06.11.04 · riemann-surfaces / open-surfaces

Null-classes O_G, O_HB, O_HD, O_AD and the classification of open surfaces

shipped3 tiersLean: nonepending prereqs

Anchor (Master): Ahlfors-Sario *Riemann Surfaces* (Princeton 1960) Ch. II §6 (originator monograph for the null-class lattice); Royden 1952 *Some remarks on open Riemann surfaces* (originator separating example); Sario-Nakai *Classification Theory of Riemann Surfaces* (Springer Grundlehren 164, 1970); Tsuji *Potential Theory in Modern Function Theory* Ch. III-V

Intuition Beginner

Some surfaces are roomy and some are thin. On a roomy surface you can write down many well-behaved functions that vary from place to place: temperatures that settle into a steady pattern, smooth flows that carry finite energy. On a surface that is too thin, the only such well-behaved functions are the constant ones — the surface refuses to carry any variation at all. Sorting open surfaces by exactly which kinds of nice functions they refuse to carry is what the classification theory does.

The everyday picture is an infinite sheet of metal. Heat the sheet so that the temperature stays between zero and one hundred degrees everywhere and never runs away to infinity. On a thin sheet, equilibrium forces the temperature to be the same everywhere; the sheet cannot hold a bounded temperature pattern that changes. On a fatter sheet it can. A bounded steady temperature is a bounded harmonic function, so this is a test that some surfaces pass and some fail.

There are several such tests, each using a different family of functions: bounded harmonic ones, finite-energy harmonic ones, bounded holomorphic ones. A surface that fails one test is called null for that family. The tests are not independent: failing a strong test forces failing the weaker ones, which lines the families up into a ladder of nullness. Climbing that ladder is the way to classify open surfaces.

Visual Beginner

A Venn-style nesting diagram. The innermost region is labelled $O_{G}$ (parabolic surfaces, the thinnest), sitting inside a larger region $O_{H B}$ , sitting inside the largest region $O_{H D}$ . A parallel chain $O_{G} \subset O_{A B} \subset O_{A D}$ runs alongside, with a dotted bridge marking that the top boxes $O_{H D}$ and $O_{A D}$ coincide while the middle boxes $O_{H B}$ and $O_{A B}$ do not. Sample surfaces are dropped into the gaps between rings: the plane in $O_{G}$ , and Royden's and Toki's surfaces sitting in the thin shells that show the rings are genuinely different.

The picture is the whole subject in one frame: a hierarchy of nested boxes, each box a family of surfaces too thin to carry a particular kind of function, and named separating surfaces living in the shells that keep the boxes apart.

Worked example Beginner

Take the simplest open surface, the whole complex plane (think of it as the flat infinite sheet). Ask the first test: does it carry a bounded steady temperature that varies?

Pick any harmonic function on the plane that stays between two fixed bounds, say between $- 1$ and $1$ . Liouville's theorem from one-variable function theory says a harmonic function bounded on the entire plane must be constant. So the plane carries no varying bounded harmonic function: it passes into the box $O_{H B}$ .

Now ask whether the plane even carries a Green's function — a steady temperature with a single hot spike at one point that decays toward zero far away. On the plane it does not: the natural candidate $lo g (1/∣ z ∣)$ grows without bound instead of decaying, so no Green's function exists. A surface with no Green's function is called parabolic, and these form the smallest box $O_{G}$ . The plane is parabolic, so it lands in $O_{G}$ .

Finally test bounded holomorphic functions. A holomorphic function bounded on the whole plane is constant, again by Liouville. So the plane is in $O_{A B}$ as well.

What this tells us: the plane fails every test at once and sits in the innermost box. It is the model of a maximally thin open surface. The interesting surfaces are the ones that fail some tests but pass others, and those are what force the boxes to be genuinely different sizes.

Check your understanding Beginner

Formal definition Intermediate+

Let $W$ be an open (connected, non-compact) Riemann surface. Five families of functions on $W$ organise the classification, following Ahlfors-Sario Ch. II §6 ^{[Ahlfors-Sario 1960]}. Write $H B (W)$ for the bounded harmonic functions, $H D (W)$ for the harmonic functions of finite Dirichlet integral

D_{W} (u) = \int_{W} d u \land * d u = \int_{W} (u_{x}^{2} + u_{y}^{2}) d x d y < \infty,

$A B (W)$ for the bounded holomorphic functions, and $A D (W)$ for the holomorphic functions with $D_{W} (Re f) + D_{W} (Im f) < \infty$ . The Hodge star $*$ on a Riemann surface sends $a d x + b d y$ to $- b d x + a d y$ ; it is conformally invariant in real dimension two, so the Dirichlet integral depends only on the conformal (Riemann-surface) structure of $W$ , not on a metric.

