4. The proof presented in this paper is motivated by the recent successful applications of the hyperbolic area method in the new proof of the absence of wandering domains in rational dynamics (Sullivan’s No Wandering Domains Theorem [15], see [Ye26b]) and the absence of bounded orbit wandering domains for entire, meromorphic, and, more crucially, locally defined holomorphic maps [18 DPSU26, PRW26].

Remark 5 (Further generalizations). We believe that the argument in this paper may be further extended to establish the Quasi-Additivity and Quasi-Invariance Laws of [8] in terms of extremal width.

AI assistance. ChatGPT was used in exploratory discussions regarding the content of this manuscript based on our previous work [DPSU26]. The final results and text were written by the author, who takes all responsibility (and blame, if necessary) for the content.

Some preliminary lemmas

Let XX be a hyperbolic Riemann surface. We will denote by ϱX\varrho_X the local density of the complete hyperbolic metric on XX (normalized to have curvature −1-1). For a curve γ⊂X\gamma\subset X, we will denote

ℓX(γ)=∫γϱX(z)∣dz∣\ell_X(\gamma) = \int_{\gamma} \varrho_X(z)|dz|

the hyperbolic length of this curve; and similarly, for a measurable subset E⊆XE \subseteq X, we will write

Area⁡X(E)=∫EϱX2(z)∣dz∣2\operatorname{Area}_X(E) = \int_E \varrho_X^2(z)|dz|^2

its hyperbolic area; here, z=x+iyz=x+iy is the local holomorphic coordinate on XX, and ∣dz∣2=dxdy|dz|^2=dxdy denotes the Euclidean area element.

We will need the concept of Dirichlet integrals and their relation to the conformal modulus of an annulus; see [8 Appendix]. See also [10 §6.4] for the introduction to this concept. We specify the definitions suitably for our purposes. Let Y⋐X⊂CY \Subset X \subset\mathbb{C} be a pair of open, compactly contained Jordan disks. Consider a continuous function h:X→Rh:X\to\mathbb{R} that is equal to 11 on YY, equal to 00 on ∂X\partial X, and is harmonic in X∖YX\setminus Y; we will call hh the equilibrium potential of the pair (X,Y)(X,Y). For such a potential, we can consider its Dirichlet energy integral

∫X∣∇h(z)∣2∣dz∣2=∫X∖Y‾∣∇h(z)∣2∣dz∣2.\int_X |\nabla h(z)|^2 |dz|^2 = \int_{X\setminus\overline{Y}} |\nabla h(z)|^2 |dz|^2.

It is easy to see that this integral does not depend on the choice of a conformal metric on XX, i.e., if ϱ(z)∣dz∣\varrho(z)|dz| is a new smooth metric on XX, then

∫X∣∇h(z)∣2∣dz∣2=∫X∥∇ϱh(z)∥ϱ2⋅ϱ(z)2∣dz∣2.(1)\int_X |\nabla h(z)|^2 |dz|^2 = \int_X \|\nabla_{\varrho} h(z)\|_{\varrho}^2 \cdot\varrho(z)^2 |dz|^2. \tag*{(1)}

where ∥⋅∥ϱ\lVert\cdot\rVert_{\varrho} and ∇ϱ\nabla_{\varrho} denote the norm and gradient in the metric ϱ\varrho. And therefore, it is also invariant under conformal transformations of XX. For t∈(0,1)t \in(0,1), we will write {h=t}\{h=t\} for the level curve in X∖Y‾X\setminus\overline{Y} corresponding to h=th=t. This curve is smooth. We adopt a similar notation for the super- and sub-level sets of hh.

Since Δh=0\Delta h=0 in X∖Y‾X\setminus\overline{Y}, the divergence theorem on the set {a<h<b}⊂X∖Y‾\{a<h<b\}\subset X\setminus\overline{Y} gives

0=∫{a<h<b}Δh(z)∣dz∣2=∫{h=b}∣∇h(z)∣∣dz∣−∫{h=a}∣∇h(z)∣∣dz∣.0=\int_{\{a<h<b\}}\Delta h(z)\lvert dz\rvert^{2}=\int_{\{h=b\}}\lvert\nabla h(z)\rvert\lvert dz\rvert-\int_{\{h=a\}}\lvert\nabla h(z)\rvert\lvert dz\rvert.

