Survey of sheaf cohomology on Riemann surfaces

Intuition [Beginner]

You have already met sheaf cohomology four times under four different names on a Riemann surface. The Čech construction takes an open cover and records what fails to glue from local data. The Dolbeault construction takes the Cauchy-Riemann operator (call it $D$) and records its kernel modulo image. The harmonic construction takes the Hodge-Laplace and records its kernel as a single representative per cohomology class. The derived-functor construction takes the global-sections functor and records its right-derived approximations. Four pictures, one answer.

The reason they agree is not a coincidence — it is the analytic-and-topological richness of a compact Riemann surface itself. The four pictures live on four different sides of the bridge: combinatorial cocycles on covers (Čech), partial differential equations on smooth forms (Dolbeault), spectral data of an elliptic operator (harmonic), and abstract category-theoretic resolutions (derived). The bridges between them are theorems with names: Cartan-Leray, the Dolbeault isomorphism, the Hodge theorem, and the derived-equals-Čech comparison.

The pay-off of having four pictures is that any computation you cannot do in one picture you can sometimes do in another. Want to count global sections of a line bundle? Use Čech on a small cover (06.04.02). Want to know that the count is finite? Use ellipticity in the harmonic picture (06.04.05). Want to relate it to the dual line bundle? Use Serre duality (06.04.04). Want to compare to algebraic geometry on a projective curve? Use Grothendieck plus GAGA. The survey is the road map for moving between pictures.

Visual [Beginner]

A schematic of a compact Riemann surface $X$ at the centre, with four arrows pointing outward to four boxes labelled "Čech cocycles", "Dolbeault forms", "Harmonic forms", and "Derived functors". Each box contains a small icon: a covered surface for Čech, a $D$ operator for Dolbeault, a Laplacian symbol for harmonic, a chain of arrows for derived functors. Two-headed arrows between every pair of boxes label the comparison theorems Cartan-Leray, Dolbeault, Hodge, and derived-equals-Čech.

Worked example [Beginner]

Take the Riemann sphere $X={\mathbb{P}}^1$ and the structure sheaf ${\mathcal{O}}_X$. Compute $H^1(X,{\mathcal{O}}_X)$ in all four pictures and check the answers match.

In the Čech picture on the standard two-set cover with two affine patches and intersection an annulus, a one-cocycle is a Laurent polynomial in the overlap coordinate $z$. The coboundary of a pair $(f(z),g(w))$ on the two patches is $g(z^{-1})-f(z)$, and every Laurent polynomial decomposes uniquely as a polynomial in $z$ minus a polynomial in $w=1/z$, so the coboundary is surjective. The first Čech cohomology ${\check{H}}^1$ is therefore $0$.

In the Dolbeault picture, $H^{0,1}({\mathbb{P}}^1,\mathcal{O})=0$ because the Cauchy-Pompeiu formula on the sphere lets every smooth $(0,1)$-form be written as the Cauchy-Riemann operator applied to a smooth function (the global Cauchy-Riemann solvability holds on ${\mathbb{P}}^1$ for the structure sheaf). So $H^{0,1}=0$.

In the harmonic picture, the harmonic-form space ${\mathcal{H}}^{0,1}({\mathbb{P}}^1,\mathcal{O})=0$ because the genus of ${\mathbb{P}}^1$ is $0$, and the harmonic anti-holomorphic 1-forms on a compact Riemann surface form a $g$-dimensional space.

In the derived-functor picture, $R^1\Gamma ({\mathbb{P}}^1,\mathcal{O})$ is computed by any acyclic resolution; the Dolbeault resolution by smooth $(0,q)$-forms gives the Dolbeault answer, which is $0$.

What this tells us: all four answers agree at $0$, as expected from the comparison theorems. The same checking procedure for the canonical bundle ${\omega }_{{\mathbb{P}}^1}=\mathcal{O}(-2)$ would give $H^1=1$ in all four pictures (the Serre-dual side of $H^0({\mathcal{O}}_X)=\mathbb{C}$). The four pictures are bookkeeping conventions for one underlying invariant.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a compact Riemann surface and $\mathcal{F}$ a sheaf of ${\mathcal{O}}_X$-modules on $X$. The four pictures of sheaf cohomology on $X$ are the following.

(1) Čech. For an open cover $\mathcal{U}=\{U_i{\}}_{i\in I}$ indexed by a totally ordered set, the Čech cochain complex is

$$C^p(\mathcal{U},\mathcal{F})=\prod _{i_0<\cdots <i_p}\Gamma (U_{i_0\cdots i_p},\mathcal{F}),$$

with the alternating-sum-of-restriction differential. The Čech cohomology of $\mathcal{F}$ on $X$ is the colimit over refinements,

$${\check{H}}^p(X,\mathcal{F})={\mathrm{colim}}_{\mathcal{U}}{H}^p(C^{\bullet }(\mathcal{U},\mathcal{F})).$$

The line-bundle case $\mathcal{F}=\mathcal{O}(L)$ is 06.04.02.