To each family $X Y \in {H B, H D, A B, A D}$ attach a null-class $O_{X Y}$ :

W \in O_{X Y} ⟺ every f \in X Y (W) is constant.

Separately, $W \in O_{G}$ means $W$ carries no Green's function: there is no positive harmonic function $g_{W} (\cdot, p)$ on $W ∖ {p}$ with a logarithmic pole $g_{W} (z, p) = - lo g ∣ z - p ∣ + O (1)$ at $p$ that is the smallest such positive superharmonic function. A surface in $O_{G}$ is parabolic; one not in $O_{G}$ is hyperbolic. The defining data of $O_{G}$ — the Green's function, harmonic measure, and capacity — are developed in 06.11.03; the Royden compactification used below is built in 06.11.05.

The Dirichlet integral is a conformal invariant, $H D \subset H B$ fails in general (a Dirichlet-finite harmonic function need not be bounded, and conversely), and $A B \subset H B$ , $A D \subset H D$ hold by taking real parts. The notation $O_{X Y}$ is read "the surfaces null for class $X Y$ "; smaller class of functions yields larger null-class, because there is less to force constant.

Counterexamples to common slips

A surface in $O_{H B}$ need not be in $O_{G}$ . Parabolicity (no Green's function) is strictly stronger than the absence of non-constant bounded harmonic functions: there are hyperbolic surfaces (Green's function exists) on which every bounded harmonic function is still constant.
$H D$ is not contained in $H B$ . The function $u (z) = Re lo g z = lo g ∣ z ∣$ on an annulus-type end has finite Dirichlet integral over a suitable end yet is unbounded; Dirichlet-finiteness and boundedness are independent constraints on a harmonic function.
The Dirichlet integral $D_{W} (u)$ uses the conformal structure only; rescaling the metric does not change it. A common slip is to treat $D_{W}$ as metric-dependent and conclude null-class membership changes under quasiconformal deformation — it does not change under conformal change, though quasiconformal maps can move a surface between classes.

Key theorem with proof Intermediate+

Theorem (the null-class inclusion lattice; Ahlfors-Sario 1960). For open Riemann surfaces the strict inclusions

O_{G} \subset O_{H B} \subset O_{H D}, O_{G} \subset O_{A B} \subset O_{A D},

hold, together with the identification $O_{H D} = O_{A D}$ , while $O_{H B} \neq = O_{A B}$ . Every inclusion displayed is proper.

Proof. The inclusions split into elementary containments and the harder properness and identification statements.

Containments. Suppose $W \in O_{G}$ , so $W$ is parabolic. On a parabolic surface every positive superharmonic function is constant (the maximum principle holds up to the ideal boundary because harmonic measure of the ideal boundary vanishes — this is the content of 06.11.03). A bounded harmonic $u$ has $u - in f u \geq 0$ harmonic and bounded, hence superharmonic and bounded, hence constant; so $u$ is constant and $W \in O_{H B}$ . This gives $O_{G} \subset O_{H B}$ . For $O_{H B} \subset O_{H D}$ , observe that a Dirichlet-finite harmonic function on $W$ has, by the Royden compactification 06.11.05, boundary values on the harmonic boundary, and the harmonic boundary supports the same Poisson representation that controls bounded harmonic functions; concretely, if every bounded harmonic function is constant then the harmonic boundary is a single point, and a Dirichlet-finite harmonic function with one-point harmonic boundary is constant. Hence $W \in O_{H B} \Rightarrow W \in O_{H D}$ . For the analytic chain, the function-class containments $A B \subset H B$ and $A D \subset H D$ give the null-class inclusions $O_{H B} \subset O_{A B}$ and $O_{H D} \subset O_{A D}$ by monotonicity (Proposition 1 below), while $O_{G} \subset O_{A B} \subset O_{A D}$ follows from the same parabolic argument applied to $Re f$ and $Im f$ for holomorphic $f$ .

Identification $O_{H D} = O_{A D}$ . The inclusion $O_{H D} \subset O_{A D}$ is the containment just proved. For the reverse, let $W \in O_{A D}$ and let $u \in H D (W)$ . The conjugate differential $* d u$ is closed because $u$ is harmonic ( $d * d u = Δ u d V = 0$ ); on a surface with no $A D$ -obstruction the period class of $* d u$ over the homology of $W$ vanishes after passing to the Dirichlet-finite harmonic boundary, so $* d u = d v$ for a single-valued harmonic conjugate $v$ with $D_{W} (v) = D_{W} (u) < \infty$ . Then $f = u + i v$ is holomorphic with finite Dirichlet integral, so $f \in A D (W)$ , hence $f$ is constant by hypothesis, hence $u$ is constant. Thus $O_{A D} \subset O_{H D}$ , and equality holds. The harmonic conjugate exists because Dirichlet-finiteness kills the periods: this is exactly where the finiteness of $D_{W} (u)$ is load-bearing.