Thus, the value of the integral ∫{h=t}∣∇h(z)∣∣dz∣\int_{\{h=t\}}\lvert\nabla h(z)\rvert\lvert dz\rvert does not depend on the choice of tt. Therefore, by the co-area formula,

∫X∣∇h(z)∣2∣dz∣2=∫01(∫{h=t}∣∇h∣∣dz∣)dt=∫{h=t}∣∇h∣∣dz∣∀t∈(0,1).(2)\int_X\lvert\nabla h(z)\rvert^2\lvert dz\rvert^2=\int_0^1\left(\int_{\{h=t\}}\lvert\nabla h\rvert\lvert dz\rvert\right)dt=\int_{\{h=t\}}\lvert\nabla h\rvert\lvert dz\rvert\qquad\forall t\in(0,1). \tag*{(2)}

Note that the last identity is also independent of the choice of a conformal metric on XX. In particular,

∫X∣∇h(z)∣2∣dz∣2=1b−a∫{a<h<b}∣∇h(z)∣2∣dz∣2∀0≤a<b≤1(3)\int_X\lvert\nabla h(z)\rvert^2\lvert dz\rvert^2=\frac{1}{b-a}\int_{\{a<h<b\}}\lvert\nabla h(z)\rvert^2\lvert dz\rvert^2\qquad\forall0\leq a<b\leq1 \tag*{(3)}

Finally, the Dirichlet integral of the equilibrium potential of (X,Y)(X,Y) is related to the conformal modulus mod⁡(X∖Y‾)\operatorname{mod}(X\setminus\overline{Y}) of the annulus X∖Y‾X\setminus\overline{Y} as

1mod⁡(X∖Y‾)=∫X∣∇h∣2∣dz∣2=inf⁡v:X→Rv∣Y‾=1∫X∣∇v(z)∣2∣dz∣2,(4)\frac{1}{\operatorname{mod}(X\setminus\overline{Y})}=\int_X\lvert\nabla h\rvert^2\lvert dz\rvert^2=\inf_{\substack{v:X\to\mathbb{R}\\v\mid_{\overline{Y}}=1}}\int_X\lvert\nabla v(z)\rvert^2\lvert dz\rvert^2, \tag*{(4)}

where the last infimum is taken over all locally Lipschitz functions vv with compact support in XX that are equal to 1 on Y‾\overline{Y}. The last equality comes from the variational viewpoint on the Dirichlet problem. We refer to [8 Appendix] and the classical source [1] for the background on conformal moduli.

Lemma 2 ([DPSU26]). Let X⊂CX\subset\mathbb{C} be a topological disk, and Q⊂XQ\subset X be a finite collection of points. Then for every measurable set E⊆X∖QE\subseteq X\setminus Q

0≤(Area⁡X∖Q−Area⁡X)(E)≤2π⋅#Q.(5)0\leq(\operatorname{Area}_{X\setminus Q}-\operatorname{Area}_{X})(E)\leq2\pi\cdot\#Q. \tag*{(5)}

Here, the area gain is understood in the integral form

(Area⁡X∖Q−Area⁡X)(E):=∫E(ϱX∖Q2(z)−ϱX2(z))∣dz∣2.(\operatorname{Area}_{X\setminus Q}-\operatorname{Area}_{X})(E):=\int_E\left(\varrho_{X\setminus Q}^{2}(z)-\varrho_X^{2}(z)\right)\lvert dz\rvert^2.

Lemma 3. Let Y⋐XY\Subset X be a pair of open Jordan disks, and let γ⊂X∖Y‾\gamma\subset X\setminus\overline{Y} be a smooth simple closed curve enclosing YY. Then

1mod⁡(X∖Y‾)≤ℓX(γ)+2π3.(6)\frac{1}{\operatorname{mod}(X\setminus\overline{Y})}\leq\ell_X(\gamma)+\frac{2\pi}{3}. \tag*{(6)}

Proof. Since the modulus and the hyperbolic length are conformal invariants, we can assume that XX is the hyperbolic plane (modeled in the upper half plane H\mathbb{H} with metric ϱH(z)=1/y\varrho_{\mathbb{H}}(z)=1/y, z=x+iyz=x+iy). Let D⋐HD\Subset\mathbb{H} be a domain with rectifiable boundary. Then

Area⁡H(D)≤ℓH(∂D).(7)\operatorname{Area}_{\mathbb{H}}(D)\leq\ell_{\mathbb{H}}(\partial D). \tag*{(7)}