(2) Dolbeault / harmonic. When $\mathcal{F}=\mathcal{O}(L)$ for a holomorphic line bundle $L$, the smooth $(0,q)$-form sheaves ${\mathcal{A}}^{0,q}(L)=C^{\infty }(L\otimes {\overline{T^{\ast }X}}^{\otimes q})$ form a fine resolution

$$0\to \mathcal{O}(L)\to {\mathcal{A}}^{0,0}(L)\stackrel{\bar{\partial }}{\to }{\mathcal{A}}^{0,1}(L)\to 0,$$

and the Dolbeault cohomology is the cohomology of the global-section complex

$$H_{\bar{\partial }}^{0,q}(X,L)=\frac{\ker (\bar{\partial }:{\Omega }^{0,q}(X,L)\to {\Omega }^{0,q+1}(X,L))}{\mathrm{im}(\bar{\partial }:{\Omega }^{0,q-1}(X,L)\to {\Omega }^{0,q}(X,L))}.$$

With a Hermitian metric on $X$ and $L$, the Hodge-Laplace ${\Delta }_{\bar{\partial }}$ has finite-dimensional kernel ${\mathcal{H}}^q(X,L)$ canonically isomorphic to $H_{\bar{\partial }}^{0,q}(X,L)$. The framework is 06.04.05.

(3) Derived-functor. Sheaves of ${\mathcal{O}}_X$-modules form an abelian category with enough injectives. The right-derived functors of the global-sections functor $\Gamma (X,-)$ define

$$H^q(X,\mathcal{F})=R^q\Gamma (X,\mathcal{F})$$

(Grothendieck 1957 Tôhoku). A short exact sequence of sheaves induces a long exact sequence in $H^{\ast }$.

(4) Singular / topological for the constant sheaf $\mathbb{Z}$, $\mathbb{R}$, or $\mathbb{C}$. Singular cochains on the underlying topological surface compute $H^{\ast }(X,\mathbb{Z})$; for $\mathbb{C}$-coefficients, the de Rham complex of smooth real-valued differential forms computes $H_{dR}^{\ast }(X,\mathbb{C})\cong H^{\ast }(X,\mathbb{C})$ via integration of forms over chains.

Theorem (the four pictures agree). On a compact Riemann surface $X$ and a coherent analytic sheaf $\mathcal{F}$, the natural maps

$${\check{H}}^q(X,\mathcal{F})\cong R^q\Gamma (X,\mathcal{F})\cong H_{\bar{\partial }}^{0,q}(X,\mathcal{F})\cong {\mathcal{H}}^q(X,\mathcal{F})$$

are isomorphisms in every degree. For the constant sheaf $\mathbb{C}$, all of these further match $H^{dR}(X, \mathbb{C}) \cong H^*(X, \mathbb{C})$viadeRham,andtheHodgedecompositionidentifies$H^k(X, \mathbb{C}) = \bigoplus{p + q = k} H^{p, q}(X)$.*

The four-way identification is the core of the survey: Cartan-Leray for Čech-vs-derived, Dolbeault for derived-vs-Dolbeault, Hodge for Dolbeault-vs-harmonic, de Rham for $\mathbb{C}$-coefficients, and Hodge symmetry for the bidegree decomposition.

Counterexamples to common slips

• The four-picture equivalence requires paracompactness (for Čech-vs-derived) and coherence (for Dolbeault-vs-derived). On a non-paracompact space Čech can disagree with derived; on a non-coherent sheaf the Dolbeault resolution is not finite.
• The harmonic picture requires compactness and a Kähler metric for the Hodge-Laplace to have finite-dimensional kernel and the Hodge decomposition to hold. A Riemann surface is automatically Kähler (every Hermitian metric is closed in real dimension two), but in higher dimension the Kähler hypothesis is a substantive geometric restriction.
• The Hodge decomposition $H^k=⨁H^{p,q}$ assumes Hodge symmetry $h^{p,q}=h^{q,p}$, which fails on non-Kähler compact complex manifolds (the Hopf surface ${\mathbb{C}}^2\setminus \{0\}/\mathbb{Z}$ has $h^{0,1}=1$ but $h^{1,0}=0$).
• Serre duality requires the sheaf to be coherent; for non-coherent sheaves, the dual side is not finite-dimensional and the perfect-pairing statement breaks down.
• GAGA requires projective in the algebraic-side hypothesis, not merely proper. On a complete-but-not-projective scheme the analytic-vs-algebraic comparison can fail.

Key theorem with proof [Intermediate+]

Theorem (four-picture comparison on a compact Riemann surface). Let $X$ be a compact Riemann surface and $L\to X$ a holomorphic line bundle. The canonical maps

$${\check{H}}^q(X,L)\to R^q\Gamma (X,\mathcal{O}(L))\to H_{\bar{\partial }}^{0,q}(X,L)\to {\mathcal{H}}^q(X,L)$$

are isomorphisms of finite-dimensional complex vector spaces for every $q\ge 0$. For $q\ge 2$ all four spaces vanish.

Proof. The argument has four steps, each invoking a comparison theorem proved elsewhere.

Step 1 — Čech equals derived. For a sheaf $\mathcal{F}$ on a paracompact Hausdorff space, the canonical map ${\check{H}}^q(X,\mathcal{F})\to R^q\Gamma (X,\mathcal{F})$ is an isomorphism in every degree (Cartan-Leray). A compact Riemann surface is paracompact and Hausdorff, so the comparison applies to $\mathcal{F}=\mathcal{O}(L)$ and gives the first isomorphism.

Step 2 — derived equals Dolbeault. The Dolbeault resolution

$$0\to \mathcal{O}(L)\to {\mathcal{A}}^{0,0}(L)\stackrel{\bar{\partial }}{\to }{\mathcal{A}}^{0,1}(L)\to 0$$

is a resolution by fine sheaves: the ${\mathcal{A}}^{0,q}(L)$ admit smooth partitions of unity, hence are acyclic for the global-sections functor, and the resolution is exact at the sheaf level by the local $\bar{\partial }$-Poincaré lemma (Dolbeault-Grothendieck). Acyclic resolutions compute right-derived functors, so $R^q\Gamma (X,\mathcal{O}(L))=H^q(\Gamma (X,{\mathcal{A}}^{0,\bullet }(L)))=H_{\bar{\partial }}^{0,q}(X,L)$. The second isomorphism is the Dolbeault identification.

Step 3 — Dolbeault equals harmonic. Equip $X$ with a Hermitian metric $\omega$ and $L$ with a Hermitian metric $h$. The Hodge-Laplace ${\Delta }_{\bar{\partial }}=\bar{\partial }{\bar{\partial }}^{\ast }+{\bar{\partial }}^{\ast }\bar{\partial }$ on $L$-valued $(0,q)$-forms is a self-adjoint elliptic operator of order two with compact resolvent on the compact $X$. The Hodge decomposition (proved in 06.04.05)

$$L_{(0,q)}^2(X,L)={\mathcal{H}}^q(X,L)\oplus \overline{\mathrm{im}\bar{\partial }}\oplus \overline{\mathrm{im}{\bar{\partial }}^{\ast }}$$

and the Dolbeault isomorphism identify ${\mathcal{H}}^q(X,L)$ with $H_{\bar{\partial }}^{0,q}(X,L)$. The third isomorphism is the harmonic-projection theorem.

Step 4 — vanishing and finiteness. On a real-two-dimensional manifold, ${\mathcal{A}}^{0,q}(L)=0$ for $q\ge 2$, so $H_{\bar{\partial }}^{0,q}(X,L)=0$ for $q\ge 2$ and the four-picture equivalence forces vanishing of all four spaces in those degrees. Finiteness of ${\mathcal{H}}^q(X,L)$ for $q=0,1$ follows from ellipticity of ${\Delta }_{\bar{\partial }}$ on a compact $X$. Transferring along the chain of isomorphisms, ${\check{H}}^q$, $R^q\Gamma$, and $H_{\bar{\partial }}^{0,q}$ are all finite-dimensional. $\square$

The four-step structure mirrors the proofs in 06.04.02 (steps 1, 2, 4) and 06.04.05 (step 3); the survey collates the inputs into a single comparison statement.

Bridge. The four-picture comparison is the analytic-side foundation for Serre duality on a curve 06.04.04: the residue-trace pairing ${\check{H}}^1(X,K_X)\cong \mathbb{C}$ is computed in the Čech picture, while the perfect-pairing structure $H^q(X,L)\cong H^{n-q}(X,K_X\otimes L^{-1}{)}^{\ast }$ uses the harmonic-projection theorem from the Dolbeault side. Combined with 06.04.01 (Riemann-Roch), the Euler-characteristic identity $\chi (L)=\mathrm{deg}L+1-g$ is read in any picture: as an alternating sum of Čech cohomology dimensions, as the index of the elliptic complex $\bar{\partial }:{\mathcal{A}}^{0,0}(L)\to {\mathcal{A}}^{0,1}(L)$, or as the alternating dimension of harmonic kernels. Putting these together, the foundational insight is that on a compact Riemann surface every cohomological invariant of a coherent sheaf has four concrete computations available — combinatorial cocycles, $\bar{\partial }$-cohomology, harmonic forms, and derived functors — and the survey is the dictionary for choosing the right one in any application. The dimension-one case generalises to higher-dimensional Cartan-Serre, Kodaira-Hodge, and Grothendieck cohomology, with Stein and compact-Kähler cases as the two extremal regimes that govern complex analysis and complex algebraic geometry respectively.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib does not currently formalise the four-picture comparison theorem on a compact Riemann surface as a single coherent statement. A proposed signature, in Lean 4 / Mathlib syntax, sketching the target:

[object Promise]

The proof depends on names that do not currently exist in Mathlib (the coherent analytic sheaf category on a Riemann surface, the Čech cochain functor, the Dolbeault and harmonic functors, and the comparison-isomorphism web between them). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

The four-picture comparison on a compact Riemann surface is the dimension-one specialisation of a theorem in every complex dimension. The general formulation, due to Cartan-Serre on Stein and compact-Kähler manifolds in the 1950s and refined by Grothendieck on schemes in 1957, is the analytic-and-algebraic backbone of complex geometry.

Higher-dimensional generalisations. On a Stein manifold $M$ of complex dimension $n$, Cartan-Serre Theorems A and B (Cartan 1951–53 séminaire; Cartan-Serre 1953 CRAS 237) give $H^q(M,\mathcal{F})=0$ for every coherent analytic $\mathcal{F}$ and every $q\ge 1$. The four pictures all agree and all vanish in positive degree. On a compact Kähler manifold $(M,\omega )$, the Hodge decomposition $H^k(M,\mathbb{C})={⨁}_{p+q=k}{H}^{p,q}(M)$ holds with $h^{p,q}(M)=h^{q,p}(M)=h^{n-p,n-q}(M)$ (Hodge symmetry, Serre duality), and Kodaira's vanishing theorem $H^q(M,K_M\otimes L)=0$ for $q>0$ and $L$ positive specialises to the Riemann-surface case as Serre vanishing $H^1(X,L)=0$ for $\mathrm{deg}L\ge 2g-1$. On a noetherian scheme $X$ over a field $k$, Grothendieck's framework (1957 Tôhoku) replaces analytic sheaves by quasi-coherent or coherent sheaves of ${\mathcal{O}}_X$-modules; the derived-functor picture and the Čech picture (on a separated $X$, Cartan's comparison theorem from Hartshorne III §4) agree, and Serre duality holds on a smooth proper $X$ via Grothendieck-Hartshorne residues and duality.

GAGA. Serre 1956 Géométrie algébrique et géométrie analytique (Ann. Inst. Fourier 6) proved that on a smooth projective variety $X$ over $\mathbb{C}$, the analytification functor $\mathrm{Coh}(X)\to \mathrm{Coh}(X^{\mathrm{an}})$ is an equivalence of categories, and the cohomology comparison $H^q(X,\mathcal{F})\to H^q(X^{\mathrm{an}},{\mathcal{F}}^{\mathrm{an}})$ is an isomorphism in every degree. On a smooth projective curve $X$ over $\mathbb{C}$, GAGA matches Hartshorne's algebraic Čech computations on the standard affine cover with Forster's analytic Čech computations on a Stein cover, identifying the algebraic and analytic four-picture surveys as one combined story. The proof uses Serre's projective-vanishing theorem (Serre 1955 FAC III.4): for $L$ ample on a projective $X$, $H^q(X,\mathcal{F}\otimes L^{\otimes m})=0$ for $m\gg 0$ and $q\ge 1$; the same vanishing holds analytically by Kodaira, and the comparison is built up by induction on dimension using long exact sequences.

Hodge bridge. On a compact Riemann surface, the four pictures of sheaf cohomology of ${\mathcal{O}}_X$, ${\omega }_X$, the constant sheaf $\mathbb{C}$, and the Dolbeault and de Rham complexes all interlock via the Hodge decomposition. Specifically:

• $H^1(X,{\mathcal{O}}_X)\cong H^{0,1}(X)$ (Dolbeault for the structure sheaf).
• $H^0(X,{\omega }_X)\cong H^{1,0}(X)$ (holomorphic 1-forms).
• $H^1(X,\mathbb{C})\cong H^{1,0}(X)\oplus H^{0,1}(X)\cong {\mathbb{C}}^{2g}$ (Hodge decomposition, de Rham comparison).
• $H^{0,1}(X)=\overline{H^{1,0}(X)}$ (Hodge symmetry; conjugate-linear identification).

Combining these, $\dim H^1(X,{\mathcal{O}}_X)=\dim H^0(X,{\omega }_X)=g$, the genus identification. Serre duality $H^1(X,L)\cong H^0(X,{\omega }_X\otimes L^{-1}{)}^{\ast }$ is the global instance of the Hodge perfect-pairing structure on de Rham cohomology specialised to $L={\mathcal{O}}_X$ and twisted to arbitrary $L$.

Cohomology table (compact Riemann surface of genus $g$).

sheaf $\mathcal{F}$	$h^0(\mathcal{F})$	$h^1(\mathcal{F})$	regime
${\mathcal{O}}_X$	$1$	$g$	structure sheaf
${\omega }_X$	$g$	$1$	canonical bundle
$\mathcal{O}(D)$, $\mathrm{deg}D>2g-2$	$\mathrm{deg}D-g+1$	$0$	non-special (Serre vanishing)
$\mathcal{O}(D)$, $\mathrm{deg}D<0$	$0$	$h^0({\omega }_X\otimes \mathcal{O}(-D))$	dual non-special
$\mathcal{O}(D)$, $0\le \mathrm{deg}D\le 2g-2$	RR + Serre + speciality	RR + Serre + speciality	special

The first three rows are immediate from Riemann-Roch and Serre duality; the fourth row uses Serre duality to convert $H^1$ on a negative-degree bundle into $H^0$ on a positive-degree one; the fifth row is the speciality range, where Brill-Noether theory and the Clifford bound ($h^0(L)\le \mathrm{deg}L/2+1$ for special $L$) refine the dimension counts further. Each row is a worked computation in any of the four pictures.

Failure modes (where the analogy breaks).

• Non-Kähler compact complex manifolds (Hopf surface, Iwasawa manifold, Calabi-Eckmann manifolds): Hodge symmetry $h^{p,q}=h^{q,p}$ fails, the harmonic-vs-de Rham comparison loses the bidegree decomposition, and the Hodge spectral sequence does not degenerate at $E_1$. Dolbeault and harmonic still agree, but the bridge to topological cohomology is incomplete.
• Non-paracompact spaces: Čech-vs-derived comparison can fail. Riemann surfaces are paracompact, so this failure mode does not appear in the survey; it appears for example on the long line or on non-Hausdorff schemes.
• Non-coherent sheaves: $H^q(X,\mathcal{F})$ may be infinite-dimensional and the Dolbeault / harmonic pictures do not directly apply. Coherence is the indispensable hypothesis for the finite-dimensional bookkeeping the survey relies on.
• Higher-dimensional non-Kähler: GAGA, Kodaira vanishing, and Hodge decomposition all require the projective-Kähler hypothesis. Moishezon manifolds and class C manifolds (Fujiki) are weaker generalisations where some of the survey's identifications survive but not all.
• Positive characteristic: Hodge decomposition can fail on smooth proper varieties over $\overline{{\mathbb{F}}_p}$ (counterexamples by Mumford, Serre), although Deligne-Illusie 1987 reproved Hodge degeneration via positive-characteristic crystalline methods under a lifting hypothesis. The four-picture survey on a Riemann surface lives over $\mathbb{C}$; the higher-dimensional generalisation requires care over fields of positive characteristic.

Synthesis. The survey collects four equivalent computations of sheaf cohomology on a compact Riemann surface — Čech cocycles, Dolbeault forms, harmonic representatives, derived functors — and the comparison-theorem web (Cartan-Leray, Dolbeault, Hodge, derived-equals-Čech) that identifies them. Each picture is the right tool for some computational task: Čech for explicit cocycles on standard covers, Dolbeault for partial-differential-equation methods, harmonic for spectral and curvature arguments, derived functors for category-theoretic packaging. The dimension-one case extends to higher dimension as Cartan-Serre on Stein manifolds, Kodaira-Hodge on compact Kähler manifolds, and Grothendieck on noetherian schemes, with GAGA the bridge between the analytic and algebraic worlds on a smooth projective complex variety. Read in the opposite direction, the failure modes — non-Kähler compact complex manifolds, non-paracompact spaces, non-coherent sheaves, positive characteristic — show that the four-picture mantra is not formal: it requires the analytic-and-topological richness of a compact Kähler manifold (or a paracompact analytic space with coherent coefficients) as the load-bearing input. On a compact Riemann surface every hypothesis is automatic, and the four pictures collapse to a single finite-dimensional invariant per coherent sheaf and degree — the engine of the entire theory of compact-curve geometry, from Riemann-Roch and Serre duality through the Picard group and the Jacobian to the moduli of curves and the Hodge bundle.

Full proof set [Master]

The survey is synoptic: full proofs of the named comparison theorems live in the units they generalise, and are summarised here with citations. The integrity check below verifies that every claim in Advanced results is either proved here, proved in a referenced unit, or stated without proof with a primary citation.

Theorem (Cartan-Leray comparison). On a paracompact Hausdorff space $X$ and a sheaf $\mathcal{F}$ of abelian groups on $X$, the canonical map ${\check{H}}^q(X,\mathcal{F})\to R^q\Gamma (X,\mathcal{F})$ is an isomorphism in every degree.

Proof. Stated without proof here; full proof in Godement Topologie algébrique et théorie des faisceaux (1958) Ch. II §5, and reproduced in Hartshorne Algebraic Geometry III §4 (Theorem III.4.5) for the Noetherian-separated-scheme case. The argument: choose a Godement-resolution by canonical flabby sheaves; flabby sheaves are acyclic for both Čech and derived-functor cohomology; the resolution computes both, hence they agree. Paracompactness is used to make Čech of a flabby sheaf acyclic on a cover. $\square$

Theorem (Dolbeault isomorphism on a compact Riemann surface). For a holomorphic line bundle $L$ on a compact Riemann surface $X$, the canonical map $H^q(X,\mathcal{O}(L))\to H_{\bar{\partial }}^{0,q}(X,L)$ is an isomorphism in every degree.

Proof. Full proof in 06.04.02. The Dolbeault resolution by smooth $(0,q)$-form sheaves is fine (smooth partitions of unity), hence acyclic, and exact at the sheaf level by the local $\bar{\partial }$-Poincaré lemma (Dolbeault-Grothendieck). Acyclic resolutions compute right-derived functors. $\square$

Theorem (Hodge / harmonic-projection on a compact Riemann surface). For a Hermitian holomorphic line bundle $(L,h)$ on a compact Riemann surface with Kähler metric $\omega$, $L^2_{(0, q)}(X, L) = \mathcal{H}^q(X, L) \oplus \overline{\mathrm{im}, \bar\partial} \oplus \overline{\mathrm{im}, \bar\partial^}$orthogonally,and$\mathcal{H}^q(X, L) \cong H^{0, q}_{\bar\partial}(X, L)$.*

Proof. Full proof in 06.04.05. Ellipticity and self-adjointness of ${\Delta }_{\bar{\partial }}$ on a compact $X$ give a compact resolvent and a discrete spectrum; the spectral theorem gives the orthogonal decomposition; elliptic regularity gives smoothness of $L^2$-harmonic forms. $\square$

Theorem (Hodge decomposition on a compact Kähler manifold). For a compact Kähler manifold $(M,\omega )$ of complex dimension $n$ and constant sheaf $\mathbb{C}$, $H^k(M,\mathbb{C})={⨁}_{p+q=k}{H}^{p,q}(M)$ with Hodge symmetry $h^{p,q}(M)=h^{q,p}(M)$ and Serre duality $h^{p,q}(M)=h^{n-p,n-q}(M)$.

Proof. Stated without proof here; full proof in 06.04.03 for the Riemann-surface case, and in Voisin Hodge Theory and Complex Algebraic Geometry I §5 (Theorem 5.4) for the general compact Kähler case. The argument: the Kähler identities $[\Lambda ,\bar{\partial }]=-i{\partial }^{\ast }$ and $[L_{\omega },\partial ]=-i\bar{\partial }$ on a Kähler manifold give ${\Delta }_d=2{\Delta }_{\bar{\partial }}=2{\Delta }_{\partial }$, so harmonic forms decompose by bidegree $(p,q)$. Hodge symmetry follows from the conjugate-linear involution $\overline{\cdot }:{\mathcal{H}}^{p,q}\to {\mathcal{H}}^{q,p}$. Serre duality follows from the Hodge-star isomorphism. $\square$

Theorem (Serre duality on a curve). For a coherent sheaf $\mathcal{F}$ on a smooth projective curve $X$ over $\mathbb{C}$, the Yoneda pairing $H^1(X,\mathcal{F})\times H^0(X,{\mathcal{F}}^{\vee }\otimes {\omega }_X)\to H^1(X,{\omega }_X)\cong \mathbb{C}$ is a perfect pairing of finite-dimensional vector spaces.

Proof. Full proof in 06.04.04. Trace-and-cup-product construction; reduction to $\mathcal{F}={\mathcal{O}}_X$ via twisting; case $\mathcal{F}={\mathcal{O}}_X$ via the Hodge decomposition. $\square$

Theorem (Schwartz finiteness on a compact Riemann surface). For a coherent analytic sheaf $\mathcal{F}$ on a compact Riemann surface $X$, $\dim H^q(X,\mathcal{F})<\infty$ for every $q\ge 0$.

Proof. Sketched in Exercise 7 above. Cartan-Serre lemma reduces to vector bundles; rank-induction reduces to line bundles; the harmonic picture identifies $H^q(X,L)$ with $\ker {\Delta }_{\bar{\partial }}^{(0,q)}$ on $L$-valued forms, and ellipticity gives finite-dimensionality. Full historical proof in Forster Lectures on Riemann Surfaces §14, using a compact-perturbation Banach-space argument that bypasses the Hodge machinery. $\square$

Theorem (GAGA on a smooth projective curve over $\mathbb{C}$). For a smooth projective curve $X$ over $\mathbb{C}$ and a coherent algebraic sheaf $\mathcal{F}$ on $X$, the natural map $H^q(X,\mathcal{F})\to H^q(X^{\mathrm{an}},{\mathcal{F}}^{\mathrm{an}})$ is an isomorphism for every $q\ge 0$, and the analytification functor $\mathrm{Coh}(X)\to \mathrm{Coh}(X^{\mathrm{an}})$ is an equivalence of categories.