Non-identification $O_{H B} \neq = O_{A B}$ . Boundedness does not pass to the harmonic conjugate. A bounded harmonic $u$ can have an unbounded conjugate $v$ , so $u + i v$ need not lie in $A B$ , and the conjugation argument breaks. Ahlfors-Sario exhibit a hyperbolic surface carrying a non-constant bounded harmonic function (so $W \in / O_{H B}$ ) on which every bounded holomorphic function is constant (so $W \in O_{A B}$ ); such a surface witnesses $O_{H B} ⊊ O_{A B}$ strictly, hence $O_{H B} \neq = O_{A B}$ .

Properness of the chain. $O_{G} ⊊ O_{H B}$ : there exist hyperbolic surfaces with only constant bounded harmonic functions, so they lie in $O_{H B} ∖ O_{G}$ . $O_{H B} ⊊ O_{H D}$ is Toki's surface ^{[Toki 1952]}: a surface admitting a non-constant bounded harmonic function (not in $O_{H B}$ ) but no non-constant Dirichlet-finite harmonic function (in $O_{H D}$ ) — built from a chain of bordered pieces whose moduli grow so that bounded functions survive while the Dirichlet integral of any varying harmonic function diverges. Royden's surface ^{[Royden 1952]} separates the analytic chain. Each separating example is a concrete plane domain or covering surface with controlled end geometry. $□$

Bridge. This theorem is the apex the chapter builds toward, and the lattice it proves appears again in every later statement about open-surface type. The foundational reason the boxes nest is monotonicity of null-classes in the function family: a smaller family of test functions is harder to make non-constant, so its null-class is larger, and this is exactly the order-reversal that turns the inclusion chain of function spaces $A D \supset A B$ , $A D \supset H D \supset H B$ into the inclusion chain of surfaces. The identification $O_{H D} = O_{A D}$ generalises the compact-surface fact that harmonic and holomorphic data are interchangeable once periods vanish, and is dual to the Hodge-type pairing on a curve: the harmonic conjugate that exists here for Dirichlet-finite $u$ is the open-surface remnant of the Hodge decomposition of 06.04.03, where on a compact surface every harmonic form splits into holomorphic and anti-holomorphic parts with no boundary obstruction. The central insight is that on an open surface the obstruction to that splitting is precisely a period over the ideal boundary, and Dirichlet-finiteness is the integrability condition that makes the period vanish. Putting these together, the parabolic class $O_{G}$ from 06.11.03 sits at the bottom because parabolicity forces every positive superharmonic function to be constant, and the Royden boundary of 06.11.05 is the space on which the surviving non-constant members of $H B$ and $H D$ acquire their boundary values; the bridge is harmonic measure, which is total on a hyperbolic surface and null on a parabolic one.

Exercises Intermediate+

Exercise 4 (medium, short-answer).

Prove that the inclusion $O_{G} \subset O_{A B}$ holds: a parabolic surface carries no non-constant bounded holomorphic function.

Hint

If $f$ is bounded holomorphic, $lo g ∣ f ∣$ is subharmonic and bounded above; use that on a parabolic surface a bounded-above subharmonic function is constant.

Answer

Let $W \in O_{G}$ and $f \in A B (W)$ with $∣ f ∣ \leq M$ . Either $f$ has a zero, or $lo g ∣ f ∣$ is harmonic; in general $lo g ∣ f ∣$ is subharmonic and bounded above by $lo g M$ . On a parabolic surface every subharmonic function bounded above is constant (the maximum principle holds up to the null ideal boundary). Hence $∣ f ∣$ is constant. A holomorphic function of constant modulus on a connected surface is constant (open mapping theorem). Therefore $f$ is constant and $W \in O_{A B}$ . Rubric: full credit for the subharmonicity-of- $lo g ∣ f ∣$ argument plus the constant-modulus step.

Exercise 5 (medium, short-answer).

Explain why $O_{H D} \subset O_{A D}$ but the reverse needs the period-vanishing argument, and identify precisely the step where Dirichlet-finiteness is used.

Hint

A holomorphic function is determined by a harmonic real part plus a harmonic conjugate; the conjugate is single-valued only when certain periods vanish.

Answer

$A D \subset H D$ as function classes (real part of a Dirichlet-finite holomorphic function is Dirichlet-finite harmonic), so $O_{H D} \subset O_{A D}$ directly. For the reverse, given $u \in H D$ one forms $f = u + i v$ with $* d u = d v$ ; the conjugate $v$ is single-valued iff the periods $\oint_{γ} * d u$ vanish over a homology basis. Dirichlet-finiteness is used exactly here: $D_{W} (u) < \infty$ forces $* d u$ to be a Dirichlet-finite harmonic differential, whose periods over the Royden harmonic boundary vanish, giving a single-valued $v$ with $D_{W} (v) = D_{W} (u) < \infty$ . Then $f \in A D$ , so $f$ is constant and $u$ is constant. Rubric: full credit for naming the period-vanishing step as the place Dirichlet-finiteness enters.