Indeed, since DD can be approximated by domains with smooth boundaries so that the length and area converge, it is enough to show (7) for domains with smooth boundaries ∂D\partial D. Let (∂x,∂y)(\partial_x,\partial_y) denote the standard basis in each tangent plane of H\mathbb{H}, then a direct computation shows that the vector field −y⋅∂y-y\cdot\partial_y has hyperbolic length 11 and divergence 11. Let ν\nu be the outward-facing normal vector field along ∂D\partial D. Set dAH:=ϱH2∣dz∣2dA_{\mathbb{H}}:=\varrho_{\mathbb{H}}^2|dz|^2 and dsH:=ϱH∣dz∣ds_{\mathbb{H}}:=\varrho_{\mathbb{H}}|dz| to be the hyperbolic area and length element in H\mathbb{H}, and ⟨⋅,⋅⟩H\langle\cdot,\cdot\rangle_{\mathbb{H}}, ∥⋅∥H\|\cdot\|_{\mathbb{H}} be the hyperbolic dot product and length. Using the divergence theorem,

Area⁡H(D)=∫Ddiv⁡(−y∂y) dAH=∫∂D⟨ν,−y∂y⟩H dsH≤∫∂D∥−y∂y∥H dsH=ℓH(∂D).\operatorname{Area}_{\mathbb{H}}(D)=\int_D\operatorname{div}(-y\partial_y)\,dA_{\mathbb{H}}=\int_{\partial D}\langle\nu,-y\partial_y\rangle_{\mathbb{H}}\,ds_{\mathbb{H}}\leq\int_{\partial D}\|-y\partial_y\|_{\mathbb{H}}\,ds_{\mathbb{H}}=\ell_{\mathbb{H}}(\partial D).

Remark 6. Inequality (7) also follows from the isoperimetric inequality in the hyperbolic plane that has the form ℓH2(∂D)≥4πArea⁡H(D)+Area⁡H2(D)\ell_{\mathbb{H}}^2(\partial D)\geq4\pi\operatorname{Area}_{\mathbb{H}}(D)+\operatorname{Area}_{\mathbb{H}}^2(D).

Let Γ\Gamma be the convex hull of γ\gamma (i.e., the smallest hyperbolically convex set in H\mathbb{H} that contains γ\gamma). By definition, Γ\Gamma is compact, Y‾⊂Γ\overline{Y}\subset\Gamma. Since γ\gamma is smooth, ∂Γ\partial\Gamma is rectifiable and ℓH(∂Γ)≤ℓH(γ)\ell_{\mathbb{H}}(\partial\Gamma)\leq\ell_{\mathbb{H}}(\gamma). In particular, using (7),

Area⁡H(Γ)≤ℓH(∂Γ)≤ℓH(γ).(8)\operatorname{Area}_{\mathbb{H}}(\Gamma)\leq\ell_{\mathbb{H}}(\partial\Gamma)\leq\ell_{\mathbb{H}}(\gamma). \tag*{(8)}

Let Γs\Gamma_s be the outer parallel set for Γ\Gamma at distance s≥0s\geq0, i.e., Γs:={z∈H:dist⁡H(z,Γ)≤s}\Gamma_s:=\{z\in\mathbb{H}:\operatorname{dist}_{\mathbb{H}}(z,\Gamma)\leq s\}. By construction, Γ=Γ0\Gamma=\Gamma_0. By the Steiner formula in the hyperbolic plane (see e.g. [14 Chapter 18]),

ℓH(∂Γs)=ℓH(∂Γ)cosh⁡s+(2π+Area⁡H(Γ))sinh⁡s.(9)\ell_{\mathbb{H}}(\partial\Gamma_s)=\ell_{\mathbb{H}}(\partial\Gamma)\cosh s+(2\pi+\operatorname{Area}_{\mathbb{H}}(\Gamma))\sinh s. \tag*{(9)}

Consider the function v(z):=e−dist⁡H(z,Γ)v(z):=e^{-\operatorname{dist}_{\mathbb{H}}(z,\Gamma)}. It satisfies v∣Γ=1v|_{\Gamma}=1 and v(z)→0v(z)\to0 when z→∂Hz\to\partial\mathbb{H}. Since vv is not compactly supported in H\mathbb{H}, for ε∈(0,1)\varepsilon\in(0,1), we define its cutoff vε(z):=max⁡{v(z)−ε,0}/(1−ε)v_\varepsilon(z):=\max\{v(z)-\varepsilon,0\}/(1-\varepsilon). This function is locally Lipschitz, vε∣Γ=1v_\varepsilon|_{\Gamma}=1, and the support of vεv_\varepsilon is precisely equal to Γlog⁡(1/ε)\Gamma_{\log(1/\varepsilon)}. By conformal invariance (1) and the variational approach (4),

1mod⁡(H∖Y‾)≤∫H∥∇ϱHvε(z)∥H2 dAH=1(1−ε)2∫{v>ε}∥∇ϱHv(z)∥H2 dAH.\frac{1}{\operatorname{mod}(\mathbb{H}\setminus\overline{Y})}\leq\int_{\mathbb{H}}\|\nabla_{\varrho_{\mathbb{H}}}v_\varepsilon(z)\|_{\mathbb{H}}^2\,dA_{\mathbb{H}}=\frac{1}{(1-\varepsilon)^2}\int_{\{v>\varepsilon\}}\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}^2\,dA_{\mathbb{H}}.