Proof. Stated without proof here; full proof in Serre 1956 Géométrie algébrique et géométrie analytique (Ann. Inst. Fourier 6, 1-42) Theorem 2 + Theorem 3, and reproduced in Hartshorne Algebraic Geometry Appendix B. The argument: ample line bundles on $X^{\mathrm{an}}$ are obtained from algebraic ample line bundles via analytification; Serre's projective vanishing theorem (algebraic) and Kodaira vanishing (analytic) give the same vanishing $H^q(X,\mathcal{F}\otimes L^{\otimes m})=0$ for $m\gg 0$ and $q\ge 1$ on both sides; the analytification functor is fully faithful by the projective-embedding argument and essentially surjective by the analytic Chow theorem (every closed analytic subspace of ${\mathbb{P}}^N$ is algebraic, Chow 1949). $\square$

Theorem (long exact sequence in cohomology). A short exact sequence of sheaves $0\to {\mathcal{F}}^{\prime }\to \mathcal{F}\to {\mathcal{F}}^{\prime \prime }\to 0$ on a topological space $X$ induces a long exact sequence $\cdots \to H^q(X,{\mathcal{F}}^{\prime })\to H^q(X,\mathcal{F})\to H^q(X,{\mathcal{F}}^{\prime \prime })\to H^{q+1}(X,{\mathcal{F}}^{\prime })\to \cdots$.

Proof. Standard; the universal property of right-derived functors. Stated without proof here; full proof in Grothendieck 1957 Tôhoku §II.2 or Hartshorne III §1 (Theorem III.1.1A). The long exact sequence is the entry point for the inductive divisor-bumping arguments throughout the survey. $\square$

Connections [Master]

• Čech cohomology of holomorphic line bundles 06.04.02. The Čech picture of the survey, specialised to the line-bundle sheaf $\mathcal{O}(L)$ on a compact Riemann surface. The four-picture comparison reduces to the line-bundle Dolbeault isomorphism proven there, with the Cartan-Leray identification of Čech with derived functors as the bridge.

• Hodge decomposition on a compact Riemann surface 06.04.03. The Hodge picture of the survey, specialised to the constant sheaf $\mathbb{C}$ on a compact Riemann surface. The bidegree decomposition $H^1(X,\mathbb{C})=H^{1,0}(X)\oplus H^{0,1}(X)$ is the source of Hodge symmetry $h^{0,1}=h^{1,0}=g$ that drives the genus identifications.

• Serre duality on a curve 06.04.04. The duality theorem identifying $H^q$ on a sheaf with $H^{n-q}$ on its canonical-twisted dual. The four-picture survey is the foundational input: Serre duality lives in the cohomology computed by any of the four pictures, and the perfect-pairing structure uses the harmonic representatives from the Hodge picture.

• Hilbert-space PDE for $\bar{\partial }$ 06.04.05. The harmonic / Hodge picture of the survey, in its full Hilbert-space form. The harmonic-projection theorem is the analytic input that promotes finite-dimensionality from Schwartz's compact-perturbation Banach-space argument to the explicit elliptic-PDE picture.

• Riemann-Roch theorem on compact Riemann surfaces 06.04.01. The dimension count $\chi (L)=\mathrm{deg}L+1-g$ is read in any of the four pictures: as alternating Čech dimensions, as the Dolbeault index, as the alternating dimension of harmonic kernels, or as the alternating dimension of derived-functor groups. Riemann-Roch is the headline theorem the survey supports.

• Holomorphic line bundle on a Riemann surface 06.05.02. The objects of the survey when $\mathcal{F}=\mathcal{O}(L)$. The transition cocycle $g_{ij}$ defining $L$ is itself a Čech one-cocycle in ${\mathcal{O}}_X^{\times }$, and the Čech picture of the survey records the additive cohomology of $\mathcal{O}(L)$.

• Sheaf cohomology in algebraic geometry 04.03.01. The algebraic-side counterpart of the survey, on a noetherian scheme rather than a complex manifold. Grothendieck's derived-functor framework (1957 Tôhoku) is shared by both sides; GAGA provides the bridge for smooth projective complex curves.

• Čech cohomology on schemes 04.03.03. The algebraic-side Čech picture, with Cartan's comparison theorem (Hartshorne III.4) replacing Cartan-Leray for Noetherian separated schemes. On a smooth projective curve over $\mathbb{C}$, GAGA identifies algebraic and analytic Čech.

• Cohomology of projective space ${\mathcal{O}}_{{\mathbb{P}}^n}(d)$ 04.03.04. The standard worked example of sheaf cohomology in algebraic geometry: $H^q({\mathbb{P}}^n,\mathcal{O}(d))$ via Čech on the standard affine cover, recovering the dimensions $\left(\genfrac{}{}{0ex}{}{n+d}{n}\right)$ for $H^0$ and $\left(\genfrac{}{}{0ex}{}{-d-1}{n}\right)$ for $H^n$. The curve case $n=1$ recovers $\dim H^0({\mathbb{P}}^1,\mathcal{O}(d))=d+1$ for $d\ge 0$, the worked example of 06.04.02.