Exercise 6 (hard, short-answer).

Sketch how a Toki-type surface separates $O_{H B}$ from $O_{H D}$ : describe the end geometry that lets a non-constant bounded harmonic function survive while forcing every non-constant harmonic function to have infinite Dirichlet integral.

Hint

Build the surface as an infinite chain of bordered pieces joined along thinning necks; bounded functions only need the maximum principle, but Dirichlet energy accumulates across the necks.

Answer

Take a sequence of bordered surfaces (e.g. discs with two boundary slits) glued along a chain of connecting tubes whose conformal moduli are chosen so that harmonic measure of the ideal boundary is positive (the surface is hyperbolic and carries a non-constant bounded harmonic function, so $W \in / O_{H B}$ ), yet the tubes are made conformally long enough that any harmonic function varying from one end to the other accumulates Dirichlet energy $\sum_{n} D_{n} = \infty$ across the necks. Then no non-constant harmonic function has finite Dirichlet integral, so $W \in O_{H D}$ . The separation rests on the independence of the boundedness constraint (a maximum-principle, harmonic-measure condition) from the Dirichlet-energy constraint (an $L^{2}$ -of-the-gradient condition). Rubric: full credit for the thinning-neck construction and the contrast between harmonic-measure positivity and Dirichlet-energy divergence.

Exercise 7 (hard, short-answer).

Prove that $O_{H B} \neq = O_{A B}$ by explaining why the conjugation argument that gives $O_{H D} = O_{A D}$ fails for the bounded classes.

Hint

Boundedness of $u$ does not imply boundedness of its harmonic conjugate $v$ .

Answer

The equality $O_{H D} = O_{A D}$ used that the harmonic conjugate of a Dirichlet-finite $u$ is again Dirichlet-finite, because $D_{W} (v) = D_{W} (u)$ . For bounded classes there is no analogous transfer: a bounded harmonic $u$ can have an unbounded conjugate $v$ (already on a half-plane, $Re lo g z$ is bounded on an angular sector where $Im lo g z = ar g z$ is bounded, but on ends where the argument winds the conjugate is unbounded), so $f = u + i v$ need not be bounded and the implication $u \in H B \Rightarrow f \in A B$ fails. Ahlfors-Sario construct a hyperbolic surface with a non-constant bounded harmonic function but only constant bounded holomorphic functions, placing it in $O_{A B} ∖ O_{H B}$ . Hence $O_{H B} ⊊ O_{A B}$ . Rubric: full credit for the conjugate-unboundedness explanation and the existence of the separating surface.

Lean formalization Intermediate+

Mathlib has no theory of open Riemann surfaces, harmonic-measure-based parabolicity, the Dirichlet-finite function spaces, or the null-class predicates. A proposed signature, in Lean 4 / Mathlib syntax, sketching the target lattice statement:

-- Sketch only; no current Mathlib coverage. See lean_mathlib_gap.
import Mathlib.Geometry.Manifold.ComplexManifold

namespace Codex.RiemannSurfaces.OpenSurfaces

variable (W : Type*) [OpenRiemannSurface W]

-- Null-class predicates: every member of the named function class is constant.
def IsNullG  : Prop := ¬ ∃ p : W, HasGreenFunction W p
def IsNullHB : Prop := ∀ u : HarmonicFun W, u.Bounded → u.IsConstant
def IsNullHD : Prop := ∀ u : HarmonicFun W, u.DirichletFinite → u.IsConstant
def IsNullAB : Prop := ∀ f : HolomorphicFun W, f.Bounded → f.IsConstant
def IsNullAD : Prop := ∀ f : HolomorphicFun W, f.DirichletFinite → f.IsConstant

-- The inclusion lattice and the HD = AD identification.
theorem nullClass_lattice :
    (IsNullG W → IsNullHB W)
  ∧ (IsNullHB W → IsNullHD W)
  ∧ (IsNullG W → IsNullAB W)
  ∧ (IsNullAB W → IsNullAD W)
  ∧ (IsNullHD W ↔ IsNullAD W) := by
  sorry

end Codex.RiemannSurfaces.OpenSurfaces

The proof depends on names absent from Mathlib: the Green's function and its existence dichotomy, harmonic measure of the ideal boundary, the Dirichlet integral as a semi-norm on harmonic functions, the harmonic conjugate with controlled periods on an open surface, and the Royden compactification. Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results Master

The lattice of [the present section's theorem] is the spine of the Ahlfors-Sario classification, but the subject extends in several directions, each refining the question of how thin a surface is by widening the family of test functions or sharpening the boundary on which they live.

The full Sario-Nakai lattice. Beyond $H B, H D, A B, A D$ the classification theory of Sario-Nakai 1970 ^{[Sario-Nakai 1970]} adds the classes $H P$ (positive harmonic functions), $H B D = H B \cap H D$ , and the quasi-bounded / singular decomposition of positive harmonic functions. The refined lattice reads

O_{G} \subset O_{H P} \subset O_{H B} \subset O_{H B D} \subset O_{H D},

with $O_{G} = O_{H P}$ failing precisely when a surface carries a non-constant positive harmonic function but no Green's function — which cannot happen, giving in fact $O_{G} = O_{H P}$ , the statement that parabolicity is equivalent to the absence of non-constant positive harmonic functions. This is the Myrberg-Heins characterisation: $W$ is parabolic iff every positive superharmonic function is constant iff $W \in / O_{H P}^{c}$ . The class $O_{H B D}$ wedged between $O_{H B}$ and $O_{H D}$ is proper at both ends, established by surfaces interpolating Toki's and Royden's constructions.

Harmonic boundary and the Royden compactification. The structural explanation for the whole lattice is the Royden compactification $W^{*} = W \cup Γ$ developed in 06.11.05, where $Γ$ is the Royden boundary and $Δ \subset Γ$ the harmonic boundary — the support of harmonic measure. The Dirichlet-finite harmonic functions are exactly the harmonic functions with continuous boundary values on $Γ$ that are Dirichlet integrals of their boundary data, and the bounded harmonic functions are recovered by Poisson integration against harmonic measure on $Δ$ . In this language:

W \in O_{H B} ⟺ Δ is a single point, W \in O_{G} ⟺ Δ = \emptyset .

The lattice inclusions become statements about the size and structure of $Δ$ , and the identification $O_{H D} = O_{A D}$ becomes the statement that Dirichlet-finite harmonic differentials have vanishing periods over $Δ$ — the open-surface shadow of the period theory on a compact curve.

Type and the uniformization dichotomy. For a simply connected open surface the classification collapses to the type problem: by the uniformization theorem such a surface is conformally either $C$ (parabolic, $O_{G}$ ) or the unit disc (hyperbolic, carrying every bounded and Dirichlet-finite function in abundance). The null-class lattice is the genus-and-end refinement of this dichotomy to surfaces that are not simply connected: where uniformization gives a clean binary for the simply connected case, the null-classes measure the residual thinness of a surface whose universal cover is the disc but whose deck group is large enough to suppress some function families. The Nevanlinna and Myrberg type criteria — convergence or divergence of an integral of the modulus of an exhaustion — are the quantitative tests for $O_{G}$ membership.

Quasiconformal invariance and stability. The null-classes are conformal invariants but not topological ones: two homeomorphic surfaces can lie in different classes. They are quasiconformally quasi-invariant in the sense that a quasiconformal map preserves $O_{G}$ (parabolicity is a quasiconformal invariant, since harmonic measure of the ideal boundary being null is preserved up to bounded distortion of the Dirichlet integral) but can move a surface across the $H B$ / $H D$ boundaries when the dilatation is unbounded. This is the open-surface manifestation of the moduli dependence that becomes the Teichmüller theory of infinite-type surfaces.

Synthesis. The null-class lattice is the foundational reason open-surface theory has a classification at all: it converts the vague notion of a surface being thin into a precise, totally ordered hierarchy of function-space obstructions, and the order-reversal between test families and null-classes is exactly the engine that lines those obstructions up. Putting these together, parabolicity $O_{G}$ from 06.11.03 sits at the bottom as the absence of every non-constant positive superharmonic function; the Royden harmonic boundary $Δ$ of 06.11.05 is dual to the function spaces in the sense that the size of $Δ$ reads off membership in each class; and the identification $O_{H D} = O_{A D}$ is the central insight that, once Dirichlet-finiteness forces periods to vanish, harmonic and holomorphic data on an open surface become interchangeable just as they are on the compact surfaces of 06.04.03. This generalises the compact Hodge decomposition to the open setting with a boundary correction, and is dual to the uniformization dichotomy: where uniformization classifies simply connected surfaces into two conformal types, the null-classes classify the residual thinness of all the rest. The lattice builds toward the Sario-Nakai theory of harmonic and analytic boundaries and appears again in every later statement about the type and moduli of infinite-genus surfaces.

Full proof set Master

Proposition 1 (monotonicity of null-classes). If $F \subseteq G$ are two families of functions on open surfaces, each closed under addition of constants, then $O_{G} \subseteq O_{F}$ .