The right-hand side converges, as ε→0\varepsilon\to0, to ∫H∥∇ϱHv(z)∥H2 dAH\int_{\mathbb{H}}\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}^2\,dA_{\mathbb{H}}, thus giving the estimate

1mod⁡(H∖Y‾)≤∫H∥∇ϱHv(z)∥H2 dAH.(10)\frac{1}{\operatorname{mod}(\mathbb{H}\setminus\overline{Y})}\leq\int_{\mathbb{H}}\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}^2\,dA_{\mathbb{H}}. \tag*{(10)}

Note that ∇ϱHv(z)=−v(z)⋅∇ϱHdist⁡H(z,Γ)\nabla_{\varrho_{\mathbb{H}}}v(z)=-v(z)\cdot\nabla_{\varrho_{\mathbb{H}}}\operatorname{dist}_{\mathbb{H}}(z,\Gamma). Since the gradient of a hyperbolic distance function to a convex set always has hyperbolic norm one outside of that set, we obtain that ∥∇ϱHv(z)∥H=v(z)=e−dist⁡H(z,Γ)\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}=v(z)=e^{-\operatorname{dist}_{\mathbb{H}}(z,\Gamma)} everywhere outside of Γ\Gamma. By the co-area formula,

∫H∥∇ϱHv(z)∥H2 dAH=∫01∫{v(z)=t}∥∇ϱHv(z)∥H dsH dt.(11)\int_{\mathbb{H}}\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}^2\,dA_{\mathbb{H}}=\int_0^1\int_{\{v(z)=t\}}\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}\,ds_{\mathbb{H}}\,dt. \tag*{(11)}

We can now compute the right-hand side using the Steiner formula (9):

∫01∫{v(z)=t}∥∇ϱHv(z)∥H dsH dt=∫01t⋅ℓH({v(z)=t}) dt=∫01t⋅ℓH(∂Γ−log⁡t) dt=12∫01((1+t2)ℓH(∂Γ)+(1−t2)(2π+Area⁡H(Γ))) dt=23ℓH(∂Γ)+13(2π+Area⁡H(Γ))≤ℓH(γ)+2π3,\begin{aligned} \int_0^1\int_{\{v(z)=t\}}\|\nabla_{\varrho_{\mathbb{H}}}v(z)\|_{\mathbb{H}}\,ds_{\mathbb{H}}\,dt &=\int_0^1t\cdot\ell_{\mathbb{H}}(\{v(z)=t\})\,dt\\ &=\int_0^1t\cdot\ell_{\mathbb{H}}(\partial\Gamma_{-\log t})\,dt\\ &=\frac{1}{2}\int_0^1\left((1+t^2)\ell_{\mathbb{H}}(\partial\Gamma)+(1-t^2)(2\pi+\operatorname{Area}_{\mathbb{H}}(\Gamma))\right)\,dt\\ &=\frac{2}{3}\ell_{\mathbb{H}}(\partial\Gamma)+\frac{1}{3}(2\pi+\operatorname{Area}_{\mathbb{H}}(\Gamma))\leq\ell_{\mathbb{H}}(\gamma)+\frac{2\pi}{3}, \end{aligned}

where the last inequality follows by (8). Combining the last estimate with (10), we obtain

(mod⁡(H∖Y‾))−1≤ℓH(γ)+2π/3,\left(\operatorname{mod}\left(\mathbb{H}\setminus\overline{Y}\right)\right)^{-1}\leq\ell_{\mathbb{H}}(\gamma)+2\pi/3,

as required. □

Lemma 4. Let Y⋐XY\Subset X be a pair of open Jordan disks, and let hh be the equilibrium potential of (X,Y)(X,Y). Then, for t∈(0,1)t\in(0,1),

Area⁡X({h>t})≤1t⋅mod⁡(X∖Y‾).(12)\operatorname{Area}_{X}(\{h>t\})\leq\frac{1}{t\cdot\operatorname{mod}\left(X\setminus\overline{Y}\right)}. \tag*{(12)}

Proof. Recall that by (2) and (4), for every t∈(0,1)t\in(0,1) we have

1mod⁡(X∖Y‾)=∫h=t∣∇h(z)∣ ∣dz∣.\frac{1}{\operatorname{mod}\left(X\setminus\overline{Y}\right)}=\int_{h=t}|\nabla h(z)|\,|dz|.

