Capacity and harmonic measure of the ideal boundary

Intuition Beginner

Picture an infinite metal surface that runs off to the edge of the world — there is no rim you can touch, only a faraway "boundary at infinity." Pour heat into one spot in the middle. Two things can happen. Either the surface is so vast and open that the heat keeps leaking away forever, or the far-off boundary is so thin and pinched that the heat has nowhere to escape and pools up inside.

The first kind of surface has a fat boundary at infinity: charge or heat can run out to it. We call such a surface hyperbolic. The second kind has a boundary that is, for the purposes of heat flow, invisible — it has size zero. We call that surface parabolic.

How do we measure the "size" of a boundary we can never reach? We use two numbers. The first is the harmonic measure: it answers the question, "starting from an inside point, what fraction of the heat ends up escaping to infinity rather than returning to a fixed inner ring?" If that fraction is zero, nothing escapes. The second number is the capacity: it measures how much electric charge the boundary at infinity could hold. A boundary of capacity zero holds nothing and traps everything.

Why bother? Because these two numbers decide the entire character of the surface — whether it behaves like the cramped plane or like the wide-open disc — and they do it without ever looking at the unreachable edge directly.

Visual Beginner

Imagine a sequence of growing patches of the surface, like ripples spreading from a stone: a small inner ring stays fixed, and around it larger and larger contours expand outward toward infinity. On each patch you solve a temperature problem — set the temperature to $1$ on the current outer contour and to $0$ on the fixed inner ring, and let it settle. You get a temperature picture ${\omega }_n$ that grades smoothly from $0$ near the centre to $1$ at the moving edge.

As the outer ring marches off to infinity, the temperature pictures settle down to a limiting picture $\omega$. If that limit is flat zero everywhere, the boundary at infinity is invisible (parabolic). If the limit is a genuine positive temperature, the boundary is fat (hyperbolic).

Worked example Beginner

Take the open unit disc, with the centre as the fixed inner point and a small circle $|z|=1/4$ as the inner ring held at $0$. The "ideal boundary" is the unit circle $|z|=1$. Use the rings $|z|=r_n$ with $r_n\to 1$ as the expanding outer contours, each held at $1$.

Step 1. On the annulus between $|z|=1/4$ and $|z|=r_n$, the temperature that is $0$ on the inner ring and $1$ on the outer ring depends only on $|z|$, and the radial harmonic functions are built from $\log |z|$. So ${\omega }_n(z)=\frac{\log (|z|/(1/4))}{\log (r_n/(1/4))}=\frac{\log (4|z|)}{\log (4r_n)}$.

Step 2. Fix an interior point, say $|z|=1/2$. Then ${\omega }_n=\frac{\log 2}{\log (4r_n)}$.

Step 3. Let $r_n\to 1$. The denominator tends to $\log 4$, a finite positive number, so ${\omega }_n\to \frac{\log 2}{\log 4}=\frac{1}{2}$.

What this tells us: the limit is a positive number, not zero. Heat does escape to the unit circle. The disc is hyperbolic, and its ideal boundary is fat — exactly what we expect, since the unit circle is a real, reachable boundary with plenty of room to hold charge.

Check your understanding Beginner

Formal definition Intermediate+

Let $W$ be an open (noncompact) Riemann surface. A regular exhaustion is an increasing sequence of relatively compact subregions ${\Omega }_1⋐{\Omega }_2⋐\cdots$ with ${\bigcup }_n{\Omega }_n=W$, each ${\Omega }_n$ bounded by finitely many analytic Jordan curves, and $\overline{{\Omega }_n}\subset {\Omega }_{n+1}$. Fix a region ${\Omega }_0⋐{\Omega }_1$ with analytic boundary $\gamma =\partial {\Omega }_0$; this is the fixed inner contour. Write ${\beta }_n=\partial {\Omega }_n$ for the outer contour of the $n$-th piece. The ideal boundary of $W$ is the boundary $\partial W$ in the end compactification — the part of $\overline{W}$ approached as one leaves every ${\Omega }_n$.

