Pipeline-generated

How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

✓ 12 results · all verified · 12 also independently AI-judged

Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs; all 12 also cleared it.

Hodge Theory on Compact Riemann Surfaces

1 · Prerequisites

2 · Summary

This page develops the Dolbeault theory of holomorphic line bundles on a compact Riemann surface, with a supplied compatible Riemannian metric and Hermitian bundle metric. The maximal weak operator and its Hilbert adjoint fix the domains before local formulas and elliptic estimates establish smoothness, compactness and closed range. The resulting smooth and Hilbert Hodge decompositions give unique harmonic representatives and finite-dimensional Dolbeault cohomology.

The conjugate-linear bundle-valued Hodge star sends harmonic $(0, 1)$ -forms with values in $E$ to holomorphic sections of $K \otimes E^{*}$ . Its local Laplacian commutation and the integral wedge pairing yield complex-bilinear nondegenerate duality. Harmonic representatives depend on the metrics; the cohomology groups and their dimensions depend only on the holomorphic bundle. The analytic results explicitly retain the Axiom of Choice inherited through their Sobolev and Hilbert interfaces.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Holomorphic line bundles and meromorphic sections on a Riemann surface

Definition

Let $X$ be a Riemann surface with maximal holomorphic atlas $A$ (Riemann surfaces and holomorphic atlases). A smooth complex line bundle $π : E \to X$ is a smooth real rank-two vector bundle whose fibres carry complex vector-space structures and which has local trivializations $θ_{i} : E ∣_{U_{i}} \to U_{i} \times C$ that are complex-linear on each fibre. The transition functions are defined by $θ_{i} \circ θ_{j}^{- 1} (x, v) = (x, g_{i j} (x) v), g_{i j} : U_{i} \cap U_{j} \to C^{\times} .$ A holomorphic line bundle is such a bundle with a trivializing cover by domains of holomorphic charts for which every $g_{i j}$ is holomorphic. These functions obey $g_{i j} g_{j k} = g_{i k}$ on triple overlaps. Conversely, a holomorphic $C^{\times}$ -valued cocycle on a supplied countable open cover constructs a holomorphic line bundle: regard each scalar as its real $2 \times 2$ multiplication matrix, apply the smooth cocycle construction, and note that the resulting transitions commute with multiplication by $i$ and are holomorphic.

Write $e_{i} = θ_{i}^{- 1} (1)$ for the associated local frame. A smooth section has the form $s = f_{i} e_{i}$ locally, with $f_{i} \in C^{\infty} (U_{i}; C)$ and $f_{i} = g_{i j} f_{j}$ on overlaps. It is holomorphic when each $f_{i}$ is holomorphic; $H^{0} (X, E)$ denotes the complex vector space of holomorphic sections. A meromorphic section is a family of meromorphic functions $f_{i}$ with the same transition law. For a nonzero meromorphic section, define its order at $p$ by ${ord}_{p} (s) : = {ord}_{z_{i} (p)} (f_{i})$ in any holomorphic chart and local frame. Its divisor is $(s) : = \sum_{p \in X} {ord}_{p} (s) p$ ; this sum is locally finite and is finite when $X$ is compact.

The canonical bundle is $K : = Λ^{1, 0} T^{*} X$ . If $z_{i}, z_{j}$ are holomorphic coordinates on an overlap, then its transition function in the convention above is $g_{i j} = d z_{j} / d z_{i}$ : a local differential satisfies $f_{i} d z_{i} = f_{j} d z_{j}$ . Hence holomorphic sections of $K$ are precisely holomorphic differentials, and meromorphic sections of $K$ are meromorphic differentials (Meromorphic differentials, orders and residues). The conjugate cocycle $\overline{g_{i j}}$ defines $Λ^{0, 1} T^{*} X = \overline{K}$ , while $Λ^{0, 0} T^{*} X = X \times C$ . Thus the bundles of $E$ -valued $(0, 0)$ - and $(0, 1)$ -forms are $E$ and $Λ^{0, 1} T^{*} X \otimes E$ ; locally the latter has the form $f (z) d \overset{ˉ}{z} \otimes e$ (Bigraded complex forms and the Dolbeault operators, The exterior power bundle of the cotangent bundle).

For a holomorphic line bundle $E$ , the Dolbeault operator is ${\overset{ˉ}{\partial}}_{E} : C^{\infty} (X, E) ⟶ C^{\infty} (X, Λ^{0, 1} T^{*} X \otimes E), {\overset{ˉ}{\partial}}_{E} (f e_{i}) : = (\partial_{\overset{ˉ}{z}} f) d \overset{ˉ}{z} \otimes e_{i}$ in a holomorphic frame $e_{i}$ and coordinate $z$ . It is $C$ -linear, satisfies ${\overset{ˉ}{\partial}}_{E} (f s) = \overset{ˉ}{\partial} f \otimes s + f {\overset{ˉ}{\partial}}_{E} s$ , and ${\overset{ˉ}{\partial}}_{E} s = 0$ exactly when $s$ is holomorphic. It extends coefficientwise to $E$ -valued forms: in a holomorphic frame, ${\overset{ˉ}{\partial}}_{E} (α \otimes e_{i}) = (\overset{ˉ}{\partial} α) \otimes e_{i}$ . This extension satisfies ${\overset{ˉ}{\partial}}_{E}^{2} = 0$ and ${\overset{ˉ}{\partial}}_{E} (β \land s) = \overset{ˉ}{\partial} β \land s + (- 1)^{\deg β} β \land {\overset{ˉ}{\partial}}_{E} s$ for complex-valued forms $β$ . The complex dual $E^{*}$ has inverse holomorphic cocycle $g_{i j}^{- 1}$ and its corresponding Dolbeault operator.

Facts & Assumptions

Given: A connected Riemann surface $X$ with its maximal atlas, a smooth complex line bundle $E \to X$ , holomorphic trivializations and their transition functions, and a meromorphic section when order or divisor is discussed.

[F1]

The charts of $X$ are compatible biholomorphisms; their transitions are smooth and have nonzero derivative. A Riemann surface is a connected smooth two-manifold (Riemann surfaces and holomorphic atlases, Topological manifolds without boundary: Hausdorff, second-countable, and locally Euclidean spaces, Smooth manifolds and their smooth charts, Holomorphic functions are real analytic and smooth in their two real coordinates).

[F2]

A smooth vector bundle is locally trivial with linear fibre maps; smooth sections have smooth local components in a frame. A smooth cocycle on a supplied countable cover constructs a smooth real vector bundle (Smooth vector bundles, rank, fibres, and trivial bundles, Vector bundle charts and transition functions, Construction of a vector bundle from a smooth cocycle, Local and global frames of a vector bundle, Smooth sections, local sections, and support, Smoothness of a section is equivalent to smooth local components).

[F3]

Complex forms split by type, $d = \partial + \overset{ˉ}{\partial}$ , and the scalar Dolbeault operator obeys the square-zero and graded Leibniz identities (A smooth differential $k$ -form, The wedge product of differential forms, The exterior power bundle of the cotangent bundle, Bigraded complex forms and the Dolbeault operators, The d, partial and dbar identities).

[F4]

A meromorphic function has isolated zeros and poles unless it is identically zero; a nonzero holomorphic function has a finite zero order and a local factorization by that power. The identity theorem applies on each connected chart. The transition law for a meromorphic differential is $h_{j} (w) = h_{i} (z (w)) z^{'} (w)$ , equivalently $h_{i} d z_{i} = h_{j} d z_{j}$ (Meromorphic functions on a plane domain, Isolated singularities: removable, poles, and essential singularities, Identity theorem for holomorphic functions, The order of a zero is the exponent in its local holomorphic factorization, Holomorphic maps and meromorphic functions on Riemann surfaces, Meromorphic differentials, orders and residues).

[F5]

For a smooth complex-valued function, $\partial_{\overset{ˉ}{z}} f = 0$ is equivalent to complex differentiability and hence holomorphy; the chain rule gives the change-of-coordinate formula (The Wirtinger derivatives $\partial_{z} f$ and $\partial_{\overset{ˉ}{z}} f$ , and antiholomorphic functions, The chain rule for complex derivatives, Complex differentiability is equivalent to real total differentiability together with a complex-linear derivative, with $\partial_{\overset{ˉ}{z}} f = 0$ , or with the Cauchy–Riemann equations).

[F6]

The smooth dual of a vector bundle is defined fibrewise; for a complex line the complex-linear dual transition is the inverse scalar (Dual and Hom vector bundles).

Proof

technique · direct local verification

1.1F1F2given

The fibrewise complex-linear trivializations give smooth transitions in $C^{\times}$ and composition of the maps $θ_{i} \circ θ_{j}^{- 1}$ gives $g_{i j} g_{j k} = g_{i k}$ . For a supplied countable holomorphic cocycle, its real multiplication matrices are smooth $G L (2, R)$ transitions, so [F2] constructs the underlying smooth rank-two bundle; these matrices commute with the standard complex structure, and their holomorphicity makes the local total-space charts holomorphic.

1.2F1F3F4givenalgebra

In a local frame $e_{i}$ , the cotangent line $K$ is spanned by $d z_{i}$ . If $z_{j} = ϕ (z_{i})$ , then $d z_{j} = ϕ^{'} (z_{i}) d z_{i}$ and a differential $ω = f_{i} d z_{i} = f_{j} d z_{j}$ has $f_{i} = (d z_{j} / d z_{i}) f_{j}$ . Since $ϕ^{'}$ is holomorphic and nowhere zero, these are holomorphic line-bundle transitions; a local section of $K$ is holomorphic or meromorphic exactly when its coefficient $f_{i}$ is, so these sections are precisely the corresponding differentials. Conjugation gives the stated transitions for $\overline{K}$ , and the type decomposition gives the local formulas for $E$ -valued forms.

1.3F1F4given

Let $Z$ be the set of points having a neighborhood on which the section is zero. It is open. If $p$ is in its closure, choose a connected chart and frame around $p$ ; the meromorphic coefficient has zeros accumulating at $p$ . A pole at $p$ is impossible because its finite principal part is nonzero on a punctured neighborhood, and otherwise the holomorphic identity theorem makes the coefficient identically zero near $p$ . Thus $Z$ is closed; since $X$ is connected and the section is nonzero, $Z = \emptyset$ . Each local representative consequently has a finite Laurent order at every point. Under a change of frame it is multiplied by a holomorphic unit, and under a coordinate change its argument is composed with a biholomorphism with nonzero derivative; neither operation changes the zero or pole order. Its zeros and poles are therefore locally isolated, so the divisor is locally finite and has finite support on compact $X$ .

1.4F3F5givenalgebra

On an overlap with $z_{j} = ϕ (z_{i})$ and $e_{j} = g_{i j} e_{i}$ , a section has coefficients $f_{i} = g_{i j} f_{j}$ . The chain rule and holomorphy of $g_{i j}$ give $\overset{ˉ}{\partial} f_{i} = g_{i j} \overset{ˉ}{\partial} f_{j}$ as $(0, 1)$ -forms, so the local formula defines a global operator. The same calculation applies to $E$ -valued forms; the scalar graded Leibniz rule gives the displayed rule, and scalar ${\overset{ˉ}{\partial}}^{2} = 0$ gives ${\overset{ˉ}{\partial}}_{E}^{2} = 0$ . By [F5], its kernel on sections is exactly the holomorphic sections.

2.1F2F6step 1.1step 1.2step 1.3step 1.4∎

In the complex-linear dual frame the transition is $g_{i j}^{- 1}$ , which is holomorphic and nonvanishing, so the dual is a holomorphic line bundle and the same local construction defines ${\overset{ˉ}{\partial}}_{E^{*}}$ . The preceding coordinate and frame calculations establish all stated definitions and well-definedness claims. No choice principle is used.

DefinitionDefinition: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Hermitian metric and $L^{2}$ pairing on a compact Riemann surface

Definition

Let $X$ be a compact Riemann surface and $E \to X$ a holomorphic line bundle (Holomorphic line bundles and meromorphic sections on a Riemann surface). Its complex structure $J : T X \to T X$ is defined in a holomorphic coordinate $z = x + i y$ by $J \partial_{x} = \partial_{y}$ and $J \partial_{y} = - \partial_{x}$ . The complex orientation is the orientation for which $d x \land d y$ is positive. The complexified cotangent bundle splits into the $i$ and $- i$ eigenspaces of $J^{*}$ : $T^{*} X \otimes_{R} C = Λ^{1, 0} T^{*} X \oplus Λ^{0, 1} T^{*} X .$ The type $(p, q)$ of a form records its holomorphic and antiholomorphic factors; on a curve $q = 0, 1$ for $(0, q)$ -forms.

A Riemannian metric $g$ is compatible with $J$ when $g (J v, J w) = g (v, w)$ . In every holomorphic coordinate this is equivalent to $g = ρ (d x^{2} + d y^{2}), ρ > 0,$ and such metrics exist: if $g_{0}$ is any Riemannian metric, then $g (v, w) = \frac{1}{2} (g_{0} (v, w) + g_{0} (J v, J w))$ is compatible. The two type summands of complexified one-forms are orthogonal. The oriented Riemannian volume form and its associated density are, respectively, $d V_{g} = ρ d x \land d y = \frac{i ρ}{2} d z \land d \overset{ˉ}{z}, μ_{g} = ρ ∣ d x d y ∣ .$

A Hermitian metric $h$ on $E$ is a smooth positive-definite Hermitian form on each fibre, complex-linear in the first argument and conjugate-linear in the second. Such a metric exists: average a smooth real bundle metric $b$ by the fibre complex structure $J_{E} v : = i v$ to make it $J_{E}$ -invariant, and set $h (v, w) = b (v, w) - i b (J_{E} v, w)$ . In a holomorphic frame $e$ , its weight is the smooth positive function $ψ = h (e, e)$ . Equip the complex-linear dual $E^{*}$ with the dual Hermitian metric, so $h^{*} (e^{*}, e^{*}) = ψ^{- 1}$ .

For $q = 0, 1$ , the bundles $Λ^{0, q} T^{*} X \otimes E$ carry the pointwise Hermitian pairing induced by $g$ on forms and $h$ on $E$ , still linear in the first argument. Write $(α, β)_{C}$ for the complex-bilinear extension of the real exterior metric. The scalar Hodge star $⋆$ is the real Hodge star of Riemannian hodge star, extended $C$ -linearly; it satisfies $α \land ⋆ β = (α, β)_{C} d V_{g}$ and $⋆^{2} = (- 1)^{k (2 - k)}$ on complex $k$ -forms (Hodge star is a smooth bundle isomorphism, Hodge star squared sign).

The bundle-valued Hodge map $⋆_{F} : C^{\infty} (X, Λ^{k} T^{*} X \otimes F) ⟶ C^{\infty} (X, Λ^{2 - k} T^{*} X \otimes F^{*})$ for a Hermitian line bundle $F$ is the unique conjugate-linear map satisfying $s \land ⋆_{F} t = ⟨ s, t ⟩ d V_{g},$ where the $F$ -factor is paired with $F^{*}$ by evaluation. On $(0, q)$ -forms it has the type-correct target $Λ^{1, 1 - q} T^{*} X \otimes F^{*}$ . In a holomorphic coordinate and frame with $h (e, e) = ψ$ , $⋆_{E} (f e) = \frac{i ρ ψ}{2} \overset{ˉ}{f} d z \land d \overset{ˉ}{z} \otimes e^{*}, ⋆_{E} (u d \overset{ˉ}{z} \otimes e) = - i ψ \overset{ˉ}{u} d z \otimes e^{*} .$ For $q = 1$ , $⋆_{E}$ is an isomorphism from $Λ^{0, 1} T^{*} X \otimes E$ to $K \otimes E^{*}$ and $⋆_{E}^{- 1} = - ⋆_{E^{*}}$ on that target. With the flat metric $d x^{2} + d y^{2}$ and trivial weight, $⋆_{E} (u d \overset{ˉ}{z} \otimes e) = - i \overset{ˉ}{u} d z \otimes e^{*}$ .

For smooth compactly supported $E$ -valued $(0, q)$ -forms $s, t$ , define $⟨ s, t ⟩_{L^{2}} : = \int_{X} ⟨ s, t ⟩ d V_{g} .$ The integral is the density integral of the corresponding complex-valued density, computed on real and imaginary parts. This is the complex $L^{2}$ pairing in the library's first-variable-linear convention (The complex $L^{2}$ pairing on equivalence classes, Complex Lp classes and Euclidean test-function conventions). Its completion is denoted $L^{2} (X, Λ^{0, q} T^{*} X \otimes E)$ ; the compactly supported smooth forms are dense there, and the completion is a complex Hilbert space.

Assume the Axiom of Countable Choice ${A C}_{ω}$ (The Axiom of Countable Choice ( ${A C}_{ω}$ )). It is used only through the declared interfaces for existence of Riemannian and bundle metrics, partitions and density integration, Euclidean smooth $L^{2}$ -density, and the Hilbert completion; this item uses no full Axiom of Choice.

Facts & Assumptions

Given: A compact connected Riemann surface $X$ , its holomorphic line bundle $E$ , compatible metrics $g$ and $h$ , and ${A C}_{ω}$ .

[F1]

Holomorphic coordinate changes have complex-linear derivative, nonzero real determinant, and the tangent coordinate bases transform by the chain rule. The holomorphic charts give the connected smooth surface and the complex line-bundle data (Holomorphic line bundles and meromorphic sections on a Riemann surface, The tangent bundle as a disjoint union, Change-of-coordinate formula for tangent bases, Complex differentiability is equivalent to real total differentiability together with a complex-linear derivative, with $\partial_{\overset{ˉ}{z}} f = 0$ , or with the Cauchy–Riemann equations).

[F2]

Under countable choice every smooth manifold admits a Riemannian metric and every smooth vector bundle admits a smooth bundle metric; smooth partitions of unity exist under the same assumption (Every smooth manifold admits a riemannian metric, Every smooth vector bundle admits a smooth bundle metric, Smooth bundle metrics, Smooth partitions of unity subordinate to an open cover, Smooth partitions of unity exist on manifolds).

[F3]

The induced metric on exterior powers is determinant-normalized, so for $g = ρ (d x^{2} + d y^{2})$ one has $⟨ d z, d z ⟩_{g} = ⟨ d \overset{ˉ}{z}, d \overset{ˉ}{z} ⟩_{g} = 2 / ρ$ and $⟨ d z, d \overset{ˉ}{z} ⟩_{g} = 0$ . The real Hodge star is characterized by its wedge identity and has square sign $(- 1)^{k (2 - k)}$ (Pointwise norm and angle from a riemannian metric, Riemannian hodge star, Riemannian metrics induce metrics on dual tensor and exterior bundles, Hodge star is a smooth bundle isomorphism, Hodge star squared sign).

[F4]

The Riemannian volume density is coordinate-independent; under the complex orientation its local volume form is $d V_{g} = ρ d x \land d y$ . Its smooth density integral is the integral for a locally finite Radon measure (Pointwise Borel nonnegative densities, Integral of a compactly supported smooth density, Oriented smooth manifolds and oriented charts, Riemannian metric and riemannian manifold, Riemannian volume density, Riemannian volume of a compactly supported smooth density, The riemannian volume density is coordinate independent, Riemannian volume is the radon measure of the riemannian density, Positive smooth densities give Radon volume).

[F5]

The complex $L^{2}$ pairing is linear in the first variable, obeys Cauchy–Schwarz, and agrees with $\int f \overset{ˉ}{g}$ in local scalar coefficients. Under countable choice, Euclidean smooth compact-support functions are dense in $L^{2}$ , and the completion of an inner-product space is Hilbert (The complex $L^{2}$ pairing on equivalence classes, Complex Lp classes and Euclidean test-function conventions, Complex completeness, density, and inner product: the consumer interface, Hilbert space, The norm completion of an inner-product space is a Hilbert space, The complex $L^{2}$ pairing is well-defined and satisfies Cauchy–Schwarz).

[F6]

The complex-linear dual $E^{*}$ is defined fibrewise; its metric in the dual frame satisfies $h^{*} (e^{*}, e^{*}) = ψ^{- 1}$ (Dual and Hom vector bundles).