Set a(t):=Area⁡X({h>t})a(t):=\operatorname{Area}_{X}(\{h>t\}). By the co-area formula,

a(t)=Area⁡X(Y)+∫{h>t}ϱX(z)2∣∇h(z)∣ ∣dz∣2=∫t1∫h=sϱX(z)2∣∇h(z)∣ ∣dz∣ dsa(t)=\operatorname{Area}_{X}(Y)+\int_{\{h>t\}}\frac{\varrho_{X}(z)^{2}}{|\nabla h(z)|}\,|dz|^{2}=\int_{t}^{1}\int_{h=s}\frac{\varrho_{X}(z)^{2}}{|\nabla h(z)|}\,|dz|\,ds
⟹a′(t)=−∫{h=t}ϱX(z)2∣∇h(z)∣ ∣dz∣.\Longrightarrow\quad a'(t)=-\int_{\{h=t\}}\frac{\varrho_{X}(z)^{2}}{|\nabla h(z)|}\,|dz|.

By the Cauchy–Schwarz inequality,

(ℓX({h=t}))2=(∫{h=t}ϱX(z)∣∇h(z)∣∣∇h(z)∣ ∣dz∣)2\left(\ell_{X}(\{h=t\})\right)^{2} =\left(\int_{\{h=t\}}\frac{\varrho_{X}(z)}{\sqrt{|\nabla h(z)|}}\sqrt{|\nabla h(z)|}\,|dz|\right)^{2}
≤∫{h=t}ϱX(z)2∣∇h(z)∣ ∣dz∣⋅∫{h=t}∣∇h(z)∣ ∣dz∣=−a′(t)mod⁡(X∖Y‾).\leq\int_{\{h=t\}}\frac{\varrho_{X}(z)^{2}}{|\nabla h(z)|}\,|dz|\cdot\int_{\{h=t\}}|\nabla h(z)|\,|dz| =-\frac{a'(t)}{\operatorname{mod}\left(X\setminus\overline{Y}\right)}.

On the other hand, since ∂{h>t}={h=t}\partial\{h>t\}=\{h=t\}, inequality (7) yields:

a2(t)=(Area⁡X({h>t}))2≤(ℓX({h=t}))2≤−a′(t)mod⁡(X∖Y‾).a^{2}(t)=\left(\operatorname{Area}_{X}(\{h>t\})\right)^{2}\leq\left(\ell_{X}(\{h=t\})\right)^{2}\leq-\frac{a'(t)}{\operatorname{mod}\left(X\setminus\overline{Y}\right)}.

This differential inequality is equivalent to (1/a(t))′≥mod⁡(X∖Y‾)(1/a(t))'\geq\operatorname{mod}\left(X\setminus\overline{Y}\right). Integrating this inequality from ss to tt, where 0<s<t0<s<t gives 1/a(t)≥1/a(s)+(t−s)mod⁡(X∖Y‾)≥(t−s)mod⁡(X∖Y‾)1/a(t)\geq1/a(s)+(t-s)\operatorname{mod}\left(X\setminus\overline{Y}\right)\geq(t-s)\operatorname{mod}\left(X\setminus\overline{Y}\right). Sending s→0s\to0 gives the required inequality 1/a(t)≥tmod⁡(X∖Y‾)1/a(t)\geq t\operatorname{mod}\left(X\setminus\overline{Y}\right). □

Proof of Theorem 1

We start by extending the geometric Lemma 2 to the dynamical context. This step is similar to [DPSU26, Section 2] (and the analogous steps in [PRW26, Ye26b]).

Lemma 5. In the notation of Theorem 1, if E⊆ΛE\subseteq\Lambda is a measurable set, then

Area⁡U(E)≤d⋅Area⁡V(f(E))+2πdD.\operatorname{Area}_{U}(E)\leq d\cdot\operatorname{Area}_{V}(f(E))+2\pi dD.

Proof. Let Q⊂VQ\subset V be the set of singular values of f:U→Vf:U\to V. Removing the finite set E∩f−1(Q)E\cap f^{-1}(Q) from EE does not change the area of EE and f(E)f(E), so we may assume that f(E)∩Q=∅f(E)\cap Q=\varnothing.

We have that f:U∖f−1(Q)→V∖Qf:U\setminus f^{-1}(Q)\to V\setminus Q is a covering, and hence a local isometry in the hyperbolic metric. Hence, combined with the domain monotonicity U∖f−1(Q)⊂UU\setminus f^{-1}(Q)\subset U, in local densities we have the following estimate:

ϱU≤ϱU∖f−1(Q)=∣f′∣⋅ϱV∖Q∘f.\varrho_{U}\leq\varrho_{U\setminus f^{-1}(Q)}=|f'|\cdot\varrho_{V\setminus Q}\circ f.