For each $n$, let ${\omega }_n$ be the harmonic function on ${\Omega }_n\setminus \overline{{\Omega }_0}$ that is continuous up to the boundary with ${\omega }_n=0$ on $\gamma$ and ${\omega }_n=1$ on ${\beta }_n$. Each ${\omega }_n$ exists and is unique by the Dirichlet/Perron solution of the mixed boundary value problem on the region ${\Omega }_n\setminus \overline{{\Omega }_0}$ ^{[Ahlfors–Sario Ch. II §5]}, whose every boundary point is regular 06.01.24. The maximum principle gives $0\le {\omega }_n\le 1$ and ${\omega }_{n+1}\le {\omega }_n$ on the common domain, since ${\omega }_{n+1}\le 1={\omega }_n$ on ${\beta }_n$ and the two agree as $0$ on $\gamma$.

The harmonic measure of the ideal boundary (relative to $\gamma$) is the decreasing limit

$$\omega (z)=\underset{n\to \infty }{\lim }{\omega }_n(z),z\in W\setminus \overline{{\Omega }_0}.$$

The limit is harmonic by Harnack's theorem and independent of the choice of exhaustion. The surface $W$ is parabolic when $\omega \equiv 0$ and hyperbolic when $\omega \not\equiv 0$.

The capacity of the ideal boundary is defined through energy. For a charge (a unit mass distribution $\mu$) supported on the closed sets approaching $\partial W$, the logarithmic energy is $I(\mu )=\iint \log \frac{1}{|z-w|}d\mu (z)d\mu (w)$ measured in a fixed local parameter; the capacity is $\mathrm{cap}(\partial W)=e^{-V}$ with $V={\inf }_{\mu }I(\mu )$, and capacity zero means $V=+\infty$ ^{[Tsuji Ch. III]}.

Counterexamples to common slips

• Harmonic measure is not a single number — it is a function. The phrase "harmonic measure of the ideal boundary" names the harmonic function $\omega (z)$, whose value varies with the interior point $z$. It collapses to a number only after fixing $z$.
• Parabolic is not "small." The complex plane $\mathbb{C}$ is parabolic, yet it is the largest simply connected noncompact surface. Parabolicity is about the boundary at infinity being negligible, not about the surface being small.
• Capacity zero is not measure zero. A set can have two-dimensional area zero and still carry positive capacity; capacity is the finer, energy-based notion that governs whether harmonic functions can detect the set.

Key theorem with proof Intermediate+

Theorem (the parabolicity dichotomy). Let $W$ be an open Riemann surface with fixed inner contour $\gamma$. The following are equivalent:

(i) $W$ is parabolic, i.e. the harmonic measure $\omega \equiv 0$;

(ii) there is no Green's function $g(z,z_0)$ on $W$ with pole at $z_0$;

(iii) the ideal boundary of $W$ has capacity zero.

Proof. The harmonic measures ${\omega }_n$ decrease to $\omega \ge 0$, harmonic, with $\omega =0$ on $\gamma$. The total flux $c_n={\int }_{\gamma }\frac{\partial {\omega }_n}{\partial \nu }ds$ across $\gamma$ is a positive number — the conductance of the ring region — and decreases with $n$, to a limit $c\ge 0$.

(i) $\iff$ (ii). Build the Green's function as the increasing limit of $g_n$, where $g_n$ is harmonic on ${\Omega }_n\setminus \{z_0\}$ with the logarithmic pole at $z_0$ and $g_n=0$ on ${\beta }_n$. The maximum principle gives $g_n\le g_{n+1}$, so $g_n$ increases to a limit $g$, finite (a genuine Green's function) precisely when the $g_n$ stay bounded. Comparison of $g_n$ against a multiple of $1-{\omega }_n$ near $\gamma$ shows the $g_n$ are uniformly bounded if and only if $c>0$, i.e. if and only if $\omega \not\equiv 0$. Thus $\omega \equiv 0$ removes the Green's function and $\omega \not\equiv 0$ produces it.