[F7]

Every finite-dimensional Hermitian space satisfies Cauchy–Schwarz (Cauchy–Schwarz: $∣ ⟨ u, v ⟩ ∣ \leq ∥ u ∥ ∥ v ∥$ , with equality exactly for linearly dependent vectors).

[F8]

A compact set inside an open set admits a smooth cutoff equal to one near the compact set and supported in the open set (A manifold bump for a compact set inside an open set).

Proof

technique · direct local calculation and completion

1.1F1given

If $z_{j} = ϕ (z_{i})$ is a holomorphic change of coordinate, its real Jacobian is multiplication by the complex derivative $ϕ^{'}$ , so it commutes with multiplication by $i$ and has determinant $∣ ϕ^{'} ∣^{2} > 0$ . Thus the local rotations defining $J$ agree on overlaps and the charts orient $X$ consistently. The complexified cotangent eigenbundles are consequently well-defined, with $Λ^{0, 1} T^{*} X = \overline{K}$ and $Λ^{0, 0} T^{*} X$ trivial.

1.2F2F3F4givenalgebra

By [F2] choose a Riemannian metric $g_{0}$ and average it with $J$ ; the result is positive and satisfies $g (J v, J w) = g (v, w)$ . In a chart, writing $g = a d x^{2} + 2 b d x d y + c d y^{2}$ , this identity gives $a = c$ and $b = 0$ , hence $g = ρ (d x^{2} + d y^{2})$ with $ρ = a > 0$ . Its determinant is $ρ^{2}$ , so the Riemannian density and complex-oriented volume form have the displayed local formulas. The determinant metric on covectors makes $d z, d \overset{ˉ}{z}$ orthogonal with squared norms $2 / ρ$ , establishing the type orthogonality.

1.3F1F2givenalgebra

By [F2] the underlying real bundle of $E$ has a smooth positive real fibre metric $b$ . Its average $b_{J} (v, w) = \frac{1}{2} (b (v, w) + b (J_{E} v, J_{E} w))$ is smooth, positive and $J_{E}$ -invariant. Invariance makes $J_{E}$ orthogonal and skew-adjoint, so $b_{J} (J_{E} v, v) = 0$ ; direct substitution then shows that $h (v, w) = b_{J} (v, w) - i b_{J} (J_{E} v, w)$ is complex-linear in $v$ , conjugate-linear in $w$ , Hermitian, and satisfies $h (v, v) = b_{J} (v, v) > 0$ for $v \neq 0$ . Hence it is a smooth Hermitian bundle metric, and a holomorphic frame has smooth positive weight $ψ$ .

1.4F3F6givenalgebra

In a local frame the Hermitian dual identification sends $e$ to $ψ e^{*}$ . Combining it with the conjugate-linear metric star on complex forms gives the two displayed local formulas for $⋆_{E}$ . For $q = 0$ , $f e \land ⋆_{E} (g e) = ψ f \overset{ˉ}{g} d V_{g}$ . For $q = 1$ , $d \overset{ˉ}{z} \land d z = 2 i d x \land d y$ and $⟨ d \overset{ˉ}{z}, d \overset{ˉ}{z} ⟩_{g} = 2 / ρ$ , so $f d \overset{ˉ}{z} \otimes e \land ⋆_{E} (u d \overset{ˉ}{z} \otimes e) = (2 ψ f \overset{ˉ}{u}) d x \land d y = ⟨ f d \overset{ˉ}{z} \otimes e, u d \overset{ˉ}{z} \otimes e ⟩ d V_{g}$ . The defining pairing is nondegenerate, so these formulas determine a unique global conjugate-linear map with target bidegree $(1, 1 - q)$ . For $a d z \otimes e^{*}$ , the same calculation gives $⋆_{E^{*}} (a d z \otimes e^{*}) = i ψ^{- 1} \overset{ˉ}{a} d \overset{ˉ}{z} \otimes e^{* *}$ ; hence $⋆_{E^{*}} ⋆_{E} = - 1$ on $(0, 1)$ -forms and $⋆_{E}^{- 1} = - ⋆_{E^{*}}$ .

2.1F4F5F7step 1.4given

For compactly supported smooth $s, t$ , the pointwise pairing is smooth and the volume density is positive and smooth, so its integral is finite; linearity and conjugate symmetry follow pointwise and from the integral. If $s \neq 0$ , its pointwise squared norm is positive on a neighborhood of a point where $s$ is nonzero, so the integral is strictly positive. The defining equation for $⋆_{E}$ yields $⟨ s, t ⟩_{L^{2}} = \int_{X} s \land ⋆_{E} t$ . Pointwise Cauchy–Schwarz [F7] gives $∣ ⟨ s, t ⟩ ∣ \leq ∣ s ∣ ∣ t ∣$ , and scalar $L^{2}$ Cauchy–Schwarz [F5] then gives $∣ ⟨ s, t ⟩_{L^{2}} ∣ \leq ∥ s ∥_{L^{2}} ∥ t ∥_{L^{2}}$ .

3.1F2F4F5F8step 2.1

Choose a finite holomorphic-chart/frame cover and a subordinate smooth partition $(χ_{j})$ . For a measurable square-integrable section $s$ , each $χ_{j} s$ has compact support $K_{j} \subset U_{j}$ . In the chart, its coefficient extended by zero lies in Euclidean $L^{2}$ . By [F8], choose a cutoff $η_{j}$ supported in $U_{j}$ and equal to one near $K_{j}$ . On the compact support of $η_{j}$ , the smooth positive metric and volume weights are bounded above and below. Approximate the coefficient by Euclidean $C_{c}^{\infty}$ functions from [F5], multiply those approximants by $η_{j}$ , convert them back to sections, and extend by zero; the norm equivalence on $supp (η_{j})$ makes the resulting sections converge to $χ_{j} s$ . Summing over the finite cover proves density of compactly supported smooth sections in the measurable realization. The countable-choice completion interface in [F5] makes the completion a complex Hilbert space. The only choice assumption is ${A C}_{ω}$ through the metric, partition, density-integral and completion suppliers listed in [F2], [F4], and [F5].

4.1F5step 3.1∎

The completion of the dense inner-product space is the stated $L^{2}$ Hilbert space, and its norm is the completion of the bundle norm. The smooth compactly supported forms are dense by construction and by step 3.1, so the two descriptions agree.

DefinitionDefinition: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface

Statement

Assume the Axiom of Choice (The Axiom of Choice), carried here by the Sobolev restriction and cutoff localization interface; the partitions, $L^{2}$ completions, Hilbert adjoints, and graph-space Riesz argument use only Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Let $X$ be a compact Riemann surface, $E$ a holomorphic line bundle with Hermitian metric $h$ , and $g$ a compatible Riemannian metric (Hermitian metric and $L^{2}$ pairing on a compact Riemann surface). For $q = 0, 1$ , let $L_{q}^{2} : = L^{2} (X, Λ^{0, q} T^{*} X \otimes E)$ be the complex Hilbert completion in the first-variable-linear $L^{2}$ pairing.

Weak Dolbeault derivative. For $u \in L_{0}^{2}$ and $v \in L_{1}^{2}$ , say that $u$ has weak Dolbeault derivative $v$ when $\int_{X} v \land φ = - \int_{X} u \land {\overset{ˉ}{\partial}}_{E^{*}} φ for every φ \in C_{c}^{\infty} (X, Λ^{1, 0} T^{*} X \otimes E^{*}),$ where the $E$ and $E^{*}$ factors are paired by evaluation. Such $v$ , if it exists, is unique. The maximal Dolbeault operator is $\overset{ˉ}{D} : dom \overset{ˉ}{D} \subseteq L_{0}^{2} ⟶ L_{1}^{2}, dom \overset{ˉ}{D} : = {u \in L_{0}^{2} : \exists v \in L_{1}^{2} satisfying the weak identity above}, \overset{ˉ}{D} u : = v .$ It is densely defined and closed, and extends the smooth operator ${\overset{ˉ}{\partial}}_{E}$ defined in Holomorphic line bundles and meromorphic sections on a Riemann surface. Its Hilbert adjoint ${\overset{ˉ}{D}}^{*} : dom {\overset{ˉ}{D}}^{*} \subseteq L_{1}^{2} ⟶ L_{0}^{2}$ is the unique operator satisfying $⟨ \overset{ˉ}{D} u, w ⟩_{L^{2}} = ⟨ u, {\overset{ˉ}{D}}^{*} w ⟩_{L^{2}} (u \in dom \overset{ˉ}{D}, w \in dom {\overset{ˉ}{D}}^{*}) .$ It is closed and densely defined, and $\ker {\overset{ˉ}{D}}^{*} = (ran \overset{ˉ}{D})^{⊥} in L_{1}^{2} .$

Dolbeault Laplacian. On $L_{0}^{2} \oplus L_{1}^{2}$ define $Δ^{''} : = (\begin{array}{cc} {\overset{ˉ}{D}}^{*} \overset{ˉ}{D} & 0 \\ 0 & \overset{ˉ}{D} {\overset{ˉ}{D}}^{*} \end{array}),$ with $dom Δ^{''} = {(u_{0}, u_{1}) : u_{0} \in dom \overset{ˉ}{D}, \overset{ˉ}{D} u_{0} \in dom {\overset{ˉ}{D}}^{*}, u_{1} \in dom {\overset{ˉ}{D}}^{*}, {\overset{ˉ}{D}}^{*} u_{1} \in dom \overset{ˉ}{D}} .$ This is a self-adjoint nonnegative operator, and $⟨ Δ^{''} (u_{0}, u_{1}), (u_{0}, u_{1}) ⟩_{L^{2}} = ∥ \overset{ˉ}{D} u_{0} ∥_{L^{2}}^{2} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥_{L^{2}}^{2} .$ Consequently $\ker Δ^{''} = (\ker \overset{ˉ}{D}) \oplus (\ker {\overset{ˉ}{D}}^{*})$ . Write $H^{0, q} (X, E) : = \ker Δ_{q}^{''} \subseteq L_{q}^{2} (q = 0, 1)$ for the harmonic $(0, q)$ -forms, where $Δ_{0}^{''} = {\overset{ˉ}{D}}^{*} \overset{ˉ}{D}$ and $Δ_{1}^{''} = \overset{ˉ}{D} {\overset{ˉ}{D}}^{*}$ .

Sobolev spaces. Fix a finite holomorphic chart and frame cover $(U_{j}, e_{j})$ , a subordinate smooth partition of unity $(χ_{j})$ , and $q = 0, 1$ . For $k = 1, 2$ , the space $H^{k} (X, Λ^{0, q} T^{*} X \otimes E)$ is the completion of smooth $E$ -valued $(0, q)$ -forms in the norm $∥ u ∥_{H^{k}}^{2} : = \sum_{j} ∥ (χ_{j} u)_{e_{j}} ∥_{W^{k, 2} (U_{j})}^{2},$ where $(χ_{j} u)_{e_{j}}$ is the scalar local coefficient, including the $d {\overset{ˉ}{z}}_{j}$ coefficient when $q = 1$ . Different finite covers, frames, and subordinate partitions give equivalent norms and the same completed space. The natural inclusions $H^{2} (X, Λ^{0, q} T^{*} X \otimes E) ↪ H^{1} (X, Λ^{0, q} T^{*} X \otimes E) ↪ L_{q}^{2}$ are continuous, and smooth forms are dense in each $H^{k}$ .

Facts & Assumptions

Given: A compact Riemann surface $X$ , a holomorphic line bundle $E$ , supplied compatible metrics $g, h$ , and the stated Axiom of Choice and Countable Choice assumptions.

[F1]

In holomorphic coordinates and frames, ${\overset{ˉ}{\partial}}_{E}$ and ${\overset{ˉ}{\partial}}_{E^{*}}$ are globally defined, and the bundle-valued Hodge star identifies smooth $(0, 1)$ -forms with smooth $E^{*}$ -valued $(1, 0)$ -forms. Smooth compactly supported forms are dense in $L_{q}^{2}$ (Holomorphic line bundles and meromorphic sections on a Riemann surface, Hermitian metric and $L^{2}$ pairing on a compact Riemann surface).

[F2]

The first-variable-linear $L^{2}$ pairings are Hilbert pairings and satisfy Cauchy–Schwarz; the density integral of an exact compactly supported top form on a boundaryless manifold is zero (The complex $L^{2}$ pairing on equivalence classes, Complex completeness, density, and inner product: the consumer interface, The complex $L^{2}$ pairing is well-defined and satisfies Cauchy–Schwarz, The general Stokes theorem, A compactly supported primitive has zero total derivative integral).

[F3]

A densely defined Hilbert-space operator has a unique Hilbert adjoint; its adjoint is closed and satisfies $ran (T - z)^{⊥} = \ker (T^{*} - \overline{z})$ . A closable densely defined operator has dense adjoint domain and double adjoint equal to its closure (Adjoint of a densely defined operator, The adjoint is well defined, closed, and reverses inclusions, Closability is equivalent to density of the adjoint domain).

[F4]

A closed operator's graph norm is complete, so its domain with $⟨ x, y ⟩_{T} = ⟨ x, y ⟩ + ⟨ T x, T y ⟩$ is a Hilbert space; every bounded linear functional on a Hilbert space has a Riesz representative (Densely defined, closed and closable operators, and cores, Unbounded linear operators: domain, graph and extension, Hilbert space, Riesz representation for Hilbert spaces, The Axiom of Countable Choice ( ${A C}_{ω}$ )).

[F5]

For a densely defined symmetric operator $T$ , surjectivity of both $T - i$ and $T + i$ implies self-adjointness; for a linear subspace $M$ of a Hilbert space, $M^{⊥ ⊥} = \overline{M}$ (Symmetric, self-adjoint and essentially self-adjoint operators, Range criterion for self-adjointness, Orthogonality and the orthogonal complement, The double orthogonal complement of a subspace is its closure).

[F6]

The Euclidean $W^{k, 2}$ norm is the finite sum of the $L^{2}$ norms of weak derivatives; restriction to open sets and multiplication by smooth cutoffs are bounded, smooth coordinate changes obey the chain rule, and smooth partitions of unity exist (Integer-order Sobolev spaces and their norms, The notation $H^{k}$ and the reserved zero-boundary symbol, Bounded restriction and cutoff localisation in Sobolev spaces, The chain rule for differentials of smooth maps, Smooth partitions of unity exist on manifolds).

[F7]

Locally integrable weak derivatives are unique as almost-everywhere classes, and continuous functions and their derivatives are bounded on compact supports (Uniqueness of a weak derivative as an almost-everywhere class, A continuous real-valued function on a nonempty compact metric space is bounded and attains a greatest and a least value).

Proof

technique · define the weak operator by duality, use graph Hilbert spaces for its adjoint and Laplacian, and compare finite local Sobolev norms

1.1F1F2givenalgebra

For smooth $u \in C^{\infty} (X, E)$ and a smooth test form $φ \in C_{c}^{\infty} (X, K \otimes E^{*})$ , the product $u \land φ$ contracts to a degree-one form. On a curve its exterior derivative is its $\overset{ˉ}{\partial}$ part because the $(2, 0)$ component vanishes. Stokes and the graded Leibniz rule give $\int_{X} {\overset{ˉ}{\partial}}_{E} u \land φ = - \int_{X} u \land {\overset{ˉ}{\partial}}_{E^{*}} φ$ , so smooth $u$ has weak derivative ${\overset{ˉ}{\partial}}_{E} u$ . If two $L_{1}^{2}$ forms $v, v^{'}$ satisfy the same weak identity, their difference pairs to zero with every test. By [F1], every test is $⋆_{E} w$ for a smooth $w$ ; the star identity turns that pairing into $⟨ v - v^{'}, w ⟩_{L^{2}}$ . Density of smooth $w$ in $L_{1}^{2}$ implies $v = v^{'}$ . Thus the weak derivative is unique and $\overset{ˉ}{D}$ extends ${\overset{ˉ}{\partial}}_{E}$ .

1.2F3F4F5given

Let $B : H_{0} \to H_{1}$ be closed and densely defined between Hilbert spaces, with $B^{*}$ closed and densely defined, and set $T = B^{*} B$ on $M = {x \in dom B : B x \in dom B^{*}}$ . The graph inner product $⟨ x, y ⟩_{B} = ⟨ x, y ⟩_{H_{0}} + ⟨ B x, B y ⟩_{H_{1}}$ makes $dom B$ Hilbert by [F4]. For any $f \in H_{0}$ , Riesz applied to $x \mapsto ⟨ x, f ⟩$ gives $u \in dom B$ with $⟨ x, u ⟩ + ⟨ B x, B u ⟩ = ⟨ x, f ⟩$ for every $x \in dom B$ . Hence $⟨ B x, B u ⟩ = ⟨ x, f - u ⟩$ , so $B u \in dom B^{*}$ and $B^{*} B u = f - u$ ; therefore $u \in M$ and $(T + I) u = f$ , proving $ran (T + I) = H_{0}$ . If $y ⊥ M$ , apply this construction with $f = y$ to obtain $u \in M$ and $(T + I) u = y$ . Then $0 = ⟨ u, y ⟩ = ⟨ u, (T + I) u ⟩ = ∥ u ∥^{2} + ∥ B u ∥^{2}$ , so $u = y = 0$ ; hence $M^{⊥} = 0$ and $M$ is dense by [F5]. For $x, y \in M$ , $⟨ T x, y ⟩ = ⟨ B x, B y ⟩ = ⟨ x, T y ⟩$ ; also $⟨ T x, x ⟩ = ∥ B x ∥^{2}$ , so $T$ is symmetric and nonnegative. If $x_{n} \to x$ and $T x_{n} \to y$ , then $∥ B (x_{n} - x_{m}) ∥^{2} = ⟨ x_{n} - x_{m}, T (x_{n} - x_{m}) ⟩ \leq ∥ x_{n} - x_{m} ∥ ∥ T x_{n} - T x_{m} ∥$ , making $B x_{n}$ Cauchy. Closedness of $B$ gives $B x_{n} \to B x$ , and closedness of $B^{*}$ applied to $(B x_{n}, T x_{n})$ gives $x \in M$ and $T x = y$ . Thus $T$ is closed.

1.3F6F7given

With the fixed chart/frame/partition data, each coefficient norm in the statement is the Euclidean $W^{k, 2}$ norm of a compactly supported coefficient. On each nonempty compact support, the smooth positive metric and volume weights are bounded above and below, so these local norms are equivalent to the same coefficient norms measured against the Riemannian density; empty supports contribute zero. For a second choice, split each first partitioned term over the finitely many second charts meeting its compact support and insert the second partition. On each resulting compact overlap, coefficients transform by smooth frame and $d \overset{ˉ}{z}$ factors and by a smooth coordinate change $F$ . For a scalar coefficient $f$ , the first- and second-derivative formulas are $D_{i} (f \circ F) = (D_{a} f \circ F) D_{i} F_{a}$ and $D_{i j} (f \circ F) = (D_{a b} f \circ F) D_{i} F_{a} D_{j} F_{b} + (D_{a} f \circ F) D_{i j} F_{a}$ ; multiplying by the smooth frame and form transitions uses the Leibniz rule. All transition derivatives through order two are bounded on these compact supports by [F7], and [F6] gives bounded restriction and cutoff maps. Comparing the invariant density norms, summing finitely many terms gives $∥ u ∥_{H^{k}, 1} \leq C ∥ u ∥_{H^{k}, 2}$ for $k = 1, 2$ ; reversing the choices gives the converse. Thus the norms are equivalent and their completions have the same smooth-form identification.