This way, we obtain the right-hand side of the estimate

0≤ϱU2−∣f′∣2⋅ϱV2∘f≤∣f′∣2⋅(ϱV∖Q2−ϱV2)∘f;(13)0 \le\varrho_U^2-|f'|^2\cdot\varrho_V^2\circ f\le|f'|^2\cdot(\varrho_{V\setminus Q}^2-\varrho_V^2)\circ f; \tag*{(13)}

the left-hand side follows by the Schwarz–Pick lemma applied to the holomorphic map f:U→Vf:U\to V. By Lemma 2 applied to the set B′∖QB'\setminus Q gives

0≤∫f(E)(ϱV∖Q2(z)−ϱV2(z))∣dz∣2≤2π⋅#Q≤2πD⟹0\le\int_{f(E)}\left(\varrho_{V\setminus Q}^2(z)-\varrho_V^2(z)\right)|dz|^2\le2\pi\cdot\#Q\le2\pi D \quad\Longrightarrow
⟹Area⁡V∖Q(f(E))≤Area⁡V(f(E))+2πD.(14)\Longrightarrow\operatorname{Area}_{V\setminus Q}(f(E))\le\operatorname{Area}_V(f(E))+2\pi D. \tag*{(14)}

The multiplicity of ff over the set EE, since EE lies in Λ′\Lambda', is at most dd. Therefore, by integrating (13) over EE we obtain:

Area⁡U(E)≤dArea⁡V(f(E))+d(Area⁡V∖Q(f(E))−Area⁡V(f(E)))≤dArea⁡V(f(E))+2πdD,\operatorname{Area}_U(E)\le d\operatorname{Area}_V(f(E))+d\left(\operatorname{Area}_{V\setminus Q}(f(E))-\operatorname{Area}_V(f(E))\right)\le d\operatorname{Area}_V(f(E))+2\pi dD,

where the last inequality follows from (14). This finishes the proof of the lemma. □

Proof of Theorem 1. Let vv and bb be the equilibrium potentials for pairs (V,B)(V,B) and (B′,B)(B',B) respectively. By the maximum principle, we have that

v(z)≥b(z)∀z∈B′.(15)v(z)\ge b(z)\qquad\forall z\in B'. \tag*{(15)}

Furthermore, define

u(z)={b∘f(z),z∈Λ′,0,z∈U∖Λ′(16)u(z)= \begin{cases} b\circ f(z), & z\in\Lambda',\\ 0, & z\in U\setminus\Lambda' \end{cases} \tag*{(16)}

This function has finite Dirichlet energy.

Define a set Ω:={1/2<u<1}⊂Λ′∖Λ\Omega:=\{1/2<u<1\}\subset\Lambda'\setminus\Lambda. Let us apply Lemma 5 to Ω\Omega. Note that f(Ω)⊂{b>1/2}⊂B′f(\Omega)\subset\{b>1/2\}\subset B'. By Lemma 5 followed by inequality (15) and an application of Lemma 4, we obtain:

Area⁡U(Ω)≤dArea⁡V({b>1/2})+2πdD≤dArea⁡V({v>1/2})+2πdD≤2dmod⁡(V∖B‾)+2πdD.(17)\begin{aligned} \operatorname{Area}_U(\Omega)\le d\operatorname{Area}_V(\{b>1/2\})+2\pi dD\le d\operatorname{Area}_V(\{v>1/2\})+2\pi dD \\ &\le\frac{2d}{\operatorname{mod}(V\setminus\overline{B})}+2\pi dD. \tag*{(17)} \end{aligned}

Observe that

∫Ω∣∇u(z)∣2∣dz∣2=d∫{1/2<b<1}∣∇b(z)∣2∣dz∣2.\int_{\Omega}|\nabla u(z)|^2|dz|^2=d\int_{\{1/2<b<1\}}|\nabla b(z)|^2|dz|^2.

On the other hand, by the constant flux property (3) and relation (4), we have

∫{1/2<b<1}∣∇b(z)∣2∣dz∣2=12∫B′∣∇b(z)∣2∣dz∣2=12mod⁡(B′∖B‾).\int_{\{1/2<b<1\}}|\nabla b(z)|^2|dz|^2=\frac{1}{2}\int_{B'}|\nabla b(z)|^2|dz|^2=\frac{1}{2\operatorname{mod}(B'\setminus\overline{B})}.