(ii) $\iff$ (iii). The conductance $c$ equals (up to the fixed normalisation of $\gamma$) the reciprocal of the modulus of the ring separating $\gamma$ from $\partial W$; that modulus is $+\infty$ exactly when no positive-energy mass distribution sits on the ideal boundary. Energy infimum $V=+\infty$ is the statement $\mathrm{cap}(\partial W)=0$, and it coincides with $c=0$, which by the previous paragraph is $\omega \equiv 0$ and the absence of the Green's function. $\square$

Bridge. This theorem builds toward the entire classification theory of open surfaces, and it appears again in 06.11.06 where the conductance $c$ is recast as the reciprocal of an extremal length. The foundational reason the three conditions coincide is that each measures the same flux: harmonic measure measures it as escaping heat, the Green's function measures it as the potential of a unit pole, and capacity measures it as storable charge. This is exactly the trichotomy that the type problem 06.11.03 turns on, and putting these together identifies parabolicity with a single vanishing — the conductance $c=0$ — from which the harmonic-measure, Green's-function, and capacity statements all follow. The bridge is the conductance $c$, which generalises the elementary annulus modulus to an arbitrary ideal boundary, and the central insight is that an unreachable boundary is detected only through the flux it admits.

Exercises Intermediate+

Advanced results Master

Harmonic measure as a measure. Beyond the harmonic function $\omega (z)$, harmonic measure is genuinely a family of Borel measures $\{{\omega }_z{\}}_{z\in W}$ on the Royden or Wiener boundary. For a resolutive boundary function $f$, the Perron solution is $H_f(z)=\int fd{\omega }_z$, and the function $\omega (z)$ of the dichotomy is $H_1$ relative to the splitting that holds the inner contour at $0$. Parabolicity is the statement that the total mass of ${\omega }_z$ on the ideal boundary is $0$: the boundary is $\omega$-null. This realises the type problem inside the classification scheme of Sario and Nakai, where the null-classes $O_G\subset O_{HB}\subset O_{HD}$ rank surfaces by which spaces of harmonic functions degenerate ^{[Sario–Nakai Ch. I–III]}; $O_G$ is exactly the parabolic class.

Capacity as an extremal problem. The energy definition $V={\inf }_{\mu }I(\mu )$ has the dual extremal-length face. Writing $\mathrm{cap}(\partial W)=0$ as $V=+\infty$ matches the divergence of the modulus of the family of curves separating $\gamma$ from the ideal boundary. The equilibrium measure ${\mu }^{\star }$, when it exists (capacity positive), is the harmonic measure of the boundary seen from infinity, and its potential is the Green's function with pole there. Nevanlinna's mass-distribution method ^{[Nevanlinna Ch. V]} computes capacity as a transfinite diameter, and the energy infimum is attained by a unique equilibrium distribution whenever the capacity is positive.

The conductance and the flux. The number $c={\int }_{\gamma }{\partial }_{\nu }\omega ds$ is the harmonic-measure flux and equals the reciprocal of the extremal distance from $\gamma$ to the ideal boundary. The dichotomy theorem is the statement $c=0⟺\mathrm{cap}(\partial W)=0⟺\omega \equiv 0⟺$ no Green's function. On a hyperbolic surface, $c$ normalises the Green's function: $g(\cdot ,z_0)$ has flux $2\pi$ around its pole and flux $c$-controlled decay toward the ideal boundary.

Stability and removability. Capacity zero is stable under countable unions of capacity-zero sets and is exactly the class of removable sets for bounded harmonic functions: a compact $E$ of capacity zero is removable, so deleting it does not change the type. This is the surface-theoretic upgrade of the classical fact that polar sets are removable for the Dirichlet problem 06.01.24, and it is why parabolicity is a property of the ideal boundary modulo capacity-zero alterations.