2.1F3F5step 1.2given

If $y \in \ker (T^{*} - i)$ , choose $x \in M$ with $(T + I) x = y$ using the surjectivity in step 1.2. Since $T^{*} y = i y$ , the first-variable-linear convention gives $∥ y ∥^{2} = ⟨ (T + I) x, y ⟩ = (1 - i) ⟨ x, y ⟩$ , while $⟨ x, y ⟩ = ⟨ x, (T + I) x ⟩ = ∥ B x ∥^{2} + ∥ x ∥^{2}$ is real and nonnegative. Therefore $y = 0$ . The same argument with $- i$ gives $\ker (T^{*} + i) = 0$ . For $z = \pm i$ , symmetry gives $∥ (T - z) x ∥^{2} = ∥ T x ∥^{2} + ∥ x ∥^{2}$ ; if $(T - z) x_{n}$ converges, this estimate makes $x_{n}$ Cauchy and then $T x_{n}$ converges, so closedness of $T$ makes $ran (T - z)$ closed. By [F3], the orthogonal complements of these two ranges are the opposite deficiency kernels, hence both ranges are dense and therefore equal $H_{0}$ . The range criterion [F5] proves that $B^{*} B$ is self-adjoint.

2.2F1F2step 1.1given

Smooth forms are dense in $L_{0}^{2}$ by [F1], and step 1.1 shows they lie in $dom \overset{ˉ}{D}$ , so this domain is dense. If $u_{n} \to u$ in $L_{0}^{2}$ and $\overset{ˉ}{D} u_{n} \to v$ in $L_{1}^{2}$ , then for every smooth test $φ$ , Cauchy–Schwarz makes the wedge pairings continuous and gives $\int_{X} v \land φ = \lim_{n} \int_{X} \overset{ˉ}{D} u_{n} \land φ = - \lim_{n} \int_{X} u_{n} \land {\overset{ˉ}{\partial}}_{E^{*}} φ = - \int_{X} u \land {\overset{ˉ}{\partial}}_{E^{*}} φ$ . Thus $v$ is the weak derivative of $u$ , so the graph of $\overset{ˉ}{D}$ is closed.

2.3F1F6F7step 1.3

In the fixed chart data, each local $W^{k, 2}$ norm contains the local $L^{2}$ norm, and the positive metric weights on compact supports give $∥ u ∥_{L^{2}} \leq C ∥ u ∥_{H^{1}}$ ; the derivative sums give $∥ u ∥_{H^{1}} \leq C^{'} ∥ u ∥_{H^{2}}$ for smooth $u$ . These bounds extend the identity to continuous maps $H^{2} \to H^{1} \to L^{2}$ . They are injective: if an $H^{k}$ Cauchy sequence of smooth forms tends to zero in $L^{2}$ , each local coefficient tends to zero in $L^{2}$ while its derivatives through order $k$ have $L^{2}$ limits. Integration by parts against compactly supported smooth tests shows every such limit is a weak derivative of zero, hence vanishes by [F7]. The $H^{k}$ norm limit is therefore zero. Smooth forms are dense by the definition as completion.

3.1F3step 2.2given

Put $H = L_{0}^{2} \oplus L_{1}^{2}$ and define the single-space operator $A (u, w) = (0, \overset{ˉ}{D} u)$ on $dom A = dom \overset{ˉ}{D} \oplus L_{1}^{2}$ . By step 2.2, $A$ is densely defined and closed. For $(a, b) \in H$ , the adjoint identity $⟨ A (u, w), (a, b) ⟩_{H} = ⟨ (u, w), A^{*} (a, b) ⟩_{H}$ for all $(u, w) \in dom A$ holds exactly when $b \in dom {\overset{ˉ}{D}}^{*}$ and $A^{*} (a, b) = ({\overset{ˉ}{D}}^{*} b, 0)$ : vary $w$ first, then apply the adjoint identity on $L_{0}^{2}, L_{1}^{2}$ . If $b_{n} \to b$ and ${\overset{ˉ}{D}}^{*} b_{n} \to c$ , closedness of $A^{*}$ applied to $(0, b_{n}) \to (0, b)$ proves that ${\overset{ˉ}{D}}^{*}$ is closed. Since $A$ is closable, [F3] makes $dom A^{*} = L_{0}^{2} \oplus dom {\overset{ˉ}{D}}^{*}$ dense in $H$ , so $dom {\overset{ˉ}{D}}^{*}$ is dense in $L_{1}^{2}$ . Also $\ker A^{*} = ran (A)^{⊥}$ by [F3]; with $ran A = {0} \oplus ran \overset{ˉ}{D}$ and $A^{*} (a, b) = ({\overset{ˉ}{D}}^{*} b, 0)$ , intersecting with ${0} \oplus L_{1}^{2}$ yields $\ker {\overset{ˉ}{D}}^{*} = (ran \overset{ˉ}{D})^{⊥}$ . Finally, $A^{* *} = A$ because $A$ is closed; the same block calculation gives $A^{* *} (u, w) = (0, ({\overset{ˉ}{D}}^{*})^{*} u)$ , hence $({\overset{ˉ}{D}}^{*})^{*} = \overset{ˉ}{D}$ .

4.1F3F5step 1.2step 2.1step 3.1

Apply steps 1.2 and 2.1 to $B = \overset{ˉ}{D}$ and to $B = {\overset{ˉ}{D}}^{*}$ . Step 3.1 gives the closed dense adjoints needed for both applications and $({\overset{ˉ}{D}}^{*})^{*} = \overset{ˉ}{D}$ , so both ${\overset{ˉ}{D}}^{*} \overset{ˉ}{D}$ on $L_{0}^{2}$ and $\overset{ˉ}{D} {\overset{ˉ}{D}}^{*}$ on $L_{1}^{2}$ are self-adjoint nonnegative operators. Their direct sum on the stated domain is densely defined and symmetric; its shifts by $\pm i$ are onto because the corresponding shifts of each summand are onto, so [F5] makes the direct sum self-adjoint. The adjoint identity gives $⟨ {\overset{ˉ}{D}}^{*} \overset{ˉ}{D} u_{0}, u_{0} ⟩ = ∥ \overset{ˉ}{D} u_{0} ∥^{2}$ and $⟨ \overset{ˉ}{D} {\overset{ˉ}{D}}^{*} u_{1}, u_{1} ⟩ = ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥^{2}$ . Adding these identities proves the energy formula; it vanishes exactly when $\overset{ˉ}{D} u_{0} = 0$ and ${\overset{ˉ}{D}}^{*} u_{1} = 0$ , which proves the harmonic-kernel description.

5.1F1F2F3F4F5F6F7step 1.1step 1.2step 1.3step 2.1step 2.2step 2.3step 3.1step 4.1∎

Steps 1.1–4.1 establish the unique weak maximal operator, its densely defined closed Hilbert adjoint, the self-adjoint nonnegative block Laplacian and its energy kernel, and the choice-independent Sobolev completions with continuous inclusions. Full AC is used through the stated Sobolev restriction and cutoff localization interface [F6]; the other interfaces named in [F1]–[F5] and the partition construction use only Countable Choice.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Chern connection of a Hermitian holomorphic line bundle

Statement

Let $X$ be a Riemann surface and $E \to X$ a holomorphic line bundle with Hermitian metric $h$ (Holomorphic line bundles and meromorphic sections on a Riemann surface, Hermitian metric and $L^{2}$ pairing on a compact Riemann surface). Identify the complexified cotangent bundle with the Whitney sum $Λ^{1, 0} T^{*} X \oplus Λ^{0, 1} T^{*} X$ , and hence identify the bundle of complex-valued one-forms with values in $E$ as $(Λ^{1, 0} T^{*} X \oplus Λ^{0, 1} T^{*} X) \otimes E$ (Cotangent space and cotangent bundle as a disjoint union, Whitney sums of vector bundles, Whitney sums are smooth vector bundles).

A connection on $E$ is a $C$ -linear map $\nabla : C^{\infty} (X, E) ⟶ C^{\infty} (X, (Λ^{1, 0} \oplus Λ^{0, 1}) T^{*} X \otimes E)$ satisfying $\nabla (f s) = d f \otimes s + f \nabla s$ for $f \in C^{\infty} (X; C)$ and $s \in C^{\infty} (X, E)$ (The exterior derivative of a function is its differential). Its $(1, 0)$ - and $(0, 1)$ -parts $\nabla^{'}$ and $\nabla^{''}$ are the projections to the two summands. Extend $h$ to $E$ -valued one-forms by $⟨ α \otimes u, t ⟩_{h} = α h (u, t)$ and $⟨ s, β \otimes v ⟩_{h} = \overset{ˉ}{β} h (s, v)$ , pairing the bundle factors and conjugating the one-form coefficient in the second argument. The connection is compatible with $h$ when $d (h (s, t)) = ⟨ \nabla s, t ⟩_{h} + ⟨ s, \nabla t ⟩_{h}$ for all smooth sections $s, t$ .

There exists exactly one connection $\nabla_{E}$ such that $\nabla_{E}^{''} = {\overset{ˉ}{\partial}}_{E}$ and $\nabla_{E}$ is compatible with $h$ . It is the Chern connection. In a holomorphic frame $e$ over a holomorphic chart, with $ψ = h (e, e) > 0$ , it is $\nabla_{E} (f e) = (d f + f ψ^{- 1} \partial ψ) \otimes e,$ so $\nabla_{E}^{'} e = ψ^{- 1} \partial ψ \otimes e$ , $\nabla_{E}^{''} e = 0$ , and its connection form is $ω = ψ^{- 1} \partial ψ = \partial \log ψ$ . In another smooth frame $e^{'} = g e$ , $g \in C^{\infty} (U; C^{\times})$ , the full connection form transforms by $ω^{'} = ω + g^{- 1} d g$ .

Facts & Assumptions

Given: A Riemann surface $X$ , a holomorphic line bundle $E \to X$ , and a supplied smooth Hermitian metric $h$ on $E$ .

[F1]

In a holomorphic frame $e$ , a section is $f e$ with smooth coefficient $f$ ; the canonical Dolbeault operator is ${\overset{ˉ}{\partial}}_{E} (f e) = (\overset{ˉ}{\partial} f) \otimes e$ , and holomorphic frame changes are holomorphic nonvanishing functions (Holomorphic line bundles and meromorphic sections on a Riemann surface, Hermitian metric and $L^{2}$ pairing on a compact Riemann surface, Local and global frames of a vector bundle, Smoothness of a section is equivalent to smooth local components).

[F2]

Complex one-forms split into types and $d = \partial + \overset{ˉ}{\partial}$ ; the type projections and their Leibniz rules are coordinate-independent (Bigraded complex forms and the Dolbeault operators, The d, partial and dbar identities, A smooth differential $k$ -form, The wedge product of differential forms).

[F3]

For a smooth function $f$ , its ordinary differential is $d f$ , and the complex chain rule and Wirtinger derivatives give $\partial \log ∣ g ∣^{2} = g^{- 1} \partial g$ for every nonvanishing holomorphic $g$ (The exterior derivative of a function is its differential, The chain rule for complex derivatives, A complex domain is a nonempty connected open subset of $C$ , The Wirtinger derivatives $\partial_{z} f$ and $\partial_{\overset{ˉ}{z}} f$ , and antiholomorphic functions).

[F4]

The cotangent bundle is the disjoint union of its cotangent fibres, and the Whitney sum of the two type bundles is a smooth vector bundle (Cotangent space and cotangent bundle as a disjoint union, Whitney sums of vector bundles, Whitney sums are smooth vector bundles).

[F5]

A compact set inside an open set admits a smooth cutoff equal to one on a neighborhood of that set and supported in the open set (A manifold bump for a compact set inside an open set).

Proof

technique · direct local construction and uniqueness

1.1F1F2F3F4F5given

First, any connection in the Statement is local. If a global section $s$ vanishes near $p$ , choose a smooth cutoff $χ$ supported in that neighborhood and equal to one near $p$ by [F5]. Then $χ s = 0$ and the Leibniz rule at $p$ gives $\nabla s (p) = 0$ . A local section can be multiplied by a cutoff compactly supported in its domain and extended by zero; near any point where the cutoff is one this defines its connection independently of the extension, by locality. Thus the connection and compatibility identities apply to local frames. In a holomorphic coordinate chart and holomorphic frame $e$ , put $ω_{e} = ψ^{- 1} \partial ψ$ and define $\nabla_{E} (f e) = (d f + f ω_{e}) \otimes e$ . The target one-form bundle is smooth by [F4]. This is $C$ -linear and obeys the Leibniz rule. Its $(0, 1)$ -part is $\overset{ˉ}{\partial} f \otimes e = {\overset{ˉ}{\partial}}_{E} (f e)$ , since $ω_{e}$ has type $(1, 0)$ .

2.1F2F3step 1.1givenalgebra

For $s = f e$ and $t = g e$ , the right side of metric compatibility is $(d f + f ω_{e}) \overset{ˉ}{g} ψ + f \overline{(d g + g ω_{e})} ψ$ . Since $ω_{e} + {\overset{ˉ}{ω}}_{e} = ψ^{- 1} d ψ$ , this equals $d (f \overset{ˉ}{g} ψ) = d (h (s, t))$ . Thus the local connection is compatible with $h$ .

3.1F1F3step 1.1step 2.1algebra

If $e^{'} = g e$ is another holomorphic frame, then $ψ^{'} = ∣ g ∣^{2} ψ$ . Since $\partial \overset{ˉ}{g} = 0$ , $\partial ∣ g ∣^{2} = \overset{ˉ}{g} \partial g$ , whence $ω_{e^{'}} = \partial \log ψ^{'} = ω_{e} + g^{- 1} d g$ . For an arbitrary smooth change of frame, the same connection's form transforms by the full Leibniz rule: $\nabla_{E} (g e) = (d g + g ω_{e}) \otimes e = (ω_{e} + g^{- 1} d g) \otimes e^{'}$ . Thus the local formulas agree on holomorphic overlaps, define a global connection with the two required properties, and give the stated smooth-frame transformation.

4.1F1F2F5step 1.1step 2.1algebra∎

If $\nabla$ and $\tilde{\nabla}$ are two connections with the prescribed $(0, 1)$ -part, their difference is $C^{\infty}$ -linear by the Leibniz rule. Since $E$ has rank one, it is multiplication by an $E$ -endomorphism-valued one-form $η$ , and equality of $(0, 1)$ -parts forces $η$ to have type $(1, 0)$ . Subtracting their metric-compatibility identities gives $(η + \overset{ˉ}{η}) h (s, t) = 0$ for all $s, t$ ; taking a local nonzero frame yields $η + \overset{ˉ}{η} = 0$ . The two terms have distinct types, so each vanishes and $η = 0$ . Hence the connection is unique. No choice principle is used.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

The Dolbeault adjoint and Laplacian: local formulas and ellipticity

Statement

Assume the Axiom of Choice, inherited through the completed $L^{2}$ spaces and the Sobolev-localisation interface of The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface; the local calculations below make no new choice. Let $X$ be a compact Riemann surface, $E$ a holomorphic line bundle with Hermitian metric $h$ , and $g$ a compatible Riemannian metric. Use the spaces, maximal operator $\overset{ˉ}{D}$ , Hilbert adjoint ${\overset{ˉ}{D}}^{*}$ , and Dolbeault Laplacian $Δ^{''}$ of The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Let $⋆_{E}$ be the conjugate-linear bundle-valued Hodge map of Hermitian metric and $L^{2}$ pairing on a compact Riemann surface, characterized by $s \land ⋆_{E} t = ⟨ s, t ⟩ d V_{g}$ . In the formula below, $⋆_{E}^{- 1}$ means the inverse of its degree-zero map $E \to Λ^{1, 1} T^{*} X \otimes E^{*}$ .

For a smooth $E$ -valued $(0, 1)$ -form $t$ , define ${\overset{ˉ}{\partial}}_{E}^{*} t : = - ⋆_{E}^{- 1} {\overset{ˉ}{\partial}}_{E^{*}} (⋆_{E} t) .$ Then ${\overset{ˉ}{\partial}}_{E}^{*} t$ is smooth, every smooth $t$ lies in $dom {\overset{ˉ}{D}}^{*}$ , and ${\overset{ˉ}{D}}^{*} t = {\overset{ˉ}{\partial}}_{E}^{*} t$ . In particular, for compactly supported smooth sections $s$ and $(0, 1)$ -forms $t$ , $⟨ {\overset{ˉ}{\partial}}_{E} s, t ⟩_{L^{2}} = ⟨ s, {\overset{ˉ}{\partial}}_{E}^{*} t ⟩_{L^{2}} .$

In a holomorphic chart $z = x + i y$ and holomorphic frame $e$ with $ψ = h (e, e) > 0$ , write $g = ρ (d x^{2} + d y^{2})$ , so $d V_{g} = ρ d x d y$ . For smooth local coefficients $f, u$ , ${\overset{ˉ}{\partial}}_{E} (f e) = \frac{\partial f}{\partial \overset{ˉ}{z}} d \overset{ˉ}{z} \otimes e, {\overset{ˉ}{\partial}}_{E}^{*} (u d \overset{ˉ}{z} \otimes e) = - \frac{2}{ρ ψ} \frac{\partial (ψ u)}{\partial z} e .$ Thus $Δ_{0}^{''} f = - \frac{2}{ρ ψ} \frac{\partial}{\partial z} (ψ \frac{\partial f}{\partial \overset{ˉ}{z}}), Δ_{1}^{''} (u d \overset{ˉ}{z} \otimes e) = - 2 \frac{\partial}{\partial \overset{ˉ}{z}} (\frac{1}{ρ ψ} \frac{\partial (ψ u)}{\partial z}) d \overset{ˉ}{z} \otimes e,$ so both blocks are divergence-form operators with smooth coefficients. The smooth operator $Δ^{''}$ is formally self-adjoint of order $2$ . With the Fourier convention $σ_{F} (\partial_{x}) = i ξ_{x}$ , $σ_{F} (\partial_{y}) = i ξ_{y}$ , its principal symbol on both form degrees is $σ_{F} (Δ^{''}) (x, ξ) = \frac{1}{2 ρ (x)} ∣ ξ ∣_{e u c l}^{2} id = \frac{1}{2} ∣ ξ ∣_{g}^{2} id (ξ \neq 0),$ which is positive definite. Under the scalar-polynomial convention $p_{2} (x, ξ) = \sum_{∣ α ∣ = 2} a_{α} (x) ξ^{α}$ of Principal part and principal symbol of a scalar PDE, the same operator has $p_{2} = - \frac{1}{2} ∣ ξ ∣_{g}^{2} id$ ; this is negative definite and hence also elliptic. The first-order Fourier symbols of ${\overset{ˉ}{\partial}}_{E}$ and ${\overset{ˉ}{\partial}}_{E}^{*}$ have trivial kernel on every nonzero real covector; hence these operators are locally elliptic in the injective-symbol sense.

Facts & Assumptions

Given: A compact Riemann surface $X$ , a holomorphic line bundle $E$ with supplied positive Hermitian metric, a compatible Riemannian metric, and the operators $\overset{ˉ}{D}, {\overset{ˉ}{D}}^{*}, Δ^{''}$ from the preceding item.

[F1]

The bundle Dolbeault operators are globally defined; in a holomorphic frame $e$ , ${\overset{ˉ}{\partial}}_{E} (f e) = (\partial_{\overset{ˉ}{z}} f) d \overset{ˉ}{z} \otimes e$ , and the dual operator extends coefficientwise to $E^{*}$ -valued forms (Holomorphic line bundles and meromorphic sections on a Riemann surface).

[F2]

The bundle star is conjugate-linear, satisfies $s \land ⋆_{E} t = ⟨ s, t ⟩ d V_{g}$ , and in a chart/frame of weight $ψ$ has $⋆_{E} (f e) = \frac{i ρ ψ}{2} \overset{ˉ}{f} d z \land d \overset{ˉ}{z} \otimes e^{*}$ and $⋆_{E} (u d \overset{ˉ}{z} \otimes e) = - i ψ \overset{ˉ}{u} d z \otimes e^{*}$ (Hermitian metric and $L^{2}$ pairing on a compact Riemann surface).

[F3]

The maximal operator's domain is defined by $\int_{X} \overset{ˉ}{D} u \land φ = - \int_{X} u \land {\overset{ˉ}{\partial}}_{E^{*}} φ$ for every smooth test $φ \in C_{c}^{\infty} (X, K \otimes E^{*})$ , and ${\overset{ˉ}{D}}^{*}$ is defined by the first-variable-linear Hilbert adjoint identity (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F4]

The Wirtinger derivatives satisfy $\partial_{z} = \frac{1}{2} (\partial_{x} - i \partial_{y})$ and $\partial_{\overset{ˉ}{z}} = \frac{1}{2} (\partial_{x} + i \partial_{y})$ (The Wirtinger derivatives $\partial_{z} f$ and $\partial_{\overset{ˉ}{z}} f$ , and antiholomorphic functions).

[F5]

For a scalar local differential expression of order $2$ , the principal symbol is the homogeneous polynomial made from its top-order coefficients; a real scalar quadratic symbol is elliptic when it is nonzero for every nonzero real covector (Principal part and principal symbol of a scalar PDE, Elliptic, hyperbolic, and parabolic principal symbols).

[F6]

Full AC and its countable instances are inherited from the completed Hilbert and Sobolev-localisation interfaces stated in the preceding item; the local calculations here use no choice (The Axiom of Choice, The Axiom of Countable Choice ( ${A C}_{ω}$ ), The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F7]

The Chern connection's $(0, 1)$ component is the holomorphic Dolbeault operator; thus Demailly's smooth Chern Dolbeault Laplacian in the cited comparison has the same differential expression as the blocks computed here (Chern connection of a Hermitian holomorphic line bundle).

Proof

technique · use the weak adjoint identity to prove the formal formula, compute its chart expression, then calculate the principal symbols

1.1F1F2F3givenalgebra

For any $u \in dom \overset{ˉ}{D}$ , let $v = \overset{ˉ}{D} u$ and take the smooth test $φ = ⋆_{E} t$ in [F3]. By [F2], $⟨ v, t ⟩_{L^{2}} = \int_{X} v \land ⋆_{E} t$ , and the weak identity gives $⟨ v, t ⟩_{L^{2}} = - \int_{X} u \land {\overset{ˉ}{\partial}}_{E^{*}} ⋆_{E} t$ . The definition of ${\overset{ˉ}{\partial}}_{E}^{*}$ implies $⋆_{E} ({\overset{ˉ}{\partial}}_{E}^{*} t) = - {\overset{ˉ}{\partial}}_{E^{*}} ⋆_{E} t$ , so this is $⟨ u, {\overset{ˉ}{\partial}}_{E}^{*} t ⟩_{L^{2}}$ . Thus $t \in dom {\overset{ˉ}{D}}^{*}$ and ${\overset{ˉ}{D}}^{*} t = {\overset{ˉ}{\partial}}_{E}^{*} t$ ; smooth sections belong to $dom \overset{ˉ}{D}$ and $\overset{ˉ}{D} s = {\overset{ˉ}{\partial}}_{E} s$ , giving the stated formal identity.

2.1F1F2F4step 1.1algebra

In the chart/frame of the statement, solving the $q = 0$ star formula in [F2] for its coefficient gives $⋆_{E}^{- 1} (a d z \land d \overset{ˉ}{z} \otimes e^{*}) = \frac{2 i}{ρ ψ} \overset{ˉ}{a} e$ . For $t = u d \overset{ˉ}{z} \otimes e$ , [F2] gives $⋆_{E} t = - i ψ \overset{ˉ}{u} d z \otimes e^{*}$ , and the local dual Dolbeault formula in [F1] gives ${\overset{ˉ}{\partial}}_{E^{*}} ⋆_{E} t = i \partial_{\overset{ˉ}{z}} (ψ \overset{ˉ}{u}) d z \land d \overset{ˉ}{z} \otimes e^{*}$ . Applying the displayed inverse and conjugating the coefficient yields $⋆_{E}^{- 1} {\overset{ˉ}{\partial}}_{E^{*}} ⋆_{E} t = \frac{2}{ρ ψ} \partial_{z} (ψ u) e$ , so the negative sign in the definition gives the claimed formula for ${\overset{ˉ}{\partial}}_{E}^{*}$ .

3.1F1step 2.1given

The local Dolbeault formula in [F1] gives $Δ_{0}^{''} f = {\overset{ˉ}{\partial}}_{E}^{*} {\overset{ˉ}{\partial}}_{E} f = - \frac{2}{ρ ψ} \partial_{z} (ψ \partial_{\overset{ˉ}{z}} f)$ . Applying ${\overset{ˉ}{\partial}}_{E}$ to the formula from step 2.1 gives $Δ_{1}^{''} (u d \overset{ˉ}{z} \otimes e) = - 2 \partial_{\overset{ˉ}{z}} ((ρ ψ)^{- 1} \partial_{z} (ψ u)) d \overset{ˉ}{z} \otimes e$ . The coefficients are smooth because $ρ, ψ$ are smooth and positive.

4.1step 1.1step 3.1given

For smooth sections $s_{0}, t_{0}$ and smooth $(0, 1)$ -forms $s_{1}, t_{1}$ , the formal-adjoint identity in step 1.1 gives $⟨ Δ_{0}^{''} s_{0}, t_{0} ⟩ = ⟨ {\overset{ˉ}{\partial}}_{E} s_{0}, {\overset{ˉ}{\partial}}_{E} t_{0} ⟩ = ⟨ s_{0}, Δ_{0}^{''} t_{0} ⟩$ and $⟨ Δ_{1}^{''} s_{1}, t_{1} ⟩ = ⟨ {\overset{ˉ}{\partial}}_{E}^{*} s_{1}, {\overset{ˉ}{\partial}}_{E}^{*} t_{1} ⟩ = ⟨ s_{1}, Δ_{1}^{''} t_{1} ⟩$ . Thus the smooth differential operator is formally self-adjoint.

4.2F4F5F7step 3.1algebra

The top-order term of either Laplacian block in step 3.1 is $- \frac{2}{ρ} \partial_{z} \partial_{\overset{ˉ}{z}} = - \frac{1}{2 ρ} (\partial_{x}^{2} + \partial_{y}^{2})$ ; lower-order derivatives of $ψ$ and $ρ$ do not enter the symbol [F4, F5, step 3.1]. With $σ_{F} (\partial_{j}) = i ξ_{j}$ , the Fourier symbol is $σ_{F} (Δ^{''}) (x, ξ) = \frac{1}{2 ρ (x)} (ξ_{x}^{2} + ξ_{y}^{2}) id = \frac{1}{2} ∣ ξ ∣_{g}^{2} id$ . With the scalar-polynomial convention from [F5], $p_{2} = - \frac{1}{2 ρ (x)} (ξ_{x}^{2} + ξ_{y}^{2}) id = - \frac{1}{2} ∣ ξ ∣_{g}^{2} id$ , which is nonzero for $ξ \neq 0$ and has a definite sign. The first-order Fourier symbols are $σ_{F} ({\overset{ˉ}{\partial}}_{E}) (ξ) = \frac{i}{2} (ξ_{x} + i ξ_{y})$ and $σ_{F} ({\overset{ˉ}{\partial}}_{E}^{*}) (ξ) = - \frac{i}{ρ} (ξ_{x} - i ξ_{y})$ in the local line frames, each nonzero for nonzero real $ξ$ . By [F7], the source's Chern-connection Dolbeault operator has this same $(0, 1)$ part; the local computation itself proves the stated ellipticity.

5.1F1F2F3F4F5F6F7step 1.1step 2.1step 3.1step 4.1step 4.2∎

Steps 1.1–4.2 prove the formal adjoint identity, agreement with the Hilbert adjoint on smooth forms, both local Laplacian formulas, formal self-adjointness, and the positive Fourier and negative scalar-polynomial symbols. Full AC is inherited exactly through the preceding maximal-operator item; the local computations themselves use no choice.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface

Statement

Assume the Axiom of Choice, used through the Sobolev restriction and cutoff-localisation interface; the mollification, Hilbert-space, partition, and interior-regularity interfaces use its countable instances (The Axiom of Choice, The Axiom of Countable Choice ( ${A C}_{ω}$ ), Bounded restriction and cutoff localisation in Sobolev spaces). Let $X$ be a nonempty compact Riemann surface, $E \to X$ a holomorphic line bundle with Hermitian metric $h$ , and $g$ a compatible Riemannian metric. Use the maximal Dolbeault operator $\overset{ˉ}{D} : L_{0}^{2} \to L_{1}^{2}$ , its Hilbert adjoint ${\overset{ˉ}{D}}^{*} : L_{1}^{2} \to L_{0}^{2}$ , and the block Dolbeault Laplacian $Δ^{''}$ from The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Write a total form as $u = u_{0} + u_{1}$ , with $u_{q} \in L_{q}^{2}$ , and set $V : = dom \overset{ˉ}{D} \oplus dom {\overset{ˉ}{D}}^{*}, ∥ u ∥_{V}^{2} : = ∥ u ∥_{L^{2}}^{2} + ∥ \overset{ˉ}{D} u_{0} ∥_{L^{2}}^{2} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥_{L^{2}}^{2} .$ For each integer $k \geq 0$ , $H^{k} (X, Λ^{0, ∙} T^{*} X \otimes E)$ denotes the finite-chart Sobolev completion using the same norm formula as in the preceding item, for the fixed finite chart/frame cover and partition used there. For $k = 1, 2$ this is exactly its convention.

First-order estimate and domain. One has $V = H^{1} (X, Λ^{0, ∙} T^{*} X \otimes E), ∥ u ∥_{H^{1}} \leq C (∥ \overset{ˉ}{D} u_{0} ∥_{L^{2}} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥_{L^{2}} + ∥ u ∥_{L^{2}}) .$ The displayed right-hand norm is equivalent to the $H^{1}$ norm. In particular, $V$ is a Hilbert space in its graph norm and smooth forms are dense in it in that norm.
Second-order and higher estimates. Suppose $u \in H^{1} (X, Λ^{0, ∙} T^{*} X \otimes E)$ and $Δ^{''} u = f$ distributionally, where $f \in H^{k} (X, Λ^{0, ∙} T^{*} X \otimes E)$ and $k \geq 0$ . Then $u \in H^{k + 2}$ and for a constant $C_{k}$ , depending on $k, X, g, h$ and the fixed finite-chart norms,

∥ u ∥_{H^{k + 2}} \leq C_{k} (∥ f ∥_{H^{k}} + ∥ u ∥_{L^{2}}) .

For $k = 0$ this is the second-order Gårding estimate. In particular it applies to every $u \in dom Δ^{''}$ , with $f = Δ^{''} u$ . Distributionally means that the local scalar differential expressions of the two Laplacian blocks equal the local coefficients of $f$ . Equivalently, for all smooth test forms $v = v_{0} + v_{1}$ ,

⟨ \overset{ˉ}{D} u_{0}, \overset{ˉ}{D} v_{0} ⟩_{L^{2}} + ⟨ {\overset{ˉ}{D}}^{*} u_{1}, {\overset{ˉ}{D}}^{*} v_{1} ⟩_{L^{2}} = ⟨ f, v ⟩_{L^{2}} .

The form inner product with the additional term $⟨ u, v ⟩_{L^{2}}$ represents $I + Δ^{''}$ , not $Δ^{''}$ .

Facts & Assumptions

Given: the metrics, maximal operators, Sobolev conventions and choice assumptions in the Statement.

[F1]

In a holomorphic chart $z = x + i y$ and holomorphic frame $e$ with $g = ρ (d x^{2} + d y^{2})$ and $h (e, e) = ψ > 0$ , the local formulas are ${\overset{ˉ}{\partial}}_{E} (f e) = (\partial_{\overset{ˉ}{z}} f) d \overset{ˉ}{z} \otimes e, {\overset{ˉ}{\partial}}_{E}^{*} (a d \overset{ˉ}{z} \otimes e) = - \frac{2}{ρ ψ} \partial_{z} (ψ a) e,$ and $Δ_{0}^{''} f = - \frac{2}{ρ ψ} \partial_{z} (ψ \partial_{\overset{ˉ}{z}} f), Δ_{1}^{''} (a d \overset{ˉ}{z} \otimes e) = - 2 \partial_{\overset{ˉ}{z}} ⁣ ((ρ ψ)^{- 1} \partial_{z} (ψ a)) d \overset{ˉ}{z} \otimes e .$ The formulas agree with the Hilbert adjoint on smooth tests (The Dolbeault adjoint and Laplacian: local formulas and ellipticity).

[F2]

The weak maximal domain records the distributional $\overset{ˉ}{\partial}$ derivative, the Hilbert-adjoint domain records the distributional formal-adjoint expression, the Laplacian is the stated nonnegative block operator, and its energy pairing with smooth tests is the sum in the Statement (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F3]

The $L^{2}$ pairing is first-variable-linear; $C_{c}^{\infty} (R^{2}; C)$ is dense in complex $L^{2} (R^{2})$ , and $L^{2}$ is a Hilbert space (Complex completeness, density, and inner product: the consumer interface). A bounded linear functional on a Hilbert space has a Riesz representative (Riesz representation for Hilbert spaces).

[F4]

The Wirtinger derivatives satisfy $\partial_{\overset{ˉ}{z}} = \frac{1}{2} (\partial_{x} + i \partial_{y})$ and $\partial_{z} = \frac{1}{2} (\partial_{x} - i \partial_{y})$ (The Wirtinger derivatives $\partial_{z} f$ and $\partial_{\overset{ˉ}{z}} f$ , and antiholomorphic functions).

[F5]

Multiplication of distributions by a smooth cutoff obeys the Leibniz rule (Leibniz rule for distributions).

[F6]

Interior mollification approximates $L^{2}$ classes locally and commutes distributionally with constant-coefficient derivatives; Meyers–Serrin gives smooth approximation in finite-order Sobolev spaces (Local smooth approximation in integer-order Sobolev spaces, Meyers–Serrin density on an arbitrary open set).

[F7]

Restriction and smooth cutoffs are bounded on Sobolev spaces (Bounded restriction and cutoff localisation in Sobolev spaces).

[F8]

A smooth partition of unity subordinate to a finite chart cover exists (Smooth partitions of unity exist on manifolds).

[F9]

On a relatively compact chart domain the scalar divergence operator with principal coefficients $a^{i j} = (2 ρ)^{- 1} δ_{i j}$ , smooth lower-order coefficients, and an $H^{1}$ weak solution with datum in $H^{k}$ is uniformly elliptic and satisfies the interior $H^{k + 2}$ estimate (Local weak solutions of a divergence-form operator, Uniformly elliptic divergence-form operators and their sesquilinear forms, Interior $H^{k + 2}$ elliptic regularity).

[F10]

The Axiom of Choice supplies Countable Choice. Full AC enters this item only through the Sobolev cutoff-localisation interface; its countable instances enter through mollification, Hilbert representation and density, partitions and interior elliptic regularity (The Axiom of Choice, The Axiom of Countable Choice ( ${A C}_{ω}$ ), Bounded restriction and cutoff localisation in Sobolev spaces).

[F11]

On a Euclidean chart, $H^{k} = W^{k, 2}$ with the norm made from the $L^{2}$ classes of weak derivatives. The finite-chart global norms in the Statement use these local norms; for $k = 1, 2$ they are the preceding item’s completed spaces (The notation $H^{k}$ and the reserved zero-boundary symbol, Integer-order Sobolev spaces and their norms, The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F12]

The $L^{2}$ pairing integrates the pointwise Hermitian pairing against the Riemannian volume form (Hermitian metric and $L^{2}$ pairing on a compact Riemann surface).

[F13]

A compact set inside an open set admits a smooth cutoff equal to one near it, supported in that open set (A manifold bump for a compact set inside an open set).

[F14]

Differentials of smooth composites obey the chain rule (The chain rule for differentials of smooth maps); repeated application gives the finite-order coordinate formulas used below.

Proof

technique · derive the local first-order estimate and divergence-form expressions, then assemble them on a finite chart cover

1.1F3F4F5F6F13algebra

Let $w \in C_{c}^{\infty} (R^{2}; C)$ . Expanding with [F4] gives $4 ∣ \partial_{\overset{ˉ}{z}} w ∣^{2} = ∣ w_{x} ∣^{2} + ∣ w_{y} ∣^{2} + i (w_{y} \overline{w_{x}} - w_{x} \overline{w_{y}})$ . Integration by parts shows $\int w_{x} \overline{w_{y}} = \int w_{y} \overline{w_{x}}$ , so this integral is real and the cross term integrates to zero. Hence $∥ \nabla w ∥_{L^{2}} = 2 ∥ \partial_{\overset{ˉ}{z}} w ∥_{L^{2}}$ ; the identical calculation gives $∥ \nabla w ∥_{L^{2}} = 2 ∥ \partial_{z} w ∥_{L^{2}}$ . Now let $w \in L^{2} (R^{2})$ have compact support and distributional $\partial_{\overset{ˉ}{z}} w = h \in L^{2}$ . Choose a nonnegative smooth bump supported in the unit ball and positive near zero by [F13], and normalize its positive integral to one. Mollify $w$ with its rescalings $ρ_{ϵ}$ . Testing the weak derivative against the translated smooth kernel gives $\partial_{\overset{ˉ}{z}} w_{ϵ} = h * ρ_{ϵ}$ . Both convolutions converge in $L^{2}$ by the $k = 0$ case of [F6]: for $ϵ \leq 1$ their supports lie in one fixed compact set, so local convergence is global. The smooth identity uniformly bounds each $\partial_{j} w_{ϵ}$ . For every test $φ$ , integration by parts and Cauchy–Schwarz therefore bound $φ \mapsto - \int w \partial_{j} φ$ by $2 ∥ h ∥_{2} ∥ φ ∥_{2}$ . Density [F3] and Riesz [F3] represent each functional by an $L^{2}$ function (conjugating the representative for the bilinear weak-derivative convention); thus $w \in H^{1}$ and $∥ \nabla w ∥_{2} \leq 2 \sqrt{2} ∥ h ∥_{2}$ . Conjugation gives the same conclusion when $\partial_{z} w \in L^{2}$ . For a local $L^{2}$ coefficient with derivative $h$ , apply this compact-support result to $η w$ extended by zero, retaining $\partial_{\overset{ˉ}{z}} (η w) = η h + (\partial_{\overset{ˉ}{z}} η) w$ ; both terms are $L^{2}$ on the compact support.

1.2F1F4F9algebra

Expanding [F1] in $x, y$ , each local Laplacian block has principal part $- \frac{1}{2 ρ} (\partial_{x}^{2} + \partial_{y}^{2})$ ; derivatives of $ρ$ and $ψ$ contribute only smooth lower-order terms. Put $a^{i j} = (2 ρ)^{- 1} δ_{i j}$ . Then the block is $- \partial_{i} (a^{i j} \partial_{j} \cdot) + b^{i} \partial_{i} + c \cdot$ for smooth $b^{i}, c$ , after absorbing the derivatives of $a^{i j}$ into $b^{i}$ . On every relatively compact chart subdomain, positivity of $ρ$ gives a positive lower bound for $a^{i j} ξ_{j} \overline{ξ_{i}} / ∣ ξ ∣^{2}$ , and all coefficient derivatives are bounded there. The distributional equation in the Statement and integration by parts against compactly supported tests make each local coefficient a weak solution in the sense of [F9].

2.1F1F2F5F8F12step 1.1

In a chart/frame let $u_{0} = f e$ and $u_{1} = a d \overset{ˉ}{z} \otimes e$ . The weak maximal-domain identity [F2] gives $\partial_{\overset{ˉ}{z}} f \in L_{l o c}^{2}$ . For $u_{1} \in dom {\overset{ˉ}{D}}^{*}$ , test the adjoint identity against compactly supported smooth sections $ϕ e$ . The $L^{2}$ pairing [F12] and the formal adjoint formula [F1], interpreted distributionally by integration by parts, give ${\overset{ˉ}{D}}^{*} u_{1} = - 2 (ρ ψ)^{- 1} \partial_{z} (ψ a)$ ; hence $\partial_{z} (ψ a) \in L_{l o c}^{2}$ . Choose a partition cutoff $χ$ with compact support in the chart, using [F8]. The distributional product rule [F5] gives $\partial_{\overset{ˉ}{z}} (χ f) = χ \partial_{\overset{ˉ}{z}} f + (\partial_{\overset{ˉ}{z}} χ) f$ and $\partial_{z} (χ ψ a) = χ \partial_{z} (ψ a) + (\partial_{z} χ) ψ a$ . Apply step 1.1 to $χ f$ and, by conjugation, to $χ ψ a$ . Since $ρ, ψ$ and their inverses and first derivatives are bounded on the compact support, [F1, F12] then bounds the local $H^{1}$ norms of both coefficients by their local $L^{2}$ norms and the corresponding coefficients of $\overset{ˉ}{D} u_{0}$ and ${\overset{ˉ}{D}}^{*} u_{1}$ .

2.2F3F6F7F8F9F11F13F14step 1.2algebra

Fix $k \geq 0$ and let $(χ_{j})$ be the fixed partition used in the finite-chart norm of the Statement. Choose nested chart subdomains $U_{j}^{'} ⋐ U_{j}^{''} ⋐ U_{j}$ with $supp χ_{j} \subset U_{j}^{'}$ ; the $U_{j}^{'}$ cover $X$ because $\sum_{j} χ_{j} = 1$ . Apply the interior estimate [F9] to each local equation from step 1.2, for both $q = 0, 1$ . It gives $H^{k + 2} (U_{j}^{'})$ regularity and bounds each local norm by $C_{j} (∥ f ∥_{H^{k} (U_{j}^{''})} + ∥ u ∥_{L^{2} (U_{j}^{''})})$ . To compare these local norms with the fixed global norms at any finite order $r$ , write each coefficient as the finite sum of the partitioned coefficients in the other frames. Repeated chain [F14] and Leibniz rules express each derivative through order $r$ as a finite sum of transformed derivatives through order $r$ , multiplied by smooth transition derivatives. On compact overlaps those factors and coordinate Jacobians are bounded, with the Jacobians bounded away from zero. Changing variables therefore bounds each local $H^{r}$ norm on a relatively compact set by the global $H^{r}$ norm; restriction and [F7] give the converse bounds for partitioned coefficients. These inequalities extend from smooth forms to the completions by testing weak derivatives. Apply this with $r = k$ to control $∥ f ∥_{H^{k} (U_{j}^{''})}$ , and use [F7] to bound $χ_{j} u$ in $H^{k + 2}$ by the local estimate. Each such coefficient is a compactly supported $H^{k + 2}$ class; by [F6], approximate it smoothly in its chart, multiply by a cutoff equal to one near its support from [F13], extend by zero and sum. The comparison just proved makes these global smooth forms converge in the defining $H^{k + 2}$ norm, placing $u$ in that completion. Summing the finite estimates now proves the asserted global bound, including $k = 0$ . The same norm comparison in order $r$ gives a continuous injective inclusion $H^{r} \to L^{2}$ : if a smooth Cauchy sequence has zero $L^{2}$ limit, testing every local derivative against compactly supported tests forces all its derivative limits to zero.

3.1F1F8step 2.1algebra

Choose a finite holomorphic chart/frame cover and a subordinate partition of unity as in [F8]. Summing the finitely many local estimates of step 2.1, and using equivalence of the positive smooth metric and volume weights with Euclidean norms on each compact support, gives $∥ u ∥_{H^{1}} \leq C (∥ \overset{ˉ}{D} u_{0} ∥_{L^{2}} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥_{L^{2}} + ∥ u ∥_{L^{2}})$ . The local formulas [F1] also give the reverse bound of the graph norm by the $H^{1}$ norm.

4.1F1F2F6F7F11step 2.1step 3.1

Step 2.1 shows every $u \in V$ has local $H^{1}$ coefficients, in the finite-chart norm of [F11]. For each of the finitely many partitioned coefficients, Meyers–Serrin [F6] gives smooth approximants in its chart; multiplying them by a compactly supported cutoff equal to one near the coefficient's support preserves convergence by [F7]. Converting these compactly supported coefficients back to sections and summing gives smooth global forms converging to $u$ in the finite-chart $H^{1}$ norm, so $u \in H^{1} (X, Λ^{0, ∙} \otimes E)$ . Conversely, if $u \in H^{1}$ , choose smooth $u_{n} \to u$ in that norm by its completion definition. The first-order formulas [F1] make $\overset{ˉ}{D} u_{n, 0}$ and ${\overset{ˉ}{D}}^{*} u_{n, 1}$ converge in $L^{2}$ ; closedness of the operators [F2] gives $u \in V$ . Together with step 3.1 this proves equality of the spaces, equivalence and completeness of their norms, and density of smooth forms in the graph norm.

5.1F2step 4.1step 2.2

If $u \in dom Δ^{''}$ , [F2] gives $u \in V = H^{1}$ . For every smooth test $v$ , the Hilbert-adjoint identities yield $⟨ Δ^{''} u, v ⟩ = ⟨ \overset{ˉ}{D} u_{0}, \overset{ˉ}{D} v_{0} ⟩ + ⟨ {\overset{ˉ}{D}}^{*} u_{1}, {\overset{ˉ}{D}}^{*} v_{1} ⟩$ ; hence the operator equation is the distributional equation used in step 2.2. Taking $f = Δ^{''} u$ proves the domain corollary. Finally, the first-order form inner product adds $⟨ u, v ⟩_{L^{2}}$ , so it represents $I + Δ^{''}$ , as claimed.

6.1F10step 1.1step 1.2step 2.1step 2.2step 3.1step 4.1step 5.1∎

Steps 1.1–5.1 prove the graph-domain $H^{1}$ estimate, smooth graph-norm density, and the $H^{k + 2}$ estimates for distributional and Hilbert-domain solutions. Full AC is spent only through Sobolev cutoff localisation; the remaining Countable Choice instances are inherited from the cited analytic and Hilbert-space interfaces.

Source notes

Demailly's compact-manifold estimate is stated for a general elliptic operator and explicitly cites Hörmander for the underlying elliptic PDE theory. Hunter's Theorem 4.28 gives the corresponding interior higher-regularity theorem and refers to another source for its detailed proof. Here the higher-order estimate is proved by the finite chart reduction and the library's interior $H^{k + 2}$ theorem; the first-order graph estimate is derived directly from the local Cauchy–Riemann formulas. No claim is made that the cited source passages alone prove the bundle-valued domain statement.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Dolbeault green operator is compact on the orthogonal complement of the kernel

Statement

Assume the Axiom of Choice (The Axiom of Choice). It enters through the Sobolev localization and local compactness interfaces, and it supplies Dependent Choice for the sequential compact-operator criterion; the Lax–Milgram and compact self-adjoint norm-attainment arguments use only Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), AC implies DC implies countable choice). Let $X$ be a compact Riemann surface, let $E \to X$ be a holomorphic line bundle with Hermitian metric $h$ , and let $g$ be a compatible Riemannian metric. Use the Hilbert spaces, maximal Dolbeault operator $\overset{ˉ}{D}$ , Hilbert adjoint ${\overset{ˉ}{D}}^{*}$ , and nonnegative self-adjoint Dolbeault Laplacian $Δ^{''}$ of The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Write $H : = \ker Δ^{''}$ and $H^{⊥}$ for its orthogonal complement in $L^{2} : = L_{0}^{2} \oplus L_{1}^{2}$ . For $u = u_{0} + u_{1}$ set $V : = dom \overset{ˉ}{D} \oplus dom {\overset{ˉ}{D}}^{*}, ∥ u ∥_{V}^{2} : = ∥ u ∥_{L^{2}}^{2} + ∥ \overset{ˉ}{D} u_{0} ∥_{L^{2}}^{2} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥_{L^{2}}^{2} .$ By Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface, $V = H^{1} (X, Λ^{0, ∙} T^{*} X \otimes E)$ with equivalent norms. Give $V$ the first-variable-linear form inner product $a (u, v) : = ⟨ u, v ⟩_{L^{2}} + ⟨ \overset{ˉ}{D} u_{0}, \overset{ˉ}{D} v_{0} ⟩_{L^{2}} + ⟨ {\overset{ˉ}{D}}^{*} u_{1}, {\overset{ˉ}{D}}^{*} v_{1} ⟩_{L^{2}} .$

Boundary operator. There is a unique bounded linear operator $B : L^{2} \to V$ (A bounded linear operator between normed spaces) satisfying $a (B f, v) = ⟨ f, v ⟩_{L^{2}} (f \in L^{2}, v \in V) .$ It obeys $∥ B f ∥_{V} \leq ∥ f ∥_{L^{2}}$ and $∥ B f ∥_{H^{2}} \leq C ∥ f ∥_{L^{2}}$ , lies in $dom Δ^{''}$ , and satisfies $(I + Δ^{''}) B f = f$ . As an operator on $L^{2}$ , $B$ is injective, self-adjoint and positive; $B ∣_{H} = I$ , $\ker (I - B) = H$ , and $B (H^{⊥}) \subseteq H^{⊥}$ .
Compactness. The operator $B : L^{2} \to L^{2}$ is compact.
Green operator. The restriction of $I - B$ to $H^{⊥}$ is boundedly invertible. The operator $G : = B (I - B)^{- 1} : H^{⊥} ⟶ H^{⊥}$ is compact, self-adjoint and positive, has trivial kernel, and obeys $ran G = dom Δ^{''} \cap H^{⊥}, \overline{ran G} = H^{⊥} .$ It satisfies the Green identities $Δ^{''} G f = f (f \in H^{⊥}), G Δ^{''} u = u (u \in dom Δ^{''} \cap H^{⊥}),$ and maps $H^{⊥}$ boundedly into $H^{2}$ .

Facts & Assumptions

Given: The compact Riemann surface, the supplied Hermitian and Riemannian metrics, the operators and spaces in the Statement, and full AC.

[F1]

The pointwise Hermitian $L^{2}$ pairing is first-variable-linear, its completion is a complex Hilbert space, and smooth forms are dense; Hilbert space means a complete inner-product space (Hermitian metric and $L^{2}$ pairing on a compact Riemann surface, Hilbert space).

[F2]

The maximal operator and its adjoint are densely defined and closed, $Δ^{''}$ is self-adjoint and nonnegative with the block-composition domain, and for $u \in dom Δ^{''}$ , $⟨ Δ^{''} u, u ⟩ = ∥ \overset{ˉ}{D} u_{0} ∥_{L^{2}}^{2} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥_{L^{2}}^{2}$ (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F3]

The first-order estimate identifies $V$ with $H^{1}$ , proves smooth graph-norm density, and identifies the form equation against smooth tests with the distributional equation for $Δ^{''}$ . A distributional solution $u \in H^{1}$ of $Δ^{''} u = f \in L^{2}$ lies in $H^{2}$ and obeys $∥ u ∥_{H^{2}} \leq C_{0} (∥ f ∥_{L^{2}} + ∥ u ∥_{L^{2}})$ (Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface).

[F4]

The local formulas for $\overset{ˉ}{D}$ and ${\overset{ˉ}{D}}^{*}$ have smooth coefficients. If $u \in H^{2}$ , then their local coefficients applied to $u$ lie in $H^{1}$ : weak derivatives of an $H^{2}$ coefficient are $H^{1}$ , and multiplication by a smooth coefficient preserves $H^{1}$ by the distributional Leibniz rule. Consequently $\overset{ˉ}{D} u_{0} \in dom {\overset{ˉ}{D}}^{*}$ and ${\overset{ˉ}{D}}^{*} u_{1} \in dom \overset{ˉ}{D}$ , so $u \in dom Δ^{''}$ (The Dolbeault adjoint and Laplacian: local formulas and ellipticity, Leibniz rule for distributions).

[F5]

On a Hilbert space, a bounded coercive sesquilinear form and a bounded conjugate-linear functional have a unique Lax–Milgram solution; if the coercivity constant is $1$ , its form norm is at most the functional norm. A linear map is bounded when a constant controls its output norm by its input norm. Cauchy–Schwarz bounds the form and functional, and Lax–Milgram uses Countable Choice (Bounded, coercive and symmetric sesquilinear forms, The Lax--Milgram theorem, The Axiom of Countable Choice ( ${A C}_{ω}$ ), A bounded linear operator between normed spaces, Cauchy–Schwarz: $∣ ⟨ u, v ⟩ ∣ \leq ∥ u ∥ ∥ v ∥$ , with equality exactly for linearly dependent vectors).

[F6]

A sequence bounded in $W_{l o c}^{1, 2}$ on a Euclidean open set has a subsequence converging in $L_{l o c}^{2}$ under AC (Local $L^{p}$ compactness of $W_{l o c}^{1, p}$ -bounded sequences).

[F7]

A bounded linear operator is compact when every bounded sequence has an image subsequence converging in norm; the sequential characterization assumes DC, and AC implies DC (Compact linear operator, Sequential characterization of compact operators, AC implies DC implies countable choice).

[F8]

For a nonzero compact self-adjoint operator $T$ on a Hilbert space, one of $∥ T ∥$ and $- ∥ T ∥$ is an eigenvalue; self-adjointness and positivity have their bounded-operator meanings, and $∥ T ∥$ is the operator norm. This fact assumes Countable Choice (Norm point of a compact self adjoint operator is an eigenvalue up to sign, Self-adjoint, positive, unitary and normal operators, The operator norm as the least bound and as the unit-sphere or unit-ball supremum, The Axiom of Countable Choice ( ${A C}_{ω}$ )).

[F9]

If $T$ is bounded with $∥ T ∥ < 1$ on a Banach space, $I - T$ has inverse $\sum_{n \geq 0} T^{n}$ with norm at most $(1 - ∥ T ∥)^{- 1}$ (Neumann series and small perturbations of bounded inverses).

[F10]

An orthogonal complement is the subspace of vectors orthogonal to the given set and is closed (Orthogonality and the orthogonal complement, Orthogonal complements are closed).

[F11]

Every closed subspace of a Hilbert space is complete (Closed subspaces of complete metric spaces are complete; the converse under countable choice).

[F12]

Full AC supplies the Countable Choice instances used by Lax–Milgram and norm attainment (The Axiom of Choice, The Axiom of Countable Choice ( ${A C}_{ω}$ )).

[F13]

For a linear subspace of a Hilbert space, its double orthogonal complement is its closure (The double orthogonal complement of a subspace is its closure).

Proof

technique · Solve the positive form equation by Lax–Milgram, use the local Rellich theorem to prove compactness, then invert $I-B$ on the orthogonal complement of the harmonic kernel

1.1F1F3F5F12algebra

The form $a$ is Hermitian and bounded by $∥ u ∥_{V} ∥ v ∥_{V}$ , and $a (u, u) = ∥ u ∥_{V}^{2}$ , so it is coercive with constant $1$ . For each $f \in L^{2}$ , $v \mapsto ⟨ f, v ⟩_{L^{2}}$ is conjugate-linear and has norm at most $∥ f ∥_{L^{2}}$ on $V$ . Since $V$ is Hilbert by [F3], Lax–Milgram [F5] gives a unique $B f \in V$ with $a (B f, v) = ⟨ f, v ⟩$ for every $v \in V$ , and $∥ B f ∥_{V} \leq ∥ f ∥_{L^{2}}$ . Uniqueness makes $f \mapsto B f$ linear, and this estimate makes it bounded into both $V$ and $L^{2}$ .

1.2F2F3F4algebra

For every smooth test form $v$ , the defining equation of $B f$ gives $⟨ \overset{ˉ}{D} (B f)_{0}, \overset{ˉ}{D} v_{0} ⟩ + ⟨ {\overset{ˉ}{D}}^{*} (B f)_{1}, {\overset{ˉ}{D}}^{*} v_{1} ⟩ = ⟨ f - B f, v ⟩$ . By [F3] this is the distributional equation $Δ^{''} B f = f - B f$ ; the $H^{2}$ estimate in [F3] applies because $B f \in V = H^{1}$ and $f - B f \in L^{2}$ , yielding $∥ B f ∥_{H^{2}} \leq C_{0} (∥ f - B f ∥ + ∥ B f ∥) \leq 3 C_{0} ∥ f ∥$ . Fact [F4] now puts $B f$ in $dom Δ^{''}$ , so the distributional identity is the operator identity $(I + Δ^{''}) B f = f$ . Conversely, for $u \in dom Δ^{''}$ , set $f = (I + Δ^{''}) u$ . Then $z = B f - u$ lies in $dom Δ^{''}$ and $(I + Δ^{''}) z = 0$ ; taking its pairing with $z$ and using [F2] gives $0 = ∥ z ∥^{2} + ∥ \overset{ˉ}{D} z_{0} ∥^{2} + ∥ {\overset{ˉ}{D}}^{*} z_{1} ∥^{2}$ , so $z = 0$ and $B (I + Δ^{''}) u = u$ .

1.3F1algebra

The equation with $v = B g$ gives $a (B f, B g) = ⟨ f, B g ⟩$ ; Hermitian symmetry of $a$ then gives $⟨ B f, g ⟩ = ⟨ f, B g ⟩$ , so $B$ is self-adjoint. Taking $g = f$ shows $⟨ B f, f ⟩ = a (B f, B f) \geq 0$ , so $B$ is positive. If $B f = 0$ , its defining equation gives $⟨ f, v ⟩ = 0$ for every $v \in V$ ; smooth forms are dense in $L^{2}$ by [F1], hence $f = 0$ and $B$ is injective. Also $∥ B f ∥_{L^{2}} \leq ∥ B f ∥_{V} \leq ∥ f ∥_{L^{2}}$ , so $∥ B ∥ \leq 1$ .

2.1F1F2F10step 1.3algebra

If $h \in H$ , the energy identity [F2] gives $\overset{ˉ}{D} h_{0} = 0$ and ${\overset{ˉ}{D}}^{*} h_{1} = 0$ , whence $h \in V$ and $a (h, v) = ⟨ h, v ⟩$ for all $v \in V$ ; uniqueness gives $B h = h$ . Conversely, if $B x = x$ , then $x \in V$ and testing its defining equation with $v = x$ gives $∥ x ∥_{V}^{2} = ∥ x ∥_{L^{2}}^{2}$ , so both first-order terms vanish. The block domain in [F2] then gives $x \in dom Δ^{''}$ and $Δ^{''} x = 0$ . Thus $\ker (I - B) = H$ . Self-adjointness and $B ∣_{H} = I$ imply $B (H^{⊥}) \subseteq H^{⊥}$ .

2.2F3F6F7F12step 1.2algebra

Let $(f_{n})$ be bounded in $L^{2}$ and put $u_{n} = B f_{n}$ . Step 1.2 bounds $(u_{n})$ in the fixed finite-chart $H^{2}$ norm, so every partitioned local coefficient is bounded in $W_{l o c}^{1, 2}$ ; [F6] gives a subsequence converging in $L_{l o c}^{2}$ for each chart coefficient. Successively taking subsequences over the finitely many charts and degrees gives one subsequence converging in $L^{2}$ on every compact support of the fixed partition; summing these finitely many weighted coefficient norms gives convergence in global $L^{2}$ . Thus $B$ takes every bounded sequence to a sequence with a norm-convergent subsequence. By [F7] and [F12], the sequential characterization proves that $B : L^{2} \to L^{2}$ is compact.

3.1F1F7F8F9F10F11F12step 1.3step 2.1step 2.2algebra

The closed subspace $H^{⊥}$ is invariant under $B$ by step 2.1. For any bounded sequence in $H^{⊥}$ , compactness of $B$ from step 2.2 gives a subsequence whose images converge in $L^{2}$ ; its limit remains in $H^{⊥}$ by [F10], so [F7] shows that $T : = B ∣_{H^{⊥}}$ is compact on $H^{⊥}$ . This closed subspace is Hilbert by [F1, F11]. It is self-adjoint, positive and has norm at most $1$ by step 1.3. If $T = 0$ , its norm is already less than $1$ ; this includes $H^{⊥} = {0}$ . Otherwise [F8] gives an eigenvalue equal to $∥ T ∥$ or $- ∥ T ∥$ ; positivity excludes the negative value, and if $∥ T ∥ = 1$ its unit eigenvector would lie in $\ker (I - B) \cap H^{⊥} = {0}$ by step 2.1, a contradiction. Hence $r : = ∥ T ∥ < 1$ , and [F9] gives the bounded inverse $C : = (I - T)^{- 1} = \sum_{n \geq 0} T^{n}$ on $H^{⊥}$ , with $∥ C ∥ \leq (1 - r)^{- 1}$ .

4.1F1F7F9step 1.3step 2.1step 3.1algebra

The series $G : = T C = \sum_{n \geq 1} T^{n}$ converges in operator norm by [F9]. Each power $T^{n}$ is self-adjoint and positive: for $n = 2 m$ , $⟨ T^{n} x, x ⟩ = ∥ T^{m} x ∥^{2}$ , and for $n = 2 m + 1$ , it equals $⟨ T (T^{m} x), T^{m} x ⟩ \geq 0$ . Therefore $G$ is self-adjoint and positive. For any bounded sequence $(x_{n})$ in $H^{⊥}$ , boundedness of $C$ makes $(C x_{n})$ bounded; compactness of $T$ from step 3.1 gives a subsequence for which $G x_{n} = T C x_{n}$ converges, so [F7] proves compactness of $G$ . It is injective because both $T$ and $C$ are injective. For $f \in H^{⊥}$ , put $y = C f$ ; step 1.2 gives $G f = B y \in dom Δ^{''} \cap H^{⊥}$ and $Δ^{''} G f = (I - B) y = f$ .

5.1F1F2F9F10F13step 1.2step 4.1algebra∎

If $u \in dom Δ^{''} \cap H^{⊥}$ , self-adjointness of $Δ^{''}$ and $Δ^{''} h = 0$ for $h \in H$ imply $Δ^{''} u \in H^{⊥}$ . Both $u$ and $G Δ^{''} u$ lie in $dom Δ^{''}$ and have the same Laplacian by step 4.1, so their difference belongs to $H \cap H^{⊥} = {0}$ . Thus $G Δ^{''} u = u$ and $ran G = dom Δ^{''} \cap H^{⊥}$ . If $z ⊥ ran G$ , self-adjointness gives $⟨ x, G z ⟩ = 0$ for all $x \in H^{⊥}$ , hence $G z = 0$ and $z = 0$ ; [F13] now gives $\overline{ran G} = H^{⊥}$ . Finally, $∥ G f ∥_{H^{2}} = ∥ B (C f) ∥_{H^{2}} \leq 3 C_{0} ∥ C ∥ ∥ f ∥_{L^{2}}$ by step 1.2 and [F9].

Source notes

Demailly's Ch. VI §2 (2.2) states the compact Rellich inclusion on a compact manifold. The bundle-valued compactness step is assembled chartwise from the library's local compactness theorem. The construction of $B$ , the norm gap on $H^{⊥}$ , and the corrected Green range are established here from the exact operator-theoretic suppliers. Demailly's separate Ch. VI §3.3 Hodge statements assume a flat Hermitian connection and are not used for this result.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Elliptic regularity for Dolbeault harmonic forms

Statement

Assume the Axiom of Choice (The Axiom of Choice). Full AC is used through the Sobolev and smooth-data regularity interfaces; the local interior theorem and Hilbert projection interface use its countable instances (The Axiom of Countable Choice ( ${A C}_{ω}$ ), AC implies DC implies countable choice). Let $X$ be a nonempty compact Riemann surface, let $E \to X$ be a holomorphic line bundle with Hermitian metric $h$ , let $g$ be a compatible Riemannian metric, and use $\overset{ˉ}{D}$ , ${\overset{ˉ}{D}}^{*}$ , $Δ^{''}$ , and the Hilbert spaces $L_{q}^{2}$ from The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Write $Δ_{0}^{''} = {\overset{ˉ}{D}}^{*} \overset{ˉ}{D}$ , $Δ_{1}^{''} = \overset{ˉ}{D} {\overset{ˉ}{D}}^{*}$ , and $H^{0, q} (E) = \ker Δ_{q}^{''}$ . On an open set in $X$ , $H_{l o c}^{k}$ means that the coefficient in every holomorphic chart and frame is locally in the Euclidean $H^{k} = W^{k, 2}$ space (Integer-order Sobolev spaces and their norms, The notation $H^{k}$ and the reserved zero-boundary symbol). For each relatively compact open $U ⋐ Ω$ , the norms $H^{k} (U)$ are the finite chart/frame Sobolev norms; the constants may depend on this fixed finite cover.

Distributionally means that, in every holomorphic chart and frame, the scalar local differential expression for $Δ_{q}^{''} u$ equals the local coefficient of $f$ as a distribution (Weak derivative of a locally integrable function).

Local gain. Let $Ω \subseteq X$ be open, let $q \in {0, 1}$ , and let $u \in H_{l o c}^{1} (Ω, Λ^{0, q} T^{*} X \otimes E)$ satisfy $Δ_{q}^{''} u = f$ distributionally, with $f \in H_{l o c}^{k} (Ω, Λ^{0, q} T^{*} X \otimes E)$ for an integer $k \geq 0$ . Then $u \in H_{l o c}^{k + 2}$ . For open sets $Ω^{'} ⋐ Ω^{''} ⋐ Ω$ , $∥ u ∥_{H^{k + 2} (Ω^{'})} \leq C (∥ f ∥_{H^{k} (Ω^{''})} + ∥ u ∥_{L^{2} (Ω^{''})}),$ where $C$ may depend on $k$ , the nested sets, the fixed local norms, and the metrics.
Harmonic forms. Every $u \in H^{0, q} (E)$ is smooth and satisfies $\overset{ˉ}{D} u = 0$ if $q = 0$ , or ${\overset{ˉ}{D}}^{*} u = 0$ if $q = 1$ . Conversely, a smooth form in degree $q$ that satisfies the corresponding first-order equation belongs to $H^{0, q} (E)$ . For a total form $u = u_{0} + u_{1}$ , $u \in \ker Δ^{''} ⟺ \overset{ˉ}{D} u_{0} = 0 and {\overset{ˉ}{D}}^{*} u_{1} = 0.$
Smooth representatives and projection. $im (\overset{ˉ}{D}) \cap C^{\infty} (X, Λ^{0, 1} T^{*} X \otimes E) = {\overset{ˉ}{\partial}}_{E} (C^{\infty} (X, E)) .$ The Hilbert orthogonal projection $P_{H} : L_{0}^{2} \oplus L_{1}^{2} \to \ker Δ^{''}$ also maps smooth total forms to smooth total forms.

Facts & Assumptions

Given: The compact Riemann surface, supplied metrics, the maximal Dolbeault complex and its self-adjoint nonnegative Laplacian, and full AC.

[F1]

The spaces $L_{0}^{2}, L_{1}^{2}$ are Hilbert spaces; $\overset{ˉ}{D}$ and ${\overset{ˉ}{D}}^{*}$ are closed, densely defined operators; $Δ^{''}$ is self-adjoint and nonnegative on its block-composition domain; and $⟨ Δ^{''} (u_{0}, u_{1}), (u_{0}, u_{1}) ⟩ = ∥ \overset{ˉ}{D} u_{0} ∥^{2} + ∥ {\overset{ˉ}{D}}^{*} u_{1} ∥^{2},$ with $\ker Δ^{''} = (\ker \overset{ˉ}{D}) \oplus (\ker {\overset{ˉ}{D}}^{*})$ (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F2]

In a holomorphic chart and frame, the two scalar blocks have smooth coefficients and principal part $- \frac{1}{2 ρ} (\partial_{x}^{2} + \partial_{y}^{2})$ ; their full formulas are $Δ_{0}^{''} f = - \frac{2}{ρ ψ} \partial_{z} (ψ \partial_{\overset{ˉ}{z}} f), Δ_{1}^{''} (a d \overset{ˉ}{z} \otimes e) = - 2 \partial_{\overset{ˉ}{z}} ((ρ ψ)^{- 1} \partial_{z} (ψ a)) d \overset{ˉ}{z} \otimes e .$ Smooth degree-zero forms lie in $dom \overset{ˉ}{D}$ , and smooth degree-one forms lie in $dom {\overset{ˉ}{D}}^{*}$ (The Dolbeault adjoint and Laplacian: local formulas and ellipticity).

[F3]

The first-order graph domain $V = dom \overset{ˉ}{D} \oplus dom {\overset{ˉ}{D}}^{*}$ equals the finite-chart $H^{1}$ space with equivalent norms; smooth forms are dense in that graph domain (Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface).

[F4]

A divergence-form operator with smooth coefficients is uniformly elliptic on each relatively compact chart patch when its Hermitian principal matrix has a positive lower bound (Uniformly elliptic divergence-form operators and their sesquilinear forms).

[F5]

For such an operator with $a^{i j} \in W_{l o c}^{k + 1, \infty}$ and lower coefficients in $W_{l o c}^{k, \infty}$ , a local weak $H^{1}$ solution with datum in $H_{l o c}^{k}$ lies in $H_{l o c}^{k + 2}$ and satisfies the nested-domain estimate (Interior $H^{k + 2}$ elliptic regularity).

[F6]

With smooth coefficients and smooth datum, every local weak $H^{1}$ solution has a smooth representative on the open set (Smooth data give smooth interior solutions).

[F7]

The spaces $H^{k}$ are the classes with $L^{2}$ weak derivatives through order $k$ , and the local weak derivative is defined by testing against $C_{c}^{\infty}$ functions (Integer-order Sobolev spaces and their norms, The notation $H^{k}$ and the reserved zero-boundary symbol, Weak derivative of a locally integrable function).

[F8]

A closed linear subspace of a Hilbert space has a unique orthogonal decomposition; its subspace component is the Hilbert orthogonal projection (Hilbert space, Orthogonality and the orthogonal complement, Orthogonal decomposition by a closed subspace, The Hilbert orthogonal projection onto a closed subspace).

[F9]

Full AC is used through the Sobolev localization of item 5 and the smooth-data corollary; it supplies the Countable Choice assumed by the local interior theorem and orthogonal projection theorem (The Axiom of Choice, The Axiom of Countable Choice ( ${A C}_{ω}$ ), AC implies DC implies countable choice).

[F10]

A local weak solution is an $H^{1}$ coefficient satisfying the sesquilinear divergence-form identity against every compactly supported smooth test (Local weak solutions of a divergence-form operator).

Proof

technique · Reduce both Dolbeault Laplacian blocks to scalar uniformly elliptic divergence-form equations, then apply the local regularity and smooth-data suppliers

1.1F2F4algebra

In a holomorphic chart $z = x + i y$ and frame $e$ with $g = ρ (d x^{2} + d y^{2})$ and $h (e, e) = ψ > 0$ , [F2] shows that the principal part of either block is $- \frac{1}{2 ρ} (\partial_{x}^{2} + \partial_{y}^{2})$ . Set $a^{i j} = (2 ρ)^{- 1} δ_{i j}$ ; moving the derivatives of the smooth weights $ρ, ψ$ into the first- and zero-order coefficients writes each block as $- \partial_{i} (a^{i j} \partial_{j} \cdot) + b^{i} \partial_{i} + c$ . On every relatively compact chart patch, $ρ$ has positive minimum and all metric/frame coefficient derivatives are bounded, so this matrix is uniformly elliptic and all coefficients are smooth.

1.2F1algebra

If $u = u_{0} + u_{1} \in \ker Δ^{''}$ , the energy identity [F1] is a sum of two nonnegative squared norms and equals zero, so $\overset{ˉ}{D} u_{0} = 0$ and ${\overset{ˉ}{D}}^{*} u_{1} = 0$ . Conversely, if these two terms vanish, then their zero outputs lie in the opposite operator domains, so $u$ lies in the block domain of $Δ^{''}$ and $Δ^{''} u = 0$ . The same argument in each degree gives the stated degreewise characterization.

1.3F1F2F3F6F7F9F10

Let $v \in im \overset{ˉ}{D} \cap C^{\infty} (X, Λ^{0, 1} T^{*} X \otimes E)$ and choose $u \in dom \overset{ˉ}{D}$ with $\overset{ˉ}{D} u = v$ . By [F3], $u \in H^{1}$ . The defining weak identity for $\overset{ˉ}{D}$ in [F1] says that the local coefficient derivative of $u$ is the corresponding ${\overset{ˉ}{\partial}}_{E}$ expression in distributions. Since $v$ is smooth, [F2] gives $v \in dom {\overset{ˉ}{D}}^{*}$ with the stated smooth local formula for ${\overset{ˉ}{D}}^{*} v$ ; therefore $u \in dom Δ_{0}^{''}$ and $Δ_{0}^{''} u = {\overset{ˉ}{D}}^{*} v$ is exactly the local scalar block equation in distributions. Integration by parts against compactly supported tests gives the local weak identity of [F10]. Apply the smooth-data corollary [F6] in each chart. Its smooth representatives agree on chart overlaps because they represent the same section $u$ almost everywhere and are continuous, so they glue to a global smooth section $\tilde{u}$ . Weak differentiation depends only on the almost-everywhere class by [F7], hence $v = {\overset{ˉ}{\partial}}_{E} \tilde{u}$ . Conversely, every smooth section belongs to $dom \overset{ˉ}{D}$ and its Hilbert derivative is ${\overset{ˉ}{\partial}}_{E}$ by [F1, F2]. This proves the equality of smooth exact representatives. AC is used by [F6] as recorded in [F9].

2.1F2F4F7F10step 1.1

If $Δ_{q}^{''} u = f$ distributionally, the local coefficient equation from step 1.1 holds against every compactly supported smooth test. Integration by parts in the divergence term gives precisely the local weak identity of [F10]; the coefficients and the datum belong to its stated classes. Thus each local coefficient of $u$ is a local weak solution of a uniformly elliptic divergence-form equation.

2.2F1F2F3F6F9F10step 1.1step 1.2

A harmonic component belongs to $V$ by the block domain in [F1], and hence to $H^{1}$ by [F3]. Its local equation has smooth coefficients and smooth datum $f = 0$ by step 1.1; the local weak-solution definition [F10] applies, and [F6] gives a smooth representative in each chart. These representatives agree on overlaps because they represent the same global $L^{2}$ form almost everywhere and are smooth, so they give a global smooth representative. Conversely, [F1, F2] put every smooth degree-zero form in $dom \overset{ˉ}{D}$ and every smooth degree-one form in $dom {\overset{ˉ}{D}}^{*}$ . If its corresponding first-order derivative vanishes, the zero output lies in the other operator's domain, so step 1.2 puts the form in the Hilbert kernel. AC is used by [F6] as recorded in [F9].

3.1F4F5F7F9F10step 1.1step 2.1algebra

Fix $k \geq 0$ and $Ω^{'} ⋐ Ω^{''} ⋐ Ω$ . Cover $\overline{Ω^{'}}$ by finitely many chart patches $U_{j}^{'}$ and choose larger chart patches $U_{j}^{''}$ with $\overline{U_{j}^{'}} \subset U_{j}^{''} ⋐ Ω^{''}$ ; choose domains compactly contained in $Ω$ that contain each $\overline{U_{j}^{''}}$ . The local $H^{1}$ hypothesis and equation restrict to those domains by [F7, F10]. Apply [F5] with the smooth coefficients from step 1.1 on each nested chart domain. Summing the finite estimates, with the smooth metric and frame weights bounded above and below on the compact supports, gives the stated $H^{k + 2} (Ω^{'})$ estimate and local regularity. AC supplies the countable-choice hypothesis of [F5] by [F9].

3.2F1F8F9step 2.2

The harmonic space $H = \ker Δ^{''}$ is closed: if $h_{n} \in H$ and $h_{n} \to h$ in $L^{2}$ , then $Δ^{''} h_{n} = 0 \to 0$ , and closedness of the self-adjoint operator [F1] gives $h \in dom Δ^{''}$ with $Δ^{''} h = 0$ . Thus [F8] defines the orthogonal projection $P_{H}$ . For any smooth total form $w$ , $P_{H} w \in H$ , and step 2.2 shows that every element of $H$ is smooth. Therefore $P_{H}$ maps smooth forms to smooth forms, without using the later Hodge decomposition. AC supplies the projection theorem's countable-choice assumption by [F9].

4.1F5F6F8F9step 1.1step 1.2step 1.3step 2.1step 2.2step 3.1step 3.2∎

Steps 1.1 and 2.1 establish the local scalar equations and weak formulation; step 3.1 proves the nested $H^{k + 2}$ gain; steps 1.2 and 2.2 prove the harmonic characterization and smoothness; and steps 1.3 and 3.2 prove smooth exact preimages and smoothness of the harmonic projection. Full AC is used through [F6], with AC $_{ω}$ for [F5] and [F8], as stated in [F9].

Source notes

Hunter's Theorem 4.28 (printed p. 114) states higher interior regularity for uniformly elliptic divergence-form equations and explicitly refers to [9] for a detailed proof; Corollary 4.29 bootstraps smooth coefficients and data through Sobolev embedding to smoothness. This item uses the library's fully proved nested-domain theorem and smooth-data corollary for the bundle-valued equations. The formal Dolbeault adjoint and both block formulas are supplied by the preceding local-formula item, not inferred from a flat-connection Hodge theorem.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

The Dolbeault Laplacian has finite-dimensional kernel and closed range

Statement

Assume the Axiom of Choice (The Axiom of Choice). It is inherited through the construction of the boundary and Green operators in Dolbeault green operator is compact on the orthogonal complement of the kernel. It supplies Dependent Choice for the closed-range theorem and Countable Choice for the Hilbert orthogonal-decomposition theorem (The Axiom of Countable Choice ( ${A C}_{ω}$ ), AC implies DC implies countable choice). Let $X$ be a nonempty compact Riemann surface, let $E \to X$ be a holomorphic line bundle with Hermitian metric $h$ , and let $g$ be a compatible Riemannian metric. Use the Hilbert spaces and block Dolbeault Laplacian $Δ^{''}$ from The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Put $H : = \ker Δ^{''} \subset L^{2} : = L_{0}^{2} \oplus L_{1}^{2}$ . Let $B : L^{2} \to L^{2}$ and $G : H^{⊥} \to H^{⊥}$ be the compact boundary and Green operators from Dolbeault green operator is compact on the orthogonal complement of the kernel; in particular, $B$ is compact and self-adjoint, $\ker (I - B) = H$ , and $(I + Δ^{''}) B = I on L^{2}, B (I + Δ^{''}) u = u (u \in dom Δ^{''}) .$ The Green identities are $Δ^{''} G f = f$ for $f \in H^{⊥}$ and $G Δ^{''} u = u$ for $u \in dom Δ^{''} \cap H^{⊥}$ . The space $H^{2}$ below is the finite-chart Sobolev space of Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface.

Finite-dimensional harmonic space. $H$ is finite dimensional.
Closed range and orthogonal decomposition. The range $Δ^{''} (dom Δ^{''})$ is closed and equals $H^{⊥}$ , so $L^{2} = H \oplus Δ^{''} (dom Δ^{''})$ orthogonally. The range of $Δ^{''}$ on each summand $L_{q}^{2}$ , $q = 0, 1$ , is closed in that summand.
Quantitative inverse. There is a constant $c > 0$ such that $∥ Δ^{''} u ∥_{L^{2}} \geq c ∥ u ∥_{L^{2}} (u \in H^{⊥} \cap dom Δ^{''}) .$ Moreover, $Δ^{''}$ restricts to a topological isomorphism $H^{⊥} \cap dom Δ^{''} ⟶ H^{⊥}$ with inverse $G$ . On its domain use the graph norm $∥ u ∥_{H^{2}, Δ} : = ∥ u ∥_{H^{2}} + ∥ Δ^{''} u ∥_{L^{2}} .$

Facts & Assumptions

Given: The compact Riemann surface and supplied metrics; the maximal Dolbeault complex, its total Hilbert space, and its block Laplacian; the boundary and Green operators of the preceding lemma; and full AC.

[F1]

The total space $L_{0}^{2} \oplus L_{1}^{2}$ is a complex Hilbert space and hence a Banach space (Hilbert space).

[F2]

The Laplacian is the nonnegative self-adjoint block operator $Δ_{0}^{''} \oplus Δ_{1}^{''}$ on the direct-sum composition domain (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F3]

$B$ is compact, bounded and self-adjoint, $\ker (I - B) = H$ , and $(I + Δ^{''}) B = I$ on $L^{2}$ while $B (I + Δ^{''}) = I$ on $dom Δ^{''}$ . The Green operator $G$ is bounded into $H^{2}$ , has range $dom Δ^{''} \cap H^{⊥}$ , and satisfies both Green identities (Dolbeault green operator is compact on the orthogonal complement of the kernel).

[F4]

If $K$ is compact on a normed space, then $\ker (I - K)$ is finite dimensional (Kernel of identity minus compact is finite dimensional).

[F5]

If $K$ is compact on a Banach space, then $ran (I - K)$ is closed under DC (Range of identity minus compact is closed).

[F6]

For a bounded self-adjoint operator $T$ on a Hilbert space, $⟨ T x, y ⟩ = ⟨ x, T y ⟩$ for all $x, y$ ; the identity operator is self-adjoint (Self-adjoint, positive, unitary and normal operators).

[F7]

For a subset $S$ of a Hilbert space, $S^{⊥}$ consists of the vectors orthogonal to every element of $S$ ; every closed subspace of a Hilbert space has a unique orthogonal decomposition (Orthogonality and the orthogonal complement, Orthogonal decomposition by a closed subspace).

[F8]

The finite-chart $H^{2}$ norm is the Sobolev norm in the preceding Gårding theorem; the inclusion $H^{2} ↪ L^{2}$ is continuous, and $G : H^{⊥} \to H^{2}$ is bounded (Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface, The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface, Dolbeault green operator is compact on the orthogonal complement of the kernel).

[F9]

Full AC implies DC and Countable Choice (The Axiom of Choice, The Axiom of Countable Choice ( ${A C}_{ω}$ ), AC implies DC implies countable choice).

Proof

technique · Use compact-perturbation facts for $I-B$ and the exact Green identities for $\Delta''$

1.1F1F3F4

By [F3], $H = \ker (I - B)$ . Apply the compact-kernel lemma [F4] to $B$ on the normed space $L^{2}$ from [F1]. Thus $H$ is finite dimensional.

1.2F1F3F5F9algebra

Set $A : = I - B$ . For $f \in L^{2}$ , $(I + Δ^{''}) B f = f$ , so $(I - B) f = Δ^{''} B f$ . For $u \in dom Δ^{''}$ , $B (I + Δ^{''}) u = u$ , so $(I - B) (I + Δ^{''}) u = (I + Δ^{''}) u - u = Δ^{''} u$ . Both containments give $ran Δ^{''} = ran A$ . The space $L^{2}$ is Banach by [F1], and AC supplies DC for [F5]; hence this range is closed.

1.3F3F8algebra

If $H^{⊥} = {0}$ , the displayed estimate holds with $c = 1$ . Otherwise boundedness of $G : H^{⊥} \to H^{2}$ and the continuous inclusion $H^{2} ↪ L^{2}$ give some $M > 0$ with $∥ G f ∥_{L^{2}} \leq M ∥ f ∥_{L^{2}}$ for all $f \in H^{⊥}$ . For $u \in H^{⊥} \cap dom Δ^{''}$ , the Green identity gives $u = G Δ^{''} u$ , whence $∥ u ∥_{L^{2}} \leq M ∥ Δ^{''} u ∥_{L^{2}}$ . Taking $c = M^{- 1}$ proves the estimate in this case.

2.1F3F6F7F9step 1.2

The operator $A = I - B$ is bounded and self-adjoint by [F3, F6]. If $y ⊥ ran A$ , then $⟨ x, A y ⟩ = ⟨ A x, y ⟩ = 0$ for every $x \in L^{2}$ , so $A y = 0$ ; conversely, $A y = 0$ implies $y ⊥ ran A$ . Therefore $(ran A)^{⊥} = \ker A = H$ by [F3]. Apply [F7] to the closed subspace $ran A$ from step 1.2: $L^{2} = ran A \oplus H$ . Every vector in $ran A$ is orthogonal to $H$ ; if $z \in H^{⊥}$ and $z = r + h$ is this decomposition, then $h = z - r \in H \cap H^{⊥} = {0}$ . Thus $ran A = H^{⊥}$ , and step 1.2 gives $ran Δ^{''} = H^{⊥}$ . AC supplies the theorem's countable-choice hypothesis by [F9].

2.2F2step 1.2

By [F2], $Δ^{''}$ is block diagonal and its domain is the direct sum of the two block domains, so its range is $ran Δ_{0}^{''} \oplus ran Δ_{1}^{''}$ . If a sequence in either block range converges in that $L_{q}^{2}$ summand, embed it in $L_{0}^{2} \oplus L_{1}^{2}$ with zero in the other component. Closedness of the total range from step 1.2 puts the limit in the total range with that other component still zero, hence in the same block range. Both degreewise ranges are closed.

3.1F3F8∎

The two Green identities make $Δ^{''}$ and $G$ inverse bijections between $H^{⊥} \cap dom Δ^{''}$ and $H^{⊥}$ . By [F8], $∥ G f ∥_{H^{2}} \leq C ∥ f ∥_{L^{2}}$ ; together with $∥ Δ^{''} G f ∥_{L^{2}} = ∥ f ∥_{L^{2}}$ this bounds $G$ into the graph norm $∥ \cdot ∥_{H^{2}, Δ}$ . The forward map is continuous in that graph norm by its definition, so this is a topological isomorphism.

Source notes

Demailly's Ch. VI §2 (2.4), (2.6)–(2.8) states the finite-dimensional kernel, closed smooth range and orthogonal decomposition for a smooth elliptic differential operator on a compact manifold; its proof uses the Gårding and Rellich results and cites Hörmander for elliptic PDE facts. Ch. VI §7 (7.1)–(7.2) states the smooth Chern-Dolbeault decomposition and finite-dimensional Dolbeault cohomology, without giving its proof there. The present proof establishes the Hilbert-domain range and inverse claims from the preceding Green-operator lemma and the compact-perturbation suppliers; it does not import Demailly's smooth decomposition as a Hilbert-domain argument.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Hodge decomposition for Dolbeault forms on a compact Riemann surface

Statement

Assume the Axiom of Choice (The Axiom of Choice). Its countable-choice consequences are supplied by AC implies DC implies countable choice for the Green, closed-range, and elliptic-regularity results used below. Let $X$ be a nonempty compact connected Riemann surface, let $E \to X$ be a holomorphic line bundle with Hermitian metric $h$ , and let $g$ be a compatible Riemannian metric. Use the Hilbert spaces, maximal Dolbeault operator $\overset{ˉ}{D}$ , adjoint ${\overset{ˉ}{D}}^{*}$ , and block Laplacian $Δ^{''} = Δ_{0}^{''} \oplus Δ_{1}^{''}$ from The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Put $Ω^{0, q} (E) : = C^{\infty} (X, Λ^{0, q} T^{*} X \otimes E), H^{0, q} (E) : = \ker Δ_{q}^{''} \subseteq L_{q}^{2} (q = 0, 1) .$ The spaces $H^{0, q} (E)$ are finite-dimensional and consist of smooth forms by the preceding items. Let $G : H^{⊥} \to dom Δ^{''} \cap H^{⊥}$ be the Green operator of Dolbeault green operator is compact on the orthogonal complement of the kernel, where $H = H^{0, 0} (E) \oplus H^{0, 1} (E)$ . Using the orthogonal decomposition in The Dolbeault Laplacian has finite-dimensional kernel and closed range, extend it by zero on $H$ : $\tilde{G} (h + v) : = G v (h \in H, v \in H^{⊥}) .$ Then the harmonic projection is $P_{H} : = I - Δ^{''} \tilde{G} : L_{0}^{2} \oplus L_{1}^{2} ⟶ H .$ For each integer $k \geq 0$ , let $H_{q}^{k}$ be the finite-chart Sobolev completion in Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface.

Smooth and $L^{2}$ decompositions. The following are orthogonal direct sums: $Ω^{0, 0} (E) = H^{0, 0} (E) \oplus {\overset{ˉ}{\partial}}_{E}^{*} (Ω^{0, 1} (E)), Ω^{0, 1} (E) = H^{0, 1} (E) \oplus {\overset{ˉ}{\partial}}_{E} (Ω^{0, 0} (E)),$ $L_{0}^{2} = H^{0, 0} (E) \oplus ran {\overset{ˉ}{D}}^{*}, L_{1}^{2} = H^{0, 1} (E) \oplus ran \overset{ˉ}{D} .$ The two Hilbert ranges are closed. On smooth forms $P_{H}$ maps smooth total forms to smooth total forms.
Sobolev topology. For each $q \in {0, 1}$ and $k \geq 0$ , the degree- $q$ harmonic projection extends boundedly to $H_{q}^{k}$ . Hence $H_{q}^{k} = H^{0, q} (E) \oplus \ker ⁣ (P_{H} ∣_{H_{q}^{k}})$ as a topological direct sum. The smooth range summand in part 1 is closed in the $H^{k}$ topology relative to $Ω^{0, q} (E)$ , and its closure in $H_{q}^{k}$ is $\ker (P_{H} ∣_{H_{q}^{k}})$ .
Harmonic representatives. Every $\overset{ˉ}{\partial}$ -cohomology class in degree $(0, 1)$ has exactly one representative in $H^{0, 1} (E)$ . Moreover, $H^{0, 0} (E) = H^{0} (X, E)$ , the space of holomorphic sections, so every smooth $E$ -valued function splits uniquely as a holomorphic section plus a ${\overset{ˉ}{\partial}}_{E}^{*}$ -image.

Facts & Assumptions

Given: the Axiom of Choice, a nonempty compact connected Riemann surface $X$ , a holomorphic line bundle $E$ with the supplied Hermitian metric, and the compatible metric $g$ .

[F1]

The maximal Dolbeault operator is closed and densely defined; its Hilbert adjoint satisfies $⟨ \overset{ˉ}{D} u, w ⟩ = ⟨ u, {\overset{ˉ}{D}}^{*} w ⟩$ on the adjoint domains and $\ker {\overset{ˉ}{D}}^{*} = (ran \overset{ˉ}{D})^{⊥}$ (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F2]

The Laplacian kernel is $\ker Δ^{''} = (\ker \overset{ˉ}{D}) \oplus (\ker {\overset{ˉ}{D}}^{*})$ (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface).

[F3]

Orthogonal complements are defined by vanishing of the Hilbert pairing, and orthogonality is symmetric (Orthogonality and the orthogonal complement).

[F4]

The total Laplacian has finite-dimensional kernel and $L^{2} = H \oplus ran Δ^{''}$ orthogonally. The Green operator satisfies $Δ^{''} G v = v$ on $H^{⊥}$ and maps that complement into $dom Δ^{''}$ (Dolbeault green operator is compact on the orthogonal complement of the kernel, The Dolbeault Laplacian has finite-dimensional kernel and closed range).

[F5]

The orthogonal projection onto a closed Hilbert subspace is the component in its orthogonal decomposition (The Hilbert orthogonal projection onto a closed subspace).

[F6]

Harmonic forms and the harmonic projection are smooth; smooth degree-one elements of $ran \overset{ˉ}{D}$ have smooth preimages; and smooth coefficients and data give smooth interior solutions (Elliptic regularity for Dolbeault harmonic forms, Smooth data give smooth interior solutions).

[F7]

On smooth sections $\overset{ˉ}{D} = {\overset{ˉ}{\partial}}_{E}$ , the smooth Hilbert adjoint agrees with ${\overset{ˉ}{\partial}}_{E}^{*}$ , and ${\overset{ˉ}{\partial}}_{E} s = 0$ exactly when $s$ is holomorphic (Holomorphic line bundles and meromorphic sections on a Riemann surface, The Dolbeault adjoint and Laplacian: local formulas and ellipticity).

[F8]

The degree-one Laplacian has a smooth-coefficient divergence-form expression; on relatively compact chart domains its positive principal coefficient is uniformly elliptic, and the local weak-solution definition is the corresponding compact-test sesquilinear identity (The Dolbeault adjoint and Laplacian: local formulas and ellipticity, Uniformly elliptic divergence-form operators and their sesquilinear forms, Local weak solutions of a divergence-form operator).

[F9]

The graph domain $dom \overset{ˉ}{D} \oplus dom {\overset{ˉ}{D}}^{*}$ is the finite-chart $H^{1}$ space (Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface).

[F10]

For every $k \geq 0$ , the finite-chart $H_{q}^{k}$ spaces are completions of smooth forms, embed continuously into $L_{q}^{2}$ , and contain every smooth form (Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface).

[F11]

The first-variable-linear Hermitian pairing satisfies $∣ ⟨ u, v ⟩_{L^{2}} ∣ \leq ∥ u ∥_{L^{2}} ∥ v ∥_{L^{2}}$ (Cauchy–Schwarz: $∣ ⟨ u, v ⟩ ∣ \leq ∥ u ∥ ∥ v ∥$ , with equality exactly for linearly dependent vectors).

[F12]

Full AC supplies DC for the closed-range result and AC $_{ω}$ for the Green, Sobolev, local weak-solution, and smooth-data regularity interfaces (The Axiom of Choice, AC implies DC implies countable choice).

Proof

Given: the data and choice assumption in the Statement.

1.1F1F2F3F4F12given

If $f$ has degree $q$ and lies in $H^{0, q} (E)^{⊥}$ , then $Δ^{''} G f = f$ and $G f ⊥ H$ . Its other-degree component is therefore harmonic and orthogonal to the harmonic space, so it vanishes; thus $G f$ has degree $q$ . For $q = 0$ the Green identity gives $f = Δ_{0}^{''} G f = {\overset{ˉ}{D}}^{*} \overset{ˉ}{D} G f$ , so $f \in ran {\overset{ˉ}{D}}^{*}$ . Conversely, if $f = {\overset{ˉ}{D}}^{*} v$ and $u \in \ker \overset{ˉ}{D} = H^{0, 0} (E)$ , the adjoint identity gives $⟨ u, f ⟩ = ⟨ \overset{ˉ}{D} u, v ⟩ = 0$ . Thus $ran {\overset{ˉ}{D}}^{*} = H^{0, 0} (E)^{⊥}$ . For $q = 1$ , the Green identity gives $f = Δ_{1}^{''} G f = \overset{ˉ}{D} {\overset{ˉ}{D}}^{*} G f \in ran \overset{ˉ}{D}$ ; and $\ker {\overset{ˉ}{D}}^{*} = (ran \overset{ˉ}{D})^{⊥}$ implies $ran \overset{ˉ}{D} \subseteq H^{0, 1} (E)^{⊥}$ . Hence $ran \overset{ˉ}{D} = H^{0, 1} (E)^{⊥}$ . Both ranges are closed and give the stated degreewise $L^{2}$ splittings.

2.1F1F4F5F6F7F8F9F12step 1.1

Write $f = h + v$ with $h \in H$ and $v \in H^{⊥}$ . The Green identity gives $Δ^{''} \tilde{G} f = v$ , so $P_{H} f = h$ ; thus the displayed formula is the orthogonal projection. By [F6], it preserves smooth forms and the harmonic summands are smooth. If $w \in Ω^{0, 0} (E)$ is orthogonal to $H^{0, 0} (E)$ , step 1.1 gives $w = {\overset{ˉ}{D}}^{*} u$ for some $u \in dom {\overset{ˉ}{D}}^{*}$ . Since $w$ is smooth, it lies in $dom \overset{ˉ}{D}$ ; hence $u \in dom Δ_{1}^{''}$ and $Δ_{1}^{''} u = \overset{ˉ}{D} w$ is smooth. Also $u \in H^{1}$ by [F9]; the smooth-coefficient formula in [F8] is uniformly elliptic on compactly contained chart patches, and testing its distributional equation by integration by parts gives the local weak-solution identity. The smooth-data regularity in [F6] gives a smooth representative on each such patch. These representatives agree on overlaps because they represent the same section almost everywhere and are continuous, so they glue to a smooth section $\tilde{u}$ with $w = {\overset{ˉ}{\partial}}_{E}^{*} \tilde{u}$ by [F7]. Conversely every smooth ${\overset{ˉ}{\partial}}_{E}^{*}$ -image is orthogonal to $\ker \overset{ˉ}{D}$ by the adjoint identity. For degree one, step 1.1 gives every smooth form orthogonal to $H^{0, 1} (E)$ in $ran \overset{ˉ}{D}$ , and [F6] upgrades it to a smooth ${\overset{ˉ}{\partial}}_{E}$ -image; the adjoint identity gives the reverse orthogonality. This proves both smooth decompositions.

3.1F6F10F11step 2.1

Fix $q$ and $k$ , and choose a finite $L^{2}$ -orthonormal basis $e_{1}, \dots, e_{m}$ of $H^{0, q} (E)$ ; if this space is zero, take the empty basis. Each $e_{j}$ is smooth by [F6], so $P_{q} f = \sum_{j = 1}^{m} ⟨ f, e_{j} ⟩_{L^{2}} e_{j}$ and [F10]–[F11] give $∥ P_{q} f ∥_{H^{k}} \leq \sum_{j} ∥ f ∥_{L^{2}} ∥ e_{j} ∥_{L^{2}} ∥ e_{j} ∥_{H^{k}} \leq C_{k, q} ∥ f ∥_{H^{k}}$ . Thus $P_{q}$ extends to a bounded projection on $H_{q}^{k}$ , whose range is $H^{0, q} (E)$ and whose kernel is closed. The smooth complement is exactly the smooth range from step 2.1, so it is closed in the relative $H^{k}$ topology. If $f \in \ker P_{q} \subset H_{q}^{k}$ , approximate it in $H^{k}$ by smooth $f_{n}$ and set $g_{n} = f_{n} - P_{q} f_{n}$ ; boundedness gives $g_{n} \to f$ , and each $g_{n}$ is in that smooth complement. Therefore its $H^{k}$ closure is precisely $\ker P_{q}$ , proving the topological splitting.

4.1F1F2F3F7step 1.1step 2.1∎

On a Riemann surface every smooth $(0, 1)$ -form is $\overset{ˉ}{\partial}$ -closed, so its degree-one Dolbeault class is taken modulo ${\overset{ˉ}{\partial}}_{E} Ω^{0, 0} (E)$ . The degree-one smooth decomposition from step 2.1 gives a harmonic representative for each class. If two harmonic forms represent the same class, their difference lies both in $H^{0, 1} (E)$ and in its orthogonal complement, so its squared norm is zero and the representatives agree. By [F1], [F2], and [F7], $H^{0, 0} (E) = \ker \overset{ˉ}{D} \cap Ω^{0, 0} (E) = H^{0} (X, E)$ ; the degree-zero splitting in step 2.1 is therefore the asserted unique holomorphic-section decomposition.

Source notes

Demailly's Theorem 7.1 states the smooth Dolbeault decomposition for a compact Hermitian manifold and a holomorphic Hermitian bundle, but says only that it follows in a way similar to §3.3; this item supplies the Hilbert-domain range argument, the smooth preimage step, and the $H^{k}$ topology from its local suppliers. Demailly §3.3 (3.15)–(3.17) assumes a flat Hermitian connection and is not proof for the arbitrary holomorphic line bundle here. Theorem 3.31 in Looijenga and Theorems 6.14–6.15 in McMullen are de Rham comparison statements; Looijenga explicitly does not prove Theorem 3.31. Their passages are contextual only.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Dolbeault cohomology of a compact riemann surface is finite dimensional

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let $X$ be a compact Riemann surface and $E$ a holomorphic line bundle with Hermitian metric $h$ and compatible Riemannian metric $g$ as above. The Dolbeault cohomology groups of $E$ are $H^{0, 0} (X, E) : = \frac{\ker ({\overset{ˉ}{\partial}}_{E} : Ω^{0, 0} (E) \to Ω^{0, 1} (E))}{{0}} = H^{0} (X, E), H^{0, 1} (X, E) : = \frac{Ω^{0, 1} (E)}{{\overset{ˉ}{\partial}}_{E} (Ω^{0, 0} (E))},$ where the second definition uses that on a curve every $(0, 1)$ -form is $\overset{ˉ}{\partial}$ -closed because there are no $(0, 2)$ -forms; this is the kernel-modulo-image convention of Dolbeault cohomology of a domain, applied to the globally defined bundle Dolbeault complex. Write $Ω^{0, q} (E) = C^{\infty} (X, Λ^{0, q} T^{*} X \otimes E)$ and $H^{0, q} (E) = \ker Δ_{q}^{''}$ for the Hilbert harmonic kernels, whose elements are smooth by elliptic regularity. Then:

$H^{0, 0} (X, E) = H^{0} (X, E)$ is the finite-dimensional space of holomorphic sections of $E$ , and it equals the harmonic space $H^{0, 0} (E)$ .
$H^{0, 1} (X, E)$ is finite-dimensional, of dimension $h^{0, 1} (X, E) = \dim H^{0, 1} (E)$ , and the harmonic projection induces an isomorphism $H^{0, 1} (X, E) \tilde{\to} H^{0, 1} (E)$ inverse to the inclusion: every class has a unique harmonic representative.
Both dimensions are independent of the Hermitian metric $h$ and of the compatible Riemannian metric $g$ used to define the harmonic spaces, since the quotient and kernel defining $H^{0, i} (X, E)$ involve only ${\overset{ˉ}{\partial}}_{E}$ .