Hence,

∫Ω∣∇u(z)∣2∣dz∣2=d2mod⁡(B′∖B‾).(18)\int_{\Omega}|\nabla u(z)|^2|dz|^2=\frac{d}{2\operatorname{mod}(B'\setminus\overline{B})}. \tag*{(18)}

Finally, for every regular value t∈(1/2,1)t\in(1/2,1) of uu, the level set {u=t}\{u=t\} is a finite union of smooth Jordan curves, at least one of which, say γ\gamma, bounds a disk in UU that contains Λ‾\overline{\Lambda}. By Lemma 3 applied to Λ⋐U\Lambda\Subset U,

ℓU({u=t})≥ℓU(γ)≥1mod⁡(U∖Λ‾)−2π3.(19)\ell_U(\{u=t\})\ge\ell_U(\gamma)\ge\frac{1}{\operatorname{mod}(U\setminus\overline{\Lambda})}-\frac{2\pi}{3}. \tag*{(19)}

Note that there are only finitely many exceptional levels of uu as {z∈U:1/2<u(z)<1}⊆Λ′\{z \in U : 1/2 < u(z) < 1\} \subseteq\Lambda' and f∣Λ′f|_{\Lambda'} has finite degree. Therefore, by the co-area formula and the Cauchy–Schwarz inequality,

12(1mod⁡(U∖Λ‾)−2π3)≤∫1/21ℓU({u=t}) dt=∫1/21∫{u=t}ϱU(z)∣dz∣ dt=∫ΩϱU(z)∣∇u(z)∣ ∣dz∣2≤(Area⁡U(Ω)⋅∫Ω∣∇u(z)∣2 ∣dz∣2)1/2.(20)\begin{aligned} \frac{1}{2}\left(\frac{1}{\operatorname{mod}(U \setminus\overline{\Lambda})}-\frac{2\pi}{3}\right) \leq\int_{1/2}^{1} \ell_U(\{u=t\})\,dt = \int_{1/2}^{1}\int_{\{u=t\}} \varrho_U(z)|dz|\,dt \\ &= \int_{\Omega} \varrho_U(z)|\nabla u(z)|\,|dz|^2 \leq\left(\operatorname{Area}_U(\Omega)\cdot\int_{\Omega}|\nabla u(z)|^2\,|dz|^2\right)^{1/2}. \tag*{(20)} \end{aligned}

Suppose that the modulus U∖Λ‾U \setminus\overline{\Lambda} is degenerate so that

1mod⁡(U∖Λ‾)−2π3>12⋅1mod⁡(U∖Λ‾)⟺mod⁡(U∖Λ‾)<34π.\frac{1}{\operatorname{mod}(U \setminus\overline{\Lambda})}-\frac{2\pi}{3}>\frac{1}{2}\cdot\frac{1}{\operatorname{mod}(U \setminus\overline{\Lambda})} \quad\Longleftrightarrow\quad \operatorname{mod}(U \setminus\overline{\Lambda})<\frac{3}{4\pi}.

Then the left-hand side of (20) is positive, and thus by squaring (20) and combining with estimates (3.5 and (18), we obtain:

1mod⁡(U∖Λ‾)2≤16d2mod⁡(B′∖B‾)(1mod⁡(V∖B‾)+πD)(21)\frac{1}{\operatorname{mod}(U \setminus\overline{\Lambda})^2} \leq \frac{16d^2}{\operatorname{mod}(B' \setminus\overline{B})} \left(\frac{1}{\operatorname{mod}(V \setminus\overline{B})}+\pi D\right) \tag*{(21)}

In Theorem 1, it is enough to establish the second case. Under the collar assumption mod⁡(B′∖B‾)≥ηmod⁡(U∖Λ‾)\operatorname{mod}(B' \setminus\overline{B})\geq\eta\operatorname{mod}(U \setminus\overline{\Lambda}), the estimate (21) implies

1mod⁡(U∖Λ‾)≤16d2ηmod⁡(V∖B‾)+16πd2Dη.(22)\frac{1}{\operatorname{mod}(U \setminus\overline{\Lambda})} \leq \frac{16d^2}{\eta\operatorname{mod}(V \setminus\overline{B})} +\frac{16\pi d^2D}{\eta}. \tag*{(22)}

Now if U∖Λ‾U \setminus\overline{\Lambda} is further degenerate so that

1mod⁡(U∖Λ‾)>12mod⁡(U∖Λ‾)+16πd2Dη⟺mod⁡(U∖Λ‾)<η32πd2D,\frac{1}{\operatorname{mod}(U \setminus\overline{\Lambda})} > \frac{1}{2\operatorname{mod}(U \setminus\overline{\Lambda})} +\frac{16\pi d^2D}{\eta} \quad\Longleftrightarrow\quad \operatorname{mod}(U \setminus\overline{\Lambda})<\frac{\eta}{32\pi d^2D},

then from (22) we obtain the required bound

mod⁡(U∖Λ‾)>η32d2mod⁡(V∖B‾).\operatorname{mod}(U \setminus\overline{\Lambda}) > \frac{\eta}{32d^2}\operatorname{mod}(V \setminus\overline{B}).