• Stein Riemann surfaces 06.09.01. The non-compact version of the survey: every non-compact Riemann surface is Stein (Behnke-Stein 1949), so $H^q(X,\mathcal{F})=0$ for $q\ge 1$ and every coherent $\mathcal{F}$ (Cartan-Serre Theorem B). The four-picture comparison still holds, but all four spaces vanish in positive degree.

• Picard group of a Riemann surface. The classification of holomorphic line bundles via $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\times })$, the Čech cohomology of the multiplicative unit sheaf. The exponential sheaf sequence $0\to \mathbb{Z}\to {\mathcal{O}}_X\to {\mathcal{O}}_X^{\times }\to 0$ uses the four-picture identifications $H^1(X,{\mathcal{O}}_X)={\mathbb{C}}^g$ and $H^1(X,\mathbb{Z})={\mathbb{Z}}^{2g}$ to give ${\mathrm{Pic}}^0(X)\cong {\mathbb{C}}^g/{\mathbb{Z}}^{2g}$ — the Jacobian.

Historical & philosophical context [Master]

Eduard Čech introduced his combinatorial cohomology in Théorie générale de l'homologie dans un espace quelconque (Fundamenta Mathematicae 19 (1932) 149-183) ^{[Čech 1932]}, originally as a homology theory based on the nerve of an open cover. The cohomological dressing with the alternating-sum-of-restriction differential — and the application to sheaf-valued cochains — is the work of Henri Cartan in his Paris séminaire (1948-54) and the Cartan-Eilenberg book Homological Algebra (1956), with the holomorphic-bundle case worked out by Behnke-Stein 1949 and Cartan-Serre 1953.

W.V.D. Hodge's The Theory and Applications of Harmonic Integrals (Cambridge University Press, 1941) ^{[Hodge 1941]} established the harmonic-representative theorem: on a compact Riemannian manifold, every de Rham cohomology class has a unique harmonic representative minimising the $L^2$-norm. The Dolbeault refinement on compact complex manifolds was developed by Pierre Dolbeault in his 1953 thesis ^{[Dolbeault 1953]} (C. R. Acad. Sci. Paris 236, 175-177), with the Bochner-Kodaira-Nakano formula due jointly to Kodaira and Nakano in the early 1950s.

Jean-Pierre Serre's Faisceaux Algébriques Cohérents (Annals of Mathematics 61 (1955) 197-278) ^{[Serre 1955]} founded sheaf-cohomological algebraic geometry. The paper proved affine vanishing, computed $H^{\ast }({\mathbb{P}}^n,\mathcal{O}(d))$ via Čech on the standard cover, and established Serre duality. Serre's companion paper Un théorème de dualité (Comm. Math. Helv. 29 (1955) 9-26) proved Serre duality on coherent sheaves, with the curve case as the dimension-one specialisation. The 1956 paper Géométrie algébrique et géométrie analytique (Ann. Inst. Fourier 6, 1-42) ^{[Serre 1956]} proved GAGA: on a smooth projective complex variety, the analytification functor on coherent sheaves is an equivalence of categories and preserves cohomology. Together the three Serre papers from 1955-56 established the algebraic-vs-analytic dictionary that the survey records as a single comparison.

Alexandre Grothendieck's Sur quelques points d'algèbre homologique (Tôhoku Math. J. 9 (1957) 119-221) ^{[Grothendieck 1957]} reformulated sheaf cohomology as the right-derived functor of global sections in an arbitrary abelian category with enough injectives. The reformulation absorbed Cartan's séminaire, Cartan-Eilenberg's homological algebra, and Serre's FAC into one categorical framework, and provided the long-exact-sequence machinery, the spectral sequences, and the duality formulations that organise the survey.

Lars Hörmander's L^2 estimates and existence theorems for the $\bar{\partial }$ operator (Acta Math. 113 (1965) 89-152) ^{[Hörmander 1965]} established the analytic-side existence theorem for $\bar{\partial }$ on Stein manifolds with explicit weighted-norm estimates. Together with Andreotti-Vesentini's compact-Kähler treatment the same year, the four-picture comparison gained its full quantitative analytic input, and the survey's harmonic / Hodge picture became computable in the Hilbert-space framework.

Otto Forster's Lectures on Riemann Surfaces (Springer GTM 81, 1981) ^[Forster] §11-§17 codifies the analytic-Čech sheaf-cohomological treatment as the standard textbook account; Simon Donaldson's Riemann Surfaces (Oxford GTM 22, 2011) ^[Donaldson] §10-§12 presents the harmonic-projection / PDE picture, integrating the four-picture comparison into the proof of Riemann-Roch. Phillip Griffiths and Joseph Harris's Principles of Algebraic Geometry (Wiley, 1978) ^{[Griffiths-Harris]} §0.4 gives the higher-dimensional generalisation; Claire Voisin's Hodge Theory and Complex Algebraic Geometry I (Cambridge Studies 76, 2002) ^[Voisin] §1-§5 gives the modern Hodge-theoretic treatment with the spectral-sequence and mixed-Hodge-structure refinements.

Bibliography [Master]

[object Promise]