Synthesis. The conductance $c$ is the central insight that unifies this unit: it is dual to the extremal length of the separating curve family, it is exactly the energy reciprocal that defines capacity, and it generalises the elementary annulus modulus $1/\log (1/\rho )$ to an arbitrary ideal boundary. Putting these together, the foundational reason parabolic and hyperbolic split the open surfaces cleanly is that each of harmonic measure, the Green's function, and capacity reads off the single number $c$ — heat that escapes, potential of a pole, charge that the boundary stores — and the bridge is that all three vanish together. This is exactly the structure that the type problem 06.11.03 exploits and that extremal length 06.11.06 makes geometric, and it generalises forward to the null-class hierarchy $O_G\subset O_{HB}\subset O_{HD}$, where parabolicity ($O_G$) is the coarsest degeneration and finer capacity invariants distinguish the rest.

Full proof set Master

Proposition (Conductance is exhaustion-independent and monotone). Let $c_n={\int }_{\gamma }{\partial }_{\nu }{\omega }_{n}ds$. Then $c_n$ is positive, decreasing in $n$, and its limit $c={\int }_{\gamma }{\partial }_{\nu }\omega ds$ depends only on $W$ and $\gamma$, not on the exhaustion.

Proof. On the region $A_n={\Omega }_n\setminus \overline{{\Omega }_0}$ the harmonic measure ${\omega }_n$ is positive in the interior, $0$ on $\gamma$, $1$ on ${\beta }_n$, so by the Hopf boundary-point principle ${\partial }_{\nu }{\omega }_n>0$ along $\gamma$ (outward normal), giving $c_n>0$. Green's identity applied to ${\omega }_n$ and the constant $1$ on $A_n$ shows ${\int }_{\gamma }{\partial }_{\nu }{\omega }_{n}ds={\int }_{{\beta }_n}{\partial }_{\nu }{\omega }_{n}ds$: the flux is conserved across the ring, so $c_n$ is the Dirichlet integral ${\iint }_{A_n}|\nabla {\omega }_n{|}^2$ up to normalisation. Since ${\omega }_{n+1}\le {\omega }_n$ with equal boundary values on $\gamma$, the Dirichlet principle (the harmonic competitor minimises Dirichlet energy among functions with the same boundary data) gives ${\iint }_{A_n}|\nabla {\omega }_{n+1}{|}^2\le {\iint }_{A_n}|\nabla {\omega }_n{|}^2$, hence $c_{n+1}\le c_n$. Monotone and bounded below by $0$, $c_n\to c\ge 0$. Independence of the exhaustion follows from the corresponding independence of $\omega$ (Exercise 4): two exhaustions yield the same limit $\omega$, hence the same $\gamma$-flux $c$. $\square$

Proposition (Capacity zero is closed under countable unions). If $E={\bigcup }_k{E}_k$ with each $E_k$ compact of logarithmic capacity zero, then $E$ has capacity zero.

Proof. Capacity zero is equivalent to: every probability measure supported on the set has infinite logarithmic energy, equivalently the set is polar — there is a superharmonic function $s\not\equiv +\infty$ equal to $+\infty$ on the set. For each $k$, capacity zero gives a superharmonic $s_k\ge 0$ on a fixed disc $D\supset E$ with $s_k=+\infty$ on $E_k$ and $s_k(z_0)<2^{-k}$ at a fixed reference point $z_0\in D\setminus E$. Set $s={\sum }_k{s}_k$. Each partial sum is superharmonic and increasing; the limit $s$ is superharmonic by the monotone-limit theorem for superharmonic functions, and $s(z_0)\le {\sum }_k{2}^{-k}=1<+\infty$, so $s\not\equiv +\infty$. But $s=+\infty$ on every $E_k$, hence on $E$. Therefore $E$ is polar, i.e. $\mathrm{cap}(E)=0$. $\square$

Connections Master

• The type problem and the Green's-function construction of 06.11.03 supply condition (ii) of the dichotomy. There the existence of the Green's function $g(z,z_0)$ as the increasing limit of $g_n$ is the defining test; this unit reads that same test off the harmonic-measure side and the capacity side, so 06.11.03 and this unit are two faces of one classification, with the conductance $c$ as the shared invariant.

• The exhaustion machinery and the very notion of ideal boundary come from 06.11.01; the regular exhaustion ${\Omega }_1⋐{\Omega }_2⋐\cdots$ and the end compactification defined there are the substrate on which the harmonic measures ${\omega }_n$ live. Without the exhaustion theory there is no monotone limit and no $\omega$.

• The capacity reciprocal $c$ is geometrised in 06.11.06 as the extremal length / modulus of the family of curves separating $\gamma$ from the ideal boundary: capacity zero is infinite modulus. The energy-extremal characterisation of capacity given here is the analytic shadow of the conformal-invariant extremal length defined there.

• The local solvability of the mixed boundary value problem for each ${\omega }_n$, and the barrier/regular-point theory that makes capacity-zero sets removable, are exactly the Dirichlet/Perron apparatus of 06.01.24; the plane Wiener capacity and polar-set removability proved there are the local model that this unit globalises to the ideal boundary.

• The classification null-classes meet the boundary-value/uniformization stream at 06.11.04, where parabolicity controls which uniformizing maps and which spaces of analytic functions survive on the open surface; the $O_G$ membership established here is the entry point to that finer hierarchy.

Historical & philosophical context Master

The harmonic-measure and capacity description of the ideal boundary grew out of Nevanlinna's potential-theoretic recasting of value-distribution theory. Nevanlinna ^{[Nevanlinna 1953]}, in Eindeutige analytische Funktionen (2nd ed., 1953), organised the theory of harmonic measure on noncompact regions and made the mass-distribution / capacity method central to deciding whether a surface admits nonconstant bounded harmonic functions. Ahlfors and Sario ^{[Ahlfors–Sario 1960]}, in Riemann Surfaces (1960), gave the systematic exhaustion treatment used here, fixing the harmonic measure of the ideal boundary as the monotone limit of solutions to the mixed problem on each ${\Omega }_n$ and proving the parabolic/hyperbolic dichotomy through the conductance. Sario and Nakai ^{[Sario–Nakai 1970]} then erected the full classification theory, ordering the null-classes $O_G\subset O_{HB}\subset O_{HD}\subset O_{AD}$ and placing parabolicity as the class $O_G$. The energy and transfinite-diameter formulation of capacity is due to Frostman (1935) and was synthesised for function theory by Tsuji ^{[Tsuji 1959]} in Potential Theory in Modern Function Theory (1959).

Bibliography Master

@book{AhlforsSario1960,
  author = {Ahlfors, Lars V. and Sario, Leo},
  title = {Riemann Surfaces},
  publisher = {Princeton University Press},
  year = {1960},
  series = {Princeton Mathematical Series 26},
  note = {Ch. II §5: harmonic measure of the ideal boundary, capacity, parabolic/hyperbolic classification}
}

@book{Nevanlinna1953,
  author = {Nevanlinna, Rolf},
  title = {Eindeutige analytische Funktionen},
  publisher = {Springer},
  year = {1953},
  edition = {2nd},
  series = {Grundlehren der mathematischen Wissenschaften 46},
  note = {Ch. V: harmonic measure, mass distributions, capacity}
}

@book{SarioNakai1970,
  author = {Sario, Leo and Nakai, Mitsuru},
  title = {Classification Theory of Riemann Surfaces},
  publisher = {Springer},
  year = {1970},
  series = {Grundlehren der mathematischen Wissenschaften 164},
  note = {Null-classes O_G, O_HB, O_HD; capacity of the ideal boundary; parabolic = O_G}
}

@book{Tsuji1959,
  author = {Tsuji, Masatsugu},
  title = {Potential Theory in Modern Function Theory},
  publisher = {Maruzen},
  year = {1959},
  note = {Ch. III: logarithmic capacity, energy, transfinite diameter, equilibrium measure}
}

@article{Frostman1935,
  author = {Frostman, Otto},
  title = {Potentiel d'\'equilibre et capacit\'e des ensembles avec quelques applications \`a la th\'eorie des fonctions},
  journal = {Meddelanden Lunds Universitets Matematiska Seminarium},
  volume = {3},
  year = {1935},
  pages = {1--118},
  note = {Equilibrium potential and capacity; energy-minimising equilibrium measure}
}