Proof. Let $W \in O_{G}$ , so every member of $G (W)$ is constant. Since $F (W) \subseteq G (W)$ , every member of $F (W)$ is in particular a member of $G (W)$ , hence constant. Thus $W \in O_{F}$ . Applying this with $(F, G) = (A B, H B)$ , $(A D, H D)$ , $(A B, A D)$ , $(H B, H D)$ — using $A B \subseteq H B$ , $A D \subseteq H D$ , $A B \subseteq A D$ , $H B \cap (harmonic) \subseteq$ the Dirichlet-finite ones after the boundary reduction — yields the chain of inclusions among the null-classes from the chain of containments among the function families. $□$

Proposition 2 ( $O_{G} = O_{H P}$ ; parabolicity as the absence of positive harmonics). An open Riemann surface $W$ is parabolic if and only if every positive harmonic function on $W$ is constant.

Proof. If $W$ is hyperbolic it carries a Green's function $g_{W} (\cdot, p)$ , which is a positive harmonic function on $W ∖ {p}$ ; subtracting its singular part against a local potential and applying the Riesz decomposition produces a non-constant positive harmonic function on $W$ , so $W \in / O_{H P}$ . Conversely suppose $W$ carries a non-constant positive harmonic function $h$ . Normalise $in f h = 0$ . The greatest harmonic minorant construction applied to the family of potentials dominated by $h$ produces a non-constant positive superharmonic function with a Green-type singularity along an exhaustion, and the limit of the Green's functions $g_{W_{n}}$ of an exhaustion $W_{1} ⋐ W_{2} ⋐ \dots$ is then a positive harmonic function with logarithmic pole, i.e. a Green's function of $W$ ; so $W$ is hyperbolic. Contrapositively, parabolicity forces every positive harmonic function to be constant. The exhaustion limit $lim_{n} g_{W_{n}}$ is monotone and either diverges to $+ \infty$ everywhere (parabolic case) or converges to a finite Green's function (hyperbolic case); this dichotomy is the analytic core. $□$

Proposition 3 ( $O_{H D} = O_{A D}$ ). For every open Riemann surface $W$ , every Dirichlet-finite harmonic function on $W$ is constant if and only if every Dirichlet-finite holomorphic function on $W$ is constant.

Proof. ( $O_{H D} \Rightarrow O_{A D}$ ) If $f \in A D (W)$ then $u = Re f \in H D (W)$ with $D_{W} (u) \leq D_{W} (Re f) + D_{W} (Im f) < \infty$ ; if $W \in O_{H D}$ then $u$ is constant, and likewise $Im f$ , so $f$ is constant. Hence $W \in O_{A D}$ .

( $O_{A D} \Rightarrow O_{H D}$ ) Let $W \in O_{A D}$ and $u \in H D (W)$ . The differential $ω = * d u$ is harmonic and closed, with $\int_{W} ω \land * ω = D_{W} (u) < \infty$ , so $ω$ is a Dirichlet-finite harmonic differential. By the orthogonal decomposition of Dirichlet-finite harmonic differentials on the Royden compactification, the exact-plus-co-exact part of $ω$ carries no period over the harmonic boundary $Δ$ ; the period $\oint_{γ} ω$ over any cycle $γ$ equals the pairing of $ω$ with the harmonic-measure class of $γ$ , which vanishes for the Dirichlet-finite class because $ω = * d u$ is the conjugate of an exact differential $d u$ . Therefore $ω = d v$ for a single-valued harmonic $v$ on $W$ , and $D_{W} (v) = \int_{W} d v \land * d v = \int_{W} * d u \land * * d u = \int_{W} d u \land * d u = D_{W} (u) < \infty$ (using $* * = - 1$ on $1$ -forms in real dimension two and a sign that cancels in the integral of the squared norm). Then $f = u + i v$ satisfies the Cauchy-Riemann equations $d u = * d v$ , $d v = * d u$ , so $f$ is holomorphic with $D_{W} (Re f) + D_{W} (Im f) = 2 D_{W} (u) < \infty$ , giving $f \in A D (W)$ . By hypothesis $f$ is constant, hence $u = Re f$ is constant. Thus $W \in O_{H D}$ . $□$

Proposition 4 (properness via Royden's example; statement). There exists an open Riemann surface $W_{R}$ with $W_{R} \in O_{A D}$ but $W_{R} \in / O_{A B}$ , so the inclusion $O_{A B} ⊊ O_{A D}$ is proper.

Proof. Stated without full construction — see Royden 1952 ^{[Royden 1952]}. Royden builds $W_{R}$ as a plane domain obtained by deleting from the disc a sequence of closed sets of carefully tuned capacity, arranged so that the domain supports a non-constant bounded holomorphic function (so $W_{R} \in / O_{A B}$ ) while every holomorphic function of finite Dirichlet integral is constant (so $W_{R} \in O_{A D}$ ). The mechanism is that bounded holomorphic functions only require the deleted sets to have positive analytic capacity, whereas Dirichlet-finiteness imposes an $ℓ^{2}$ -summability on the capacities across the ends that the construction defeats. The companion separation $O_{H B} ⊊ O_{H D}$ is Toki's surface ^{[Toki 1952]}, and $O_{G} ⊊ O_{H B}$ is any hyperbolic surface whose harmonic boundary is a single point. The detailed capacity estimates are in Sario-Nakai 1970 ^{[Sario-Nakai 1970]} Ch. IV. $□$

Proposition 5 (parabolicity is conformally invariant and an exhaustion criterion). Membership in $O_{G}$ depends only on the conformal structure of $W$ , and $W \in O_{G}$ iff for some (equivalently every) smooth exhaustion ${W_{n}}$ the harmonic measures $ω_{n}$ of $\partial W_{n}$ seen from a fixed point tend to $1$ .

Proof. The Green's function is defined by a conformally invariant extremal problem (smallest positive superharmonic function with the prescribed logarithmic pole), so its existence is conformally invariant; harmonic measure is likewise conformally invariant. For the criterion: on the exhaustion, the harmonic measure $ω_{n} (z) = ω (z, \partial W_{n}, W_{n} ∖ \overline{W_{0}})$ is decreasing in $n$ relative to the inner boundary and its limit is the harmonic measure of the ideal boundary. $W$ is parabolic iff this limit is $1$ at one (hence every) interior point, equivalently iff $g_{W_{n}} (z, p) \to \infty$ , i.e. no finite Green's function survives. The equivalence of "some" and "every" exhaustion follows because two exhaustions are cofinal and harmonic measure is monotone under inclusion. $□$

Connections Master

Harmonic measure, capacity, and the parabolic/hyperbolic dichotomy 06.11.03. The class $O_{G}$ is defined by the absence of a Green's function, equivalently by harmonic measure of the ideal boundary being null; the entire bottom of the null-class lattice is the potential theory of 06.11.03 recast as a function-space obstruction. Proposition 5 above is exactly the exhaustion criterion proved there, applied to seat $O_{G}$ beneath $O_{H B}$ and $O_{A B}$ .
Royden compactification and the Royden boundary 06.11.05. The harmonic boundary $Δ$ of the Royden compactification is the space on which bounded and Dirichlet-finite harmonic functions acquire boundary values; the lattice statements $O_{H B} ⟺ ∣Δ∣ = 1$ and $O_{G} ⟺ Δ = \emptyset$ are the structural reformulation of the present theorem. The period-vanishing step in $O_{H D} = O_{A D}$ is a statement about Dirichlet-finite differentials on $Δ$ , so this unit and 06.11.05 are co-produced and mutually load-bearing.
Hodge decomposition on a compact Riemann surface 06.04.03. The identification $O_{H D} = O_{A D}$ is the open-surface analogue of the compact fact that harmonic $1$ -forms split into holomorphic and anti-holomorphic parts; on a compact surface there is no boundary and the periods are the classical periods of 06.04.03, while on an open surface Dirichlet-finiteness is the integrability condition that makes the harmonic conjugate single-valued. The open theory is the compact Hodge decomposition with a harmonic-boundary correction term.
Uniformization theorem 06.03.03. For simply connected open surfaces the null-class lattice collapses to the conformal dichotomy $C$ (parabolic) versus the disc (hyperbolic) supplied by uniformization 06.03.03; the null-classes are the refinement of the type problem to surfaces of higher connectivity, measuring the residual thinness that uniformization alone does not detect.
Moduli of Riemann surfaces 06.08.03. The quasiconformal quasi-invariance of the null-classes — $O_{G}$ preserved, the $H B$ / $H D$ boundaries movable under unbounded dilatation — is the entry point to the moduli theory of open surfaces 06.08.03, where the position of a surface in the lattice is a coarse invariant on the infinite-dimensional Teichmüller space of an infinite-type surface; this is the forward pointer toward the potential-theoretic study of moduli.

Historical & philosophical context Master

The classification of open Riemann surfaces by function-space null-classes is the organising achievement of Ahlfors and Sario's 1960 monograph Riemann Surfaces ^{[Ahlfors-Sario 1960]} (Princeton University Press), whose Chapter II §6 introduced the systematic notation $O_{G}, O_{H B}, O_{H D}, O_{A B}, O_{A D}$ and proved the inclusion lattice. The problem grew out of the type problem for simply connected surfaces studied by Rolf Nevanlinna and Lars Ahlfors in the 1930s — deciding whether a given simply connected covering surface is conformally the plane or the disc — and the recognition by Pekka Myrberg and Maurice Heins that the type dichotomy is the absence or presence of a Green's function. Halsey Royden's 1952 paper Some remarks on open Riemann surfaces ^{[Royden 1952]} (Annales Academiae Scientiarum Fennicae) supplied the separating surface that shows the analytic inclusions are proper, and introduced the algebra of Dirichlet-finite functions whose maximal-ideal space became the Royden compactification. Yukio Toki's 1952 On the classification of open Riemann surfaces ^{[Toki 1952]} (Osaka Mathematical Journal) gave the surface separating $O_{H B}$ from $O_{H D}$ .

The mature theory was consolidated by Leo Sario and Mitsuru Nakai in Classification Theory of Riemann Surfaces ^{[Sario-Nakai 1970]} (Springer Grundlehren 164, 1970), which extended the lattice to the positive-harmonic and quasi-bounded classes and developed the harmonic-boundary machinery in full. The potential-theoretic substrate — Green's functions, harmonic measure, and capacity on open surfaces — was systematised in Masatsugu Tsuji's Potential Theory in Modern Function Theory ^[Tsuji], whose treatment of parabolic surfaces underlies the $O_{G}$ end of the lattice. The subject is the function-theoretic counterpart of the uniformization theorem: where uniformization is a clean binary for simply connected surfaces, Ahlfors and Sario showed that for surfaces of arbitrary connectivity the right invariant is not a single conformal type but a graded hierarchy of nullness indexed by which function families a surface refuses to carry.

Bibliography Master

@book{AhlforsSario1960,
  author    = {Ahlfors, Lars V. and Sario, Leo},
  title     = {Riemann Surfaces},
  series    = {Princeton Mathematical Series},
  volume    = {26},
  publisher = {Princeton University Press},
  year      = {1960}
}

@article{Royden1952,
  author  = {Royden, Halsey L.},
  title   = {Some remarks on open {Riemann} surfaces},
  journal = {Ann. Acad. Sci. Fenn. Ser. A I Math.},
  volume  = {249/5},
  year    = {1952},
  pages   = {1--13}
}

@article{Toki1952,
  author  = {Toki, Yukio},
  title   = {On the classification of open {Riemann} surfaces},
  journal = {Osaka Math. J.},
  volume  = {4},
  year    = {1952},
  pages   = {191--201}
}

@book{SarioNakai1970,
  author    = {Sario, Leo and Nakai, Mitsuru},
  title     = {Classification Theory of {Riemann} Surfaces},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {164},
  publisher = {Springer},
  year      = {1970}
}

@book{Tsuji1959,
  author    = {Tsuji, Masatsugu},
  title     = {Potential Theory in Modern Function Theory},
  publisher = {Maruzen},
  year      = {1959}
}

@article{Heins1952,
  author  = {Heins, Maurice},
  title   = {Riemann surfaces of infinite genus},
  journal = {Ann. of Math. (2)},
  volume  = {55},
  year    = {1952},
  pages   = {296--317}
}

@book{Nevanlinna1953,
  author    = {Nevanlinna, Rolf},
  title     = {Eindeutige analytische Funktionen},
  edition   = {2},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {46},
  publisher = {Springer},
  year      = {1953}
}

Prerequisites

06.11.03 pending
06.11.05 pending

Tier anchors

beginner: Ahlfors-Sario *Riemann Surfaces* Ch. II §6 informal opening; Forster *Lectures on Riemann Surfaces* §22-§23 (parabolic vs hyperbolic) informal
intermediate: Ahlfors-Sario *Riemann Surfaces* (Princeton 1960) Ch. II §6; Sario-Nakai *Classification Theory of Riemann Surfaces* (Springer 1970) Ch. I-III
master: Ahlfors-Sario *Riemann Surfaces* (Princeton 1960) Ch. II §6 (originator monograph for the null-class lattice); Royden 1952 *Some remarks on open Riemann surfaces* (originator separating example); Sario-Nakai *Classification Theory of Riemann Surfaces* (Springer Grundlehren 164, 1970); Tsuji *Potential Theory in Modern Function Theory* Ch. III-V

References

Ahlfors-Sario 1960 — Riemann Surfaces · Princeton University Press, Princeton Mathematical Series 26, Ch. II §6 (the null-class lattice O_G, O_HB, O_HD, O_AB, O_AD)
Royden 1952 — Some remarks on open Riemann surfaces · Ann. Acad. Sci. Fenn. Ser. A I Math. 249/5, pp. 1--13 (separating example for the inclusion lattice)
Sario-Nakai 1970 — Classification Theory of Riemann Surfaces · Springer Grundlehren der mathematischen Wissenschaften 164, Ch. I-IV (systematic theory of the null-classes)
Toki 1952 — On the classification of open Riemann surfaces · Osaka Math. J. 4, pp. 191--201 (separating example O_HB vs O_HD)
Tsuji — Potential Theory in Modern Function Theory · Maruzen 1959 / Chelsea reprint, Ch. III-V (Green's function, harmonic measure, parabolic surfaces)

Estimated time

beginner: 14m
intermediate: 40m
master: 70m