Facts & Assumptions

Given: The compact Riemann surface, holomorphic line bundle, supplied compatible metrics and full Axiom of Choice in the Statement.

[F1]

The globally defined smooth bundle Dolbeault operator has square zero and its degree-zero kernel consists exactly of holomorphic sections; on a curve the degree-two target vanishes (Holomorphic line bundles and meromorphic sections on a Riemann surface).

[F2]

Dolbeault cohomology uses the quotient of the closed forms by the exact forms, with the degree-minus-one space zero (Dolbeault cohomology of a domain). This supplier states the convention on Euclidean domains; the global bundle complex here is supplied by [F1].

[F3]

The smooth decomposition is $Ω^{0, 1} (E) = H^{0, 1} (E) \oplus {\overset{ˉ}{\partial}}_{E} Ω^{0, 0} (E)$ , the harmonic projection is complex-linear, and $H^{0, 0} (E) = H^{0} (X, E)$ (Hodge decomposition for Dolbeault forms on a compact Riemann surface).

[F4]

The total Hilbert harmonic kernel is finite-dimensional and every harmonic form is smooth (The Dolbeault Laplacian has finite-dimensional kernel and closed range, Elliptic regularity for Dolbeault harmonic forms).

[F5]

Full AC is assumed and carried through the Hodge, finite-kernel and elliptic-regularity interfaces; this quotient argument introduces no new choice (The Axiom of Choice).

Proof

technique · direct

1.1F1F2F3F4F5given

The smooth complex supplied by [F1] has zero incoming space in degree zero and zero outgoing space in degree one. Thus [F2]'s kernel-modulo-image construction gives exactly the displayed groups, and the degree-zero group is $H^{0} (X, E)$ . By [F3] it equals $H^{0, 0} (E)$ ; by [F4] this harmonic subspace of the finite-dimensional total kernel is finite-dimensional.

2.1F3F4step 1.1algebra

Let $P_{1}$ be the degree-one harmonic projection. By [F3], every smooth $u$ has a unique splitting $u = h + {\overset{ˉ}{\partial}}_{E} f$ with $h \in H^{0, 1} (E)$ , and $P_{1} u = h$ . Hence $P_{1}$ vanishes on exact forms, so $[u] \mapsto P_{1} u$ is a well-defined complex-linear map from $H^{0, 1} (X, E)$ . It is surjective because each harmonic $h$ is smooth by [F4] and satisfies $P_{1} h = h$ . Its kernel is zero because $P_{1} u = 0$ forces $u = {\overset{ˉ}{\partial}}_{E} f$ . Inclusion of harmonic forms followed by passage to the quotient is its inverse. Thus the degree-one group is isomorphic to the finite-dimensional space $H^{0, 1} (E)$ , proving the dimension formula and unique harmonic representation.

3.1F1F5step 1.1step 2.1∎

The operator ${\overset{ˉ}{\partial}}_{E}$ and the smooth form spaces in [F1] are determined by the holomorphic structure, independently of $h, g$ . Their fixed kernel and quotient therefore define the same two cohomology vector spaces for every choice of these metrics. Applying steps 1.1–2.1 to each choice identifies its harmonic spaces with these fixed finite-dimensional spaces, so both dimensions agree. Full AC is inherited exactly through [F5].

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-08Open item page →

Harmonic star duality for line bundle valued dolbeault cohomology

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let $X$ be a compact Riemann surface, $E$ a holomorphic line bundle with Hermitian metric $h$ , $g$ a compatible Riemannian metric, and let $E^{*}$ be the dual line bundle with the dual Hermitian metric $h^{*}$ and the induced holomorphic structure ${\overset{ˉ}{\partial}}_{E^{*}}$ (Dual and Hom vector bundles, Smooth bundle metrics). Write $⟨ \cdot, \cdot ⟩_{L^{2}}$ for the $L^{2}$ pairing, which is $C$ -linear in the first variable (The complex $L^{2}$ pairing on equivalence classes).

The Hodge-# operator. Write $#$ for the conjugate-linear metric Hodge star $⋆_{E}$ of Hermitian metric and $L^{2}$ pairing on a compact Riemann surface on $(0, 1)$ -forms and for $⋆_{E^{*}}$ on $(1, 0)$ -forms with values in $E^{*}$ (using the canonical identification $E^{* *} = E$ ), so that $# : Ω^{0, 1} (E) \to Ω^{1, 0} (E^{*})$ is defined by the requirement $s \land # t = ⟨ s, t ⟩ d V_{g},$ the wedge pairing the $E$ - and $E^{*}$ -factors by the canonical duality (this is the $#$ -operator of the Hodge-star calculus, Riemannian hodge star, Hodge star is a smooth bundle isomorphism). Then $#$ is a conjugate-linear bundle isomorphism, $#^{2} = - 1$ on $(0, 1)$ -forms, and $s \land # s = ∣ s ∣^{2} d V_{g}$ ; consequently $\int_{X} s \land # s = ∥ s ∥_{L^{2}}^{2}$ for every smooth $s$ , and this top form is nonzero at every point where $s$ is nonzero; the inverse of the degree-one map is $- ⋆_{E^{*}}$ .
Commutation with the Laplacian, and the holomorphic image. Identify the $E^{*}$ -valued $(1, 0)$ -forms with the smooth sections of the holomorphic line bundle $F : = K \otimes E^{*}$ , carrying the induced tensor Hermitian metric (the holomorphic frame $d z \otimes e^{*}$ has squared norm $2 / (ρ ψ)$ when $g = ρ (d x^{2} + d y^{2})$ and $h (e, e) = ψ$ ), and let $Δ_{F}^{''}$ be the Dolbeault Laplacian of $F$ as in The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface. Then $# Δ_{E}^{''} = Δ_{F}^{''} #$ on smooth $(0, 1)$ -forms, so $#$ maps the harmonic space $H^{0, 1} (E) = \ker Δ_{E}^{''}$ conjugately and isomorphically onto the harmonic space of $F$ , which is exactly the space $H^{0} (X, K \otimes E^{*}) = \ker ({\overset{ˉ}{\partial}}_{F} : C^{\infty} (X, F) \to C^{\infty} (X, Λ^{0, 1} T^{*} X \otimes F))$ of holomorphic $E^{*}$ -valued $(1, 0)$ -forms (Meromorphic differentials, orders and residues).
Duality. The $C$ -bilinear pairing $B : H^{0, 1} (X, E) \times H^{0} (X, K \otimes E^{*}) ⟶ C, B (u, α) = \int_{X} u \land α,$ is well defined on Dolbeault classes (the integrand is a $2$ -form on the closed oriented surface, and Stokes' theorem shows that replacing $u$ by $u + {\overset{ˉ}{\partial}}_{E} f$ changes the integral by zero, The general Stokes theorem, A compactly supported primitive has zero total derivative integral), and it is a nondegenerate duality: the induced map $u \mapsto B (u, \cdot)$ is a $C$ -linear isomorphism $H^{0, 1} (X, E) \tilde{\to} H^{0} (X, K \otimes E^{*})^{*},$ so the dual of the Dolbeault group is $H^{0, 1} (X, E)^{*} ≅ H^{0} (X, K \otimes E^{*})$ .
Conjugations. Written on harmonic representatives, the two pairings are related by $⟨ u, v ⟩_{L^{2}} = B (u, # v) (u, v \in H^{0, 1} (E)),$ the $L^{2}$ pairing being $C$ -linear in the first variable and conjugate-linear in the second; correspondingly the Riesz map $H^{0, 1} (E) \to H^{0, 1} (E)^{*}$ , $u \mapsto ⟨ \cdot, u ⟩$ , is conjugate-linear, while the identification $u \mapsto B (u, \cdot)$ is $C$ -linear, and pulling $B (u, \cdot)$ back along the conjugate-linear map $# : H^{0, 1} (E) \to H^{0} (X, K \otimes E^{*})$ gives the functional $v \mapsto B (u, # v) = ⟨ u, v ⟩_{L^{2}}$ , which is conjugate-linear in $v$ and depends complex-linearly on $u$ . This pullback is distinct from the ordinary complex-linear dual functional $v \mapsto ⟨ v, u ⟩_{L^{2}}$ defining the Riesz map.

Facts & Assumptions

Given: The compact Riemann surface, holomorphic line bundle, supplied compatible metrics and full Axiom of Choice in the Statement.

[F1]

The canonical bundle, holomorphic dual bundle and their coefficientwise Dolbeault operators are well defined, and the kernel on smooth sections is the space of holomorphic sections (Holomorphic line bundles and meromorphic sections on a Riemann surface).

[F2]

The bundle star is conjugate-linear; its local formulas are $⋆_{E} (u d \overset{ˉ}{z} \otimes e) = - i ψ \overset{ˉ}{u} d z \otimes e^{*}$ and $⋆_{E^{*}} (a d z \otimes e^{*}) = i ψ^{- 1} \overset{ˉ}{a} d \overset{ˉ}{z} \otimes e^{* *}$ . Their composition is minus the identity, and $s \land ⋆_{E} t = ⟨ s, t ⟩ d V_{g}$ . The induced covector norm satisfies $∣ d z ∣_{g}^{2} = 2 / ρ$ (Hermitian metric and $L^{2}$ pairing on a compact Riemann surface).

[F3]

For any supplied Hermitian holomorphic line bundle of local weight $w$ , the smooth formulas are ${\overset{ˉ}{\partial}}^{*} (u d \overset{ˉ}{z}) = - 2 (ρ w)^{- 1} \partial_{z} (w u)$ , $Δ_{0}^{''} a = - 2 (ρ w)^{- 1} \partial_{z} (w \partial_{\overset{ˉ}{z}} a)$ and $Δ_{1}^{''} u = - 2 \partial_{\overset{ˉ}{z}} ((ρ w)^{- 1} \partial_{z} (w u))$ (The Dolbeault adjoint and Laplacian: local formulas and ellipticity).

[F4]

The Hilbert harmonic kernels consist of smooth forms; in degree one their equation is ${\overset{ˉ}{D}}^{*} u = 0$ , and in degree zero it is $\overset{ˉ}{D} a = 0$ . The maximal and adjoint operators agree with their smooth expressions on these smooth forms (The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface, Elliptic regularity for Dolbeault harmonic forms).

[F5]

The degree-one Dolbeault group is finite-dimensional and has a unique smooth harmonic representative; the degree-zero harmonic kernel for every holomorphic line bundle equals its holomorphic-section space (Dolbeault cohomology of a compact riemann surface is finite dimensional, Hodge decomposition for Dolbeault forms on a compact Riemann surface).

[F6]

An exact smooth top form whose primitive has compact support on a boundaryless oriented manifold has integral zero, under Countable Choice (A compactly supported primitive has zero total derivative integral).

[F7]

A complex Hilbert inner product is linear in its first variable. Under Countable Choice every bounded complex-linear functional has a unique Riesz representation $v \mapsto ⟨ v, u ⟩$ (The complex $L^{2}$ pairing on equivalence classes, Riesz representation for Hilbert spaces).

[F8]

Full AC is assumed; its countable instances supply Stokes and Riesz, and it is inherited through all harmonic regularity and Hodge interfaces. No additional choice is used in the local star or finite-dimensional argument (The Axiom of Choice).

Proof

technique · direct

1.1F2givenalgebra

By [F2], $#$ in the forward direction sends $u d \overset{ˉ}{z} \otimes e$ to $- i ψ \overset{ˉ}{u} d z \otimes e^{*}$ , is conjugate-linear, and is a smooth bundle isomorphism. In the reverse direction it sends $a d z \otimes e^{*}$ to $i ψ^{- 1} \overset{ˉ}{a} d \overset{ˉ}{z} \otimes e^{* *}$ ; substituting $a = - i ψ \overset{ˉ}{u}$ gives $- u$ , so $#^{2} = - 1$ and the inverse is minus the reverse star. The wedge identity gives $s \land # s = ∣ s ∣^{2} d V_{g}$ . A nonzero smooth $s$ has positive squared norm on a neighborhood of a point, hence its integral is $∥ s ∥_{L^{2}}^{2} > 0$ ; the pointwise top form is nonzero exactly where $s$ is nonzero.

1.2F1F6F8givenalgebra

For a smooth section $f$ of $E$ and a holomorphic section $α$ of $F$ , evaluation contracts $f α$ to a global smooth $(1, 0)$ -form. Its exterior derivative has only a $(1, 1)$ component on a curve, and locally $α = a d z \otimes e^{*}$ with $\partial_{\overset{ˉ}{z}} a = 0$ , so $d (f α) = {\overset{ˉ}{\partial}}_{E} f \land α$ . On compact $X$ the primitive $f α$ has compact support, and [F6] applied to real and imaginary parts gives $\int_{X} {\overset{ˉ}{\partial}}_{E} f \land α = 0$ . Consequently the integral defines $B ([u], α)$ independently of the smooth representative, and coefficientwise wedge and evaluation make it complex-bilinear.

2.1F1F2F3step 1.1algebra

In the holomorphic frame $d z \otimes e^{*}$ of $F = K \otimes E^{*}$ , the tensor metric has weight $w = 2 / (ρ ψ)$ by [F2]. The inverse and canonical cocycles are holomorphic, so this is a holomorphic line bundle by [F1]. For a smooth local coefficient $u$ of an $E$ -valued $(0, 1)$ -form put $a = - i ψ \overset{ˉ}{u}$ . Applying [F3] to $F$ gives $Δ_{F, 0}^{''} a = - ψ \partial_{z} (2 (ρ ψ)^{- 1} \partial_{\overset{ˉ}{z}} (- i ψ \overset{ˉ}{u})) = 2 i ψ \partial_{z} ((ρ ψ)^{- 1} \partial_{\overset{ˉ}{z}} (ψ \overset{ˉ}{u}))$ . Applying [F3] to $E$ and conjugating gives $- i ψ \overline{Δ_{E, 1}^{''} u} = 2 i ψ \partial_{z} ((ρ ψ)^{- 1} \partial_{\overset{ˉ}{z}} (ψ \overset{ˉ}{u}))$ , since $ρ, ψ$ are real. These equal coefficients prove $Δ_{F, 0}^{''} # = # Δ_{E, 1}^{''}$ on smooth forms; the local equalities are global because the operators and star are globally defined.

3.1F1F3F4F5step 1.1step 2.1

By [F4], a Hilbert harmonic degree-one form is smooth and harmonic exactly when $\partial_{z} (ψ u) = 0$ . Conjugating this equation gives $\partial_{\overset{ˉ}{z}} (ψ \overset{ˉ}{u}) = 0$ , exactly holomorphy of $a = - i ψ \overset{ˉ}{u}$ in the holomorphic frame of $F$ . Conversely, if $a$ is a holomorphic section coefficient, the inverse star gives $u = - i ψ^{- 1} \overset{ˉ}{a}$ , which is smooth, satisfies that first-order equation and therefore belongs to the Hilbert harmonic kernel by [F4]. Thus $#$ is a conjugate-linear bijection between $H^{0, 1} (E)$ and $H^{0} (X, F)$ ; the latter is the degree-zero harmonic kernel for $F$ by [F5]. This uses smooth representatives of maximal-domain kernels, not an identification of a maximal domain with smooth forms.

4.1F5F8step 1.1step 3.1step 1.2algebra

Replace each Dolbeault class by its unique harmonic representative $u$ using [F5]. If $u \neq 0$ , then $# u \in H^{0} (X, F)$ by step 3.1 and $B ([u], # u) = ∥ u ∥_{L^{2}}^{2} > 0$ by step 1.1. Conversely every nonzero $α \in H^{0} (X, F)$ equals $# v$ for a nonzero harmonic $v$ , so $B ([v], α) > 0$ . This proves nondegeneracy in both variables. To prove the asserted isomorphisms explicitly, choose an $L^{2}$ -orthonormal basis $e_{1}, \dots, e_{m}$ of the finite-dimensional harmonic space; if it is zero take the empty basis. The $# e_{j}$ form a complex basis of $H^{0} (X, F)$ by the conjugate-linear bijection, and $B ([e_{j}], # e_{k}) = δ_{j k}$ . Thus $[e_{j}] \mapsto B ([e_{j}], \cdot)$ maps a basis to its dual basis and is a complex-linear isomorphism; likewise $# e_{k} \mapsto B (\cdot, # e_{k})$ gives the complex-linear inverse-side duality $H^{0} (X, F) ≅ H^{0, 1} (X, E)^{*}$ .

5.1F2F7F8step 3.1step 1.2step 4.1algebra∎

Integrating [F2] gives $B ([u], # v) = ⟨ u, v ⟩_{L^{2}}$ for harmonic $u, v$ . Since $B$ is complex-bilinear and $#$ conjugate-linear, this expression is linear in $u$ and conjugate-linear in $v$ ; pulling back a linear functional on $H^{0} (X, F)$ along $#$ therefore gives a conjugate-linear functional on the harmonic space. Separately, $R (u) (v) = ⟨ v, u ⟩$ is complex-linear in $v$ and satisfies $R (c u) = \overset{ˉ}{c} R (u)$ . The harmonic space is finite-dimensional and hence Hilbert in its inherited pairing, so [F7] identifies this conjugate-linear map with the Riesz bijection to its ordinary complex-linear dual. This proves the stated conjugation conventions and completes all claims, with full AC inherited as in [F8].

5 · Examples, counterexamples and false statements

None yet.

Hodge Theory on Compact Riemann Surfaces

1 · Prerequisites

2 · Summary

3 · Logical flowchart

4 · Definitions, theorems and proofs

Holomorphic line bundles and meromorphic sections on a Riemann surface

Definition

Facts & Assumptions

Proof

Hermitian metric and L2L2 pairing on a compact Riemann surface

Definition

Facts & Assumptions

Proof

The maximal Dolbeault operator and its Hilbert adjoint on a compact Riemann surface

Statement

Facts & Assumptions

Proof

Chern connection of a Hermitian holomorphic line bundle

Statement

Facts & Assumptions

Proof

The Dolbeault adjoint and Laplacian: local formulas and ellipticity

Statement

Facts & Assumptions

Proof

Gårding estimates for the Dolbeault Laplacian on a compact Riemann surface

Statement

Facts & Assumptions

Proof

Source notes

Dolbeault green operator is compact on the orthogonal complement of the kernel

Statement

Facts & Assumptions

Proof

Source notes

Elliptic regularity for Dolbeault harmonic forms

Statement

Facts & Assumptions

Proof

Source notes

The Dolbeault Laplacian has finite-dimensional kernel and closed range

Statement

Facts & Assumptions

Proof

Source notes

Hodge decomposition for Dolbeault forms on a compact Riemann surface

Statement

Facts & Assumptions

Proof

Source notes

Dolbeault cohomology of a compact riemann surface is finite dimensional

Statement

Facts & Assumptions

Proof

Harmonic star duality for line bundle valued dolbeault cohomology

Statement

Facts & Assumptions

Proof

5 · Examples, counterexamples and false statements

Sources

Hermitian metric and $L^{2}$ pairing on a compact Riemann surface