This finishes the proof of the covering lemma. Our estimates start to work once the annulus U∖Λ‾U \setminus\overline{\Lambda} is ε\varepsilon-degenerate, where we can pick

ε:=η32πD3≤η32πd2D<34π.\varepsilon:= \frac{\eta}{32\pi D^3}\leq\frac{\eta}{32\pi d^2D}<\frac{3}{4\pi}.

Remark 7. If in the statement of Theorem 1 the map f:U→Vf: U \to V has a finite degree bounded by DD, then in the proof above we use the crude bound mod⁡(U∖Λ‾)≥1Dmod⁡(V∖B‾)\operatorname{mod}(U \setminus\overline{\Lambda})\geq\frac{1}{D}\operatorname{mod}(V \setminus\overline{B}) (see, e.g., [8 Lemma 4.5]) to pass the degeneration from U∖Λ‾U \setminus\overline{\Lambda} to V∖B‾V \setminus\overline{B} and this way bound the second bracket in (21) without the collar assumption. This gives the statement outlined in Remark 3.

References

Department de Matemàtiques i Informàtica, Universitat de Barcelona, Gran Via de les Corts Catalanes 585, 08007 Barcelona, Spain

Centre de Recerca Matemàtica, Edifici C, Carrer de l’Albareda, 08193 Bellaterra, Spain

Email address: [email protected]

References

  1. [1]Lars V. Ahlfors, Conformal invariants: topics in geometric function theory, McGraw-Hill Book Co., New York-Düsseldorf-Johannesburg, 1973, McGraw-Hill Series in Higher Mathematics.
  2. [4]Kostiantyn Drach and Dierk Schleicher, Rigidity of Newton dynamics, Adv. Math. 408 (2022), 108591.arxiv.org/abs/1812.11919
  3. [5]Dzmitry Dudko, On the MLC conjecture and the renormalization theory in complex dynamics, Proceedings of the International Congress of Mathematicians 2026, vol. 5, Society for Industrial and Applied Mathematics, 2026, pp. 116–135.
  4. [8]———, The Quasi-Additivity Law in conformal geometry, Ann. of Math. (2) 169 (2009), no. 2, 561–593.arxiv.org/abs/math/0505191
  5. [9]Oleg Kozlovski and Sebastian van Strien, Local connectivity and quasi-conformal rigidity of non-renormalizable polynomials, Proc. Lond. Math. Soc. (3) 99 (2009), no. 2, 275–296.DOI
  6. [10]Mikhail Lyubich, Conformal geometry and dynamics of quadratic polynomials, vol. I–II, Manuscript, https://www.math.stonybrook.edu/~mlyubich/book.pdf, 2026.
  7. [12]Pascale Roesch and Yongcheng Yin, Bounded critical Fatou components are Jordan domains for polynomials, Science China Mathematics 65 (2022), no. 2, 331–358.DOI
  8. [13]Pascale Roesch, Yongcheng Yin, and Jinsong Zeng, Rigidity of non-renormalizable Newton maps, Science China Mathematics 67 (2024), no. 4, 855–872.arxiv.org/abs/1811.09978
  9. [14]Luis A. Santaló, Integral geometry and geometric probability, 2nd ed., Cambridge Mathematical Library, Cambridge University Press, Cambridge, 2004.
  10. [15]D. Sullivan, Quasiconformal homeomorphisms and dynamics I: Solution of the Fatou–Julia problem on wandering domains, Ann. of Math. (2) 122 (1985), 401–418.DOI
  11. [16]Xiaoguang Wang, Yongcheng Yin, and Jinsong Zeng, Dynamics of Newton maps, Ergodic Theory and Dynamical Systems 43 (2023), no. 3, 1035–1080.arxiv.org/abs/1805.11478
  12. [17]Jonguk Yang, Local connectivity of polynomial Julia sets at bounded type Siegel boundaries, Preprint, https://arxiv.org/pdf/2010.14003v4, 2023.
  13. [18]Zihao Ye, Hyperbolic area methods in meromorphic dynamics and elliptic polynomial skew products, Preprint, https://arxiv.org/pdf/2609.30067v1, 2026.

Paper details

vlnc:W_hw9kwqvv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents