Čech cohomology of holomorphic line bundles

Intuition [Beginner]

Take a compact Riemann surface $X$ — a closed complex curve, like the Riemann sphere or a torus — and a holomorphic line bundle $L$ on it. A line bundle is a way of attaching a one-dimensional complex vector space to each point of $X$ so that nearby fibres are smoothly identified, with the gluing data given on overlaps of an open cover. Two questions naturally arise: how many global holomorphic sections does $L$ have, and what global obstruction prevents local sections of $L$ from gluing into one?

Čech cohomology answers both. Cover $X$ by open pieces, record sections on each piece (a zero-cochain), record local sections on overlaps (a one-cochain), and ask which overlap data refuse to come from data on the pieces. The result in degree zero is $H^0(X,L)$ — the global holomorphic sections of $L$. The result in degree one is $H^1(X,L)$ — the obstruction space, the room for local sections that cannot be globally smoothed away.

A compact Riemann surface has real dimension two, so all higher cohomology vanishes: $H^p(X,L)=0$ for $p\ge 2$. The entire cohomological story of every line bundle on $X$ lives in two finite-dimensional vector spaces. The Čech construction is the bookkeeping that turns the cover-and-transition-functions data of $L$ into the dimensions of those two spaces.

Visual [Beginner]

A schematic of a compact Riemann surface $X$ covered by three open patches $U_0,U_1,U_2$, with pairwise overlaps $U_0\cap U_1,U_1\cap U_2,U_0\cap U_2$ and a triple overlap $U_0\cap U_1\cap U_2$. A small ribbon over each patch labels a local section of the line bundle $L$; arrows on each overlap label transition functions $g_{ij}$ rescaling the ribbon between charts.

Worked example [Beginner]

Take the Riemann sphere $X={\mathbb{P}}^1$ and the line bundle $L=\mathcal{O}(2)$. Cover ${\mathbb{P}}^1$ by two pieces $U_0={\mathbb{P}}^1\setminus \{\infty \}$ and $U_1={\mathbb{P}}^1\setminus \{0\}$. Each piece is a copy of the complex line $\mathbb{C}$. The overlap $U_0\cap U_1={\mathbb{C}}^{\ast }$ is the punctured complex line, with coordinate $z$ on $U_0$ and $w=1/z$ on $U_1$.

The transition function of $\mathcal{O}(2)$ on the overlap is $z\mapsto z^2$. A holomorphic section of $\mathcal{O}(2)$ is a pair $(f_0,f_1)$ with $f_0$ holomorphic on $U_0$, $f_1$ holomorphic on $U_1$, and $f_1=z^2{f}_0$ on the overlap. Writing $f_0=a_0+a_{1}z+a_2{z}^2+a_3{z}^3+\cdots$ as a power series in $z$ and $f_1=b_0+b_{1}w+b_2{w}^2+\cdots$ as a power series in $w$, the relation $f_1=z^2{f}_0$ on the overlap reads

$$b_0+b_1{z}^{-1}+b_2{z}^{-2}+\cdots =a_0{z}^2+a_1{z}^3+a_2{z}^4+a_3{z}^5+\cdots .$$

Comparing coefficients: only $a_0,a_1,a_2$ appear on the right with non-negative powers of $z^{-1}$ matching on the left, and the $b_j$ for $j\ge 1$ would require negative powers of $z$ which are absent on the right. The solutions are exactly $f_0=a_0+a_{1}z+a_2{z}^2$ — degree-$\le 2$ polynomials — with $f_1=a_2+a_{1}w+a_0{w}^2$ on the other chart.

The space of global sections has a basis $\{1,z,z^2\}$ on the $U_0$-chart, so its dimension is three. The Čech first cohomology vanishes by direct check (or by Serre duality: $H^1(\mathcal{O}(2))\cong H^0(\mathcal{O}(-4){)}^{\ast }=0$).

What this tells us: the cover plus transition function $z\mapsto z^2$ produces a finite-dimensional space of global sections, equal to the degree-plus-one count $\dim H^0(\mathcal{O}(d))=d+1$ for $d\ge 0$. The Čech construction turned the gluing data into a polynomial-counting problem.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a compact Riemann surface, $L\to X$ a holomorphic line bundle with transition functions $g_{ij}$ on a cover $\mathcal{U}=\{U_i{\}}_{i\in I}$ indexed by a totally ordered set $I$, and $\mathcal{O}(L)$ the sheaf of holomorphic sections of $L$. For an ordered tuple $i_0<i_1<\cdots <i_p$ write $U_{i_0\cdots i_p}=U_{i_0}\cap \cdots \cap U_{i_p}$.

The Čech cochain group in degree $p$ is

$$C^p(\mathcal{U},\mathcal{O}(L))=\prod _{i_0<i_1<\cdots <i_p}\Gamma (U_{i_0\cdots i_p},L),$$

a product of holomorphic sections of $L$ over the $(p+1)$-fold intersections. The Čech differential ${\delta }^p:C^p\to C^{p+1}$ is the alternating-sum-of-restriction map

$$({\delta }^p\alpha {)}_{i_0\cdots i_{p+1}}=\sum _{k=0}^{p+1}(-1{)}^k{\alpha }_{i_0\cdots \widehat{i_k}\cdots i_{p+1}}{|}_{U_{i_0\cdots i_{p+1}}},$$

where the hat denotes omission and the restriction respects the line-bundle gluing via the transition functions $g_{ij}$. Direct computation on faces gives ${\delta }^{p+1}\circ {\delta }^p=0$, so $C^{\bullet }(\mathcal{U},\mathcal{O}(L))$ is a cochain complex of complex vector spaces. The Čech cohomology of $L$ on the cover $\mathcal{U}$ is

$${\check{H}}^p(\mathcal{U},L)=\frac{\ker ({\delta }^p)}{\mathrm{im}({\delta }^{p-1})}.$$

A refinement of $\mathcal{U}$ is a cover $\mathcal{V}=\{V_j\}$ with a refinement map $\tau :J\to I$ such that $V_j\subseteq U_{\tau (j)}$. Refinements induce chain maps whose effect on cohomology depends only on $\mathcal{V}$ refining $\mathcal{U}$, and the Čech cohomology of $L$ on $X$ is the colimit

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

In degree zero, ${\check{H}}^0(\mathcal{U},L)=H^0(X,L)=\Gamma (X,L)$ for every cover, since a zero-cocycle is a compatible family of local sections that glues to a global section of $L$ by the sheaf axiom.

Theorem (Čech cohomology on a compact Riemann surface). For a holomorphic line bundle $L$ on a compact Riemann surface $X$:

• ${\check{H}}^p(X,L)=0$ for every $p\ge 2$ (dimensional vanishing).
• ${\check{H}}^p(X,L)$ is finite-dimensional over $\mathbb{C}$ for $p=0,1$ (Schwartz-Cartan-Serre finiteness).
• The canonical comparison map to Dolbeault cohomology ${\check{H}}^p(X,L)\to H_{\bar{\partial }}^{0,p}(X,L)$ is an isomorphism in every degree.

Equivalent forms.

• Sheaf-cohomology identification: ${\check{H}}^p(X,L)=H^p(X,\mathcal{O}(L))$, the derived-functor cohomology of the sheaf of holomorphic sections; the comparison is an isomorphism on a paracompact Hausdorff space, and a Riemann surface is paracompact.
• Dolbeault identification: the fine resolution $0\to \mathcal{O}(L)\to C^{\infty }(L)\stackrel{\bar{\partial }}{\to }{C}^{\infty }(L\otimes \overline{T^{\ast }X})\to 0$ on a Riemann surface gives ${\check{H}}^p(X,L)\cong H_{\bar{\partial }}^{0,p}(X,L)$ via the standard fine-sheaf-acyclicity argument.
• Serre duality form: combined with 06.04.04, ${\check{H}}^1(X,L)\cong H^0(X,K_X\otimes L^{-1}{)}^{\ast }$ where $K_X$ is the canonical line bundle.

Counterexamples to common slips

• The cocycle condition $\delta \alpha =0$ on a one-cochain $\alpha =({\alpha }_{ij})$ on triple overlaps reads ${\alpha }_{jk}-{\alpha }_{ik}+{\alpha }_{ij}=0$, with the orientation convention determined by the alternating sign $(-1{)}^k$ in the differential. Reversing the convention reverses signs.
• The transition functions $g_{ij}$ of $L$ themselves form a one-cocycle in ${\mathcal{O}}_X^{\times }$ (multiplicatively, on the unit sheaf), not in $\mathcal{O}(L)$ (additively, on the section sheaf). The Picard group $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\times })$ classifies line bundles; the cohomology $H^1(X,\mathcal{O}(L))$ instead measures the failure of $L$ to be a complete linear system.
• The dimensional vanishing $H^p=0$ for $p\ge 2$ uses real dimension two of the underlying surface. On a higher-dimensional complex manifold the same Čech construction generates non-zero higher cohomology, and the dimensional ceiling rises with the complex dimension.
• Without compactness, $H^0(X,L)$ may be infinite-dimensional (consider $L={\mathcal{O}}_X$ on $X=\mathbb{C}$: holomorphic functions on $\mathbb{C}$ form an infinite-dimensional space). The finiteness of $H^p(X,L)$ for $p=0,1$ on a compact Riemann surface is a substantive input, due to Schwartz's compact-perturbation theorem on Banach spaces (Forster §14).

Key theorem with proof [Intermediate+]

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

$${\check{H}}^p(X,L)⟶H_{\bar{\partial }}^{0,p}(X,L)$$

is an isomorphism for every $p\ge 0$. In particular ${\check{H}}^p(X,L)=0$ for $p\ge 2$, and ${\check{H}}^0,{\check{H}}^1$ are finite-dimensional complex vector spaces.

Proof. The argument has four steps: identify Čech cohomology with derived-functor sheaf cohomology, set up the Dolbeault fine resolution of $\mathcal{O}(L)$, compute the derived-functor cohomology from the resolution, and read off the dimensional vanishing and finiteness.

Step 1 — Čech equals derived-functor cohomology. For a sheaf $\mathcal{F}$ on a paracompact Hausdorff space $X$, the canonical map ${\check{H}}^p(X,\mathcal{F})\to H^p(X,\mathcal{F})$ to derived-functor sheaf cohomology is an isomorphism in every degree (Cartan-Leray). A compact Riemann surface is paracompact and Hausdorff, so this comparison applies to $\mathcal{F}=\mathcal{O}(L)$. The identification is realised by the colimit over refinements, with any sufficiently fine cover already computing the answer.

Step 2 — Dolbeault fine resolution. Let ${\mathcal{A}}^{0,0}(L)=C^{\infty }(L)$ denote the sheaf of $C^{\infty }$ sections of $L$, and ${\mathcal{A}}^{0,1}(L)=C^{\infty }(L\otimes \overline{T^{\ast }X})$ the sheaf of $C^{\infty }$ $L$-valued $(0,1)$-forms. The $\bar{\partial }$-operator on $L$, defined locally on a trivialising chart by $\bar{\partial }(\sigma \cdot e_{\alpha })=(\bar{\partial }\sigma )\cdot e_{\alpha }$ for a smooth section $\sigma$ and a holomorphic frame $e_{\alpha }$, is well-defined: the transition between trivialisations is by holomorphic transition functions $g_{ij}$, and $\bar{\partial }$ commutes with multiplication by holomorphic functions. The kernel of $\bar{\partial }:{\mathcal{A}}^{0,0}(L)\to {\mathcal{A}}^{0,1}(L)$ is exactly $\mathcal{O}(L)$, the sheaf of holomorphic sections. By the $\bar{\partial }$-Poincaré lemma (Dolbeault-Grothendieck), every $\bar{\partial }$-closed $(0,1)$-form with values in $L$ is locally $\bar{\partial }$-exact on a small disc, so the sequence

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

is exact as a sequence of sheaves. The sheaves ${\mathcal{A}}^{0,q}(L)$ are fine — they admit smooth partitions of unity — and fine sheaves are acyclic for the global-sections functor, so this is a fine resolution of $\mathcal{O}(L)$.

Step 3 — derived-functor cohomology from the resolution. Acyclic resolutions compute derived-functor cohomology: applying global sections to the resolution and taking cohomology yields

$$H^p(X,\mathcal{O}(L))\cong H^p(\Gamma (X,{\mathcal{A}}^{0,\bullet }(L)),\bar{\partial }).$$

The right-hand side is the Dolbeault cohomology of $L$, denoted $H_{\bar{\partial }}^{0,p}(X,L)$. Composing with the Čech-versus-derived-functor comparison from Step 1 gives the canonical Dolbeault identification

$${\check{H}}^p(X,L)\cong H_{\bar{\partial }}^{0,p}(X,L).$$

Step 4 — vanishing and finiteness. The resolution has length two: ${\mathcal{A}}^{0,q}(L)=0$ for $q\ge 2$ on a real two-dimensional manifold, since a $(0,q)$-form requires $q$ independent anti-holomorphic differentials and on a curve there is only one. Hence $H_{\bar{\partial }}^{0,p}(X,L)=0$ for $p\ge 2$, and ${\check{H}}^p(X,L)=0$ for $p\ge 2$ by the comparison.

For finiteness: the Hodge theorem on a compact Hermitian manifold (a Riemann surface is Kähler for any Hermitian metric, since the Kähler form is top-dimensional and automatically closed) identifies $H_{\bar{\partial }}^{0,p}(X,L)$ with the space of $\bar{\partial }$-harmonic $(0,p)$-forms with values in $L$. The $\bar{\partial }$-Laplacian ${\Delta }_{\bar{\partial }}$ on $L$ is a self-adjoint elliptic operator of order two on the compact Riemann surface; elliptic regularity gives a finite-dimensional kernel. Therefore $H_{\bar{\partial }}^{0,p}(X,L)$, and hence ${\check{H}}^p(X,L)$, is finite-dimensional for $p=0,1$. $\square$

The four-step structure follows Donaldson §10-§11 (fine-resolution and harmonic-projection route) and dovetails with Forster §12-§17 (Čech-from-the-bottom-up route via Schwartz's compact-perturbation finiteness theorem). The two routes prove the same theorem; Forster's elementary Banach-space argument avoids the Hodge machinery, while Donaldson's harmonic-projection argument couples cleanly to the Hodge decomposition of 06.04.03.

Bridge. The construction here is the input to 06.04.04 (Serre duality on a curve): the Čech cohomology ${\check{H}}^1(X,L)$ pairs against the global sections of the canonical-twisted dual, with the residue trace map $H^1(X,K_X)\to \mathbb{C}$ supplying the perfect-pairing structure. Combined with 06.04.01 (Riemann-Roch), the dimension count $\dim H^0(L)-\dim H^1(L)=\mathrm{deg}L+1-g$ becomes a sharp two-sided identity once $\dim H^1(L)$ is read as $\dim H^0(K_X\otimes L^{-1})$ via the Serre pairing on the Čech side. The same Čech construction governs the Mittag-Leffler problem (Forster §26): an obstruction class in ${\check{H}}^1(X,L)$ encodes whether prescribed local principal-parts data on $X$ can be realised by a global meromorphic section. The Cousin-I and Cousin-II problems reformulate as the vanishing of specific Čech-$H^1$ obstructions, and the same vanishing controls the Picard group $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\times })$ that classifies line bundles up to isomorphism. Putting these together, the foundational insight is that on a compact Riemann surface every cohomological obstruction to assembling local holomorphic data into a global object lives in a single finite-dimensional Čech-$H^1$ space; the four-step proof identifies that space with Dolbeault cohomology and harmonic representatives, opening the door to Serre duality, Riemann-Roch, and the entire dimension-counting apparatus of compact-curve geometry.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib does not currently formalise Čech cohomology of a holomorphic line bundle on a compact Riemann surface as a first-class object. A proposed signature, in Lean 4 / Mathlib syntax, sketching the target statement:

[object Promise]

The proof depends on names that do not currently exist in Mathlib (the holomorphic-line-bundle category on a Riemann surface, the Čech-cochain functor, the $\bar{\partial }$-operator on $L$-valued forms, the Dolbeault fine resolution, and Schwartz's compact-perturbation finiteness). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

Čech cohomology of a holomorphic line bundle on a compact Riemann surface is the dimension-one analytic case of a theorem in every complex dimension. The general formulation, due to Henri Cartan and Jean-Pierre Serre in the 1950s and refined by the Behnke-Grauert school (Behnke-Stein 1949, Grauert 1958), states that for a holomorphic vector bundle $E$ on a complex manifold $M$ and the sheaf $\mathcal{O}(E)$ of holomorphic sections, the Čech cohomology ${\check{H}}^p(M,\mathcal{O}(E))$ agrees with derived-functor sheaf cohomology and with Dolbeault cohomology $H_{\bar{\partial }}^{0,p}(M,E)$. On a compact $M$, all ${\check{H}}^p$ are finite-dimensional; on a Stein $M$ (holomorphically convex non-compact), Theorem B (Cartan-Serre) gives ${\check{H}}^p(M,\mathcal{O}(E))=0$ for every $p\ge 1$. The dimension-one curve case below is the bridge between these two regimes.

Schwartz finiteness via compact perturbation. Forster §14 proves the finiteness ${\dim }_{\mathbb{C}}{\check{H}}^1(X,\mathcal{O}(L))<\infty$ on a compact Riemann surface by an elementary Banach-space argument that avoids the Hodge machinery. Choose a finite cover $\mathcal{U}=\{U_i\}$ of $X$ by relatively compact discs $U_i⋐V_i$ with refinement $\mathcal{V}=\{V_i\}$; the restriction map $C^1(\mathcal{V},\mathcal{O}(L))\to C^1(\mathcal{U},\mathcal{O}(L))$ is a compact operator between Banach spaces (Bergman-space restriction is compact for nested precompact discs). Schwartz's theorem on compact perturbations of surjective Fredholm maps then implies ${\check{H}}^1(\mathcal{V},\mathcal{O}(L))$ is finite-dimensional, and the colimit is bounded. The argument is independent of harmonic theory, but reaches the same dimension count.

Dolbeault and harmonic representatives. The Hodge theorem on a compact Hermitian manifold (a Riemann surface is automatically Kähler) identifies $H_{\bar{\partial }}^{0,p}(X,L)$ with the kernel of the $\bar{\partial }$-Laplacian ${\Delta }_{\bar{\partial }}$ on $C^{\infty }$ $L$-valued $(0,p)$-forms. On a Riemann surface, ${\Delta }_{\bar{\partial }}=\frac{1}{2}{\Delta }_d$ when $L={\mathcal{O}}_X$ (the Kähler identity, see 06.04.03), and the harmonic representatives are exactly the kernel of $\bar{\partial }$ in the $C^{\infty }$-section sheaf modulo $\bar{\partial }$-image. For $L={\mathcal{O}}_X$: harmonic $(0,0)$-forms are constants, harmonic $(0,1)$-forms are anti-holomorphic, and the dimension count $\dim H^{0,1}(X)=g$ recovers the genus from the Hodge decomposition.

Computational tool — short exact sequences. For a divisor $D$ on $X$, the inclusion $\mathcal{O}(-D)\hookrightarrow {\mathcal{O}}_X$ has cokernel a sum of skyscraper sheaves at the points of $D$. The associated long exact sequence in Čech cohomology relates ${\check{H}}^{\ast }(X,\mathcal{O}(-D))$ to ${\check{H}}^{\ast }(X,{\mathcal{O}}_X)$ and the local data of $D$:

$$0\to H^0(\mathcal{O}(-D))\to H^0({\mathcal{O}}_X)\to \underset{p\in D}{⨁}\mathbb{C}\to H^1(\mathcal{O}(-D))\to H^1({\mathcal{O}}_X)\to 0.$$

Twisting by an arbitrary line bundle $L$ gives the parallel sequence for $\mathcal{O}(L-D)$ relative to $\mathcal{O}(L)$. Iterating these sequences as $D$ varies produces an inductive backbone: every line bundle is connected to ${\mathcal{O}}_X$ via a chain of single-point twists, and the long exact sequence transfers cohomological data along the chain. This is the engine of the Riemann-Roch proof on a curve (Donaldson §11): start with the structure sheaf $L={\mathcal{O}}_X$, where $\dim H^0=1$ and $\dim H^1=g$, and bump along divisors to read off $\chi (L)=\mathrm{deg}L+1-g$ from the connecting maps.

Mittag-Leffler and Cousin obstructions. Forster's emphasis (§26-§27) frames Čech-$H^1$ as the obstruction theory for globalising local meromorphic data. Given prescribed principal parts $({\mu }_p{)}_{p\in S}$ at finitely many points $S\subset X$, when is there a global meromorphic section $f$ of $L$ with those principal parts? The data $({\mu }_p)$ is a section of $\mathcal{M}(L)/\mathcal{O}(L)$; the connecting map $\delta :H^0(\mathcal{M}(L)/\mathcal{O}(L))\to {\check{H}}^1(X,L)$ records the obstruction class. Cousin I is the additive version: prescribe sums of principal parts; obstruction in ${\check{H}}^1(X,{\mathcal{O}}_X)$. Cousin II is the multiplicative version: prescribe local divisors; obstruction in ${\check{H}}^2(X,\mathbb{Z})$ via the exponential sheaf sequence. On a non-compact Riemann surface (Behnke-Stein 1949: every non-compact Riemann surface is Stein), all such obstructions vanish, so Mittag-Leffler and both Cousin problems are universally solvable. On a compact $X$, the obstruction is genuine and finite-dimensional, controlled by the Riemann-Roch / Serre-duality apparatus.

Picard group via Čech-$H^1$. The line bundles on $X$ up to isomorphism form the Picard group $\mathrm{Pic}(X)$. The exponential sheaf sequence $0\to \mathbb{Z}\to {\mathcal{O}}_X\stackrel{\exp (2\pi i\cdot )}{\to }{\mathcal{O}}_X^{\times }\to 0$ produces the long exact sequence

$$\cdots \to H^1(X,\mathbb{Z})\to H^1(X,{\mathcal{O}}_X)\to H^1(X,{\mathcal{O}}_X^{\times })\stackrel{c_1}{\to }{H}^2(X,\mathbb{Z})\to H^2(X,{\mathcal{O}}_X)\to \cdots ,$$

with $H^2(X,{\mathcal{O}}_X)=0$ by dimensional vanishing. The first Chern class $c_1:\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\times })\to H^2(X,\mathbb{Z})=\mathbb{Z}$ is the degree of a line bundle. The kernel of $c_1$, the degree-zero Picard group ${\mathrm{Pic}}^0(X)$, is the quotient $H^1(X,{\mathcal{O}}_X)/H^1(X,\mathbb{Z})={\mathbb{C}}^g/\Lambda$ — the Jacobian variety. The Čech-cohomological description of every line bundle by its transition cocycle $g_{ij}$ is the input data for this entire identification.

Synthesis. The Čech construction on a compact Riemann surface is the engine that converts the combinatorial data of a line bundle — open cover, transition functions, local sections — into the finite-dimensional cohomological dimensions $\dim H^0(L)$ and $\dim H^1(L)$. Read in the opposite direction, the four-step proof of the Čech-Dolbeault comparison turns the analytic data of $\bar{\partial }$-harmonic forms into combinatorial cocycles, and the dimensional vanishing of $H^p$ for $p\ge 2$ is read off from the real-two-dimensionality of the underlying surface. The geometric content of every line-bundle theorem on a curve — Riemann-Roch, Serre duality, Brill-Noether, Picard, Mittag-Leffler, Cousin — is a statement about these two finite-dimensional Čech-cohomology spaces and their pairings under the canonical line-bundle structure. Putting these together, the full apparatus of compact-curve geometry — divisor counting, period theory, Jacobians and theta divisors, the moduli of line bundles — is organised by a single Čech complex of length two on a single fine-enough cover, with the colimit over refinements producing the cover-independent answer that the abstract sheaf-cohomology framework demands. The dimension-one case generalises to the Cartan-Serre-Behnke-Grauert framework that controls the cohomology of every holomorphic vector bundle on every complex manifold, with Stein and compact-Kähler cases as the two extremal regimes.

Full proof set [Master]

Lemma ($\bar{\partial }$-Poincaré, Dolbeault-Grothendieck on a disc). Let $D\subset \mathbb{C}$ be an open disc and $\eta =f(z,\bar{z})d\bar{z}$ a smooth $(0,1)$-form on $D$. Then there exists a smooth function $u$ on $D$ with $\bar{\partial }u=\eta$.

Proof. Define $u(z)=\frac{1}{2\pi i}{\int }_D\frac{f(\zeta ,\bar{\zeta })}{\zeta -z}d\zeta \wedge d\bar{\zeta }$. The standard Cauchy-Pompeiu formula gives $\bar{\partial }u=fd\bar{z}$ as smooth functions on $D$; the singularity of $1/(\zeta -z)$ is integrable in two real dimensions, so the integral is well-defined and smooth in $z$. The identity is the local solvability statement for $\bar{\partial }$ on a disc. The same construction works for $L$-valued forms: choose a holomorphic frame $e_{\alpha }$ of $L$ on $D$, write $\eta =f\cdot e_{\alpha }\otimes d\bar{z}$, solve $\bar{\partial }v=fd\bar{z}$ scalar-wise, and set $u=v\cdot e_{\alpha }$; then $\bar{\partial }u=(\bar{\partial }v)\cdot e_{\alpha }=\eta$ since the frame is holomorphic. $\square$

Lemma (fine-sheaf acyclicity). On a paracompact Hausdorff space $X$, a sheaf admitting smooth partitions of unity has $H^q(X,\mathcal{F})=0$ for all $q\ge 1$.

Proof. Smooth partitions of unity provide a contracting chain homotopy on the canonical Godement-resolution complex computing sheaf cohomology, killing all positive-degree cohomology. The argument is standard (Godement Topologie algébrique et théorie des faisceaux 1958, Ch. II §3); the sheaves ${\mathcal{A}}^{0,q}(L)$ on a Riemann surface admit smooth partitions of unity since the underlying manifold does. $\square$

Lemma (compact-Riemann-surface ellipticity of ${\Delta }_{\bar{\partial }}$). On a compact Riemann surface $X$ with a Hermitian metric and a holomorphic line bundle $L$, the $\bar{\partial }$-Laplacian $\Delta_{\bar\partial} = \bar\partial \bar\partial^ + \bar\partial^* \bar\partial$isaself-adjointellipticoperatorofordertwoon$C^\inftyL$-valued$(0, p)$-forms with finite-dimensional kernel.*

Proof. The principal symbol of ${\Delta }_{\bar{\partial }}$ at a covector $\xi$ is $|\xi {|}^2$ (after the local-to-global gluing via the Hermitian metric on $L$ and the Kähler form on $X$), positive and non-degenerate, so ${\Delta }_{\bar{\partial }}$ is elliptic. Self-adjointness follows from $(\bar{\partial }{)}^{\ast }=-{\bar{\partial }}^{\ast }$ and standard formal-adjoint computations. On a closed manifold, an elliptic self-adjoint operator has discrete spectrum with finite-dimensional eigenspaces; in particular the kernel $\ker ({\Delta }_{\bar{\partial }})$ is finite-dimensional. $\square$

Theorem (Čech-Dolbeault comparison on a compact Riemann surface). Statement and proof as in the Intermediate-tier Key theorem section.

Proof. The Intermediate-tier proof goes through using the three lemmas above as packaged inputs: the $\bar{\partial }$-Poincaré lemma (Lemma 1) ensures the Dolbeault complex is a sheaf-theoretic resolution of $\mathcal{O}(L)$; fine-sheaf acyclicity (Lemma 2) makes it an acyclic resolution computing derived-functor cohomology; ellipticity of ${\Delta }_{\bar{\partial }}$ (Lemma 3) gives the finite-dimensionality of the harmonic representatives. The Čech-versus-derived-functor comparison on a paracompact Hausdorff space (Cartan-Leray) supplies the bridge to Čech cohomology. Dimensional vanishing $H^p=0$ for $p\ge 2$ comes from ${\mathcal{A}}^{0,q}(L)=0$ for $q\ge 2$ on a real two-manifold. $\square$

Corollary (cohomology of $\mathcal{O}(d)$ on ${\mathbb{P}}^1$). On ${\mathbb{P}}^1$, $\dim H^0({\mathbb{P}}^1,\mathcal{O}(d))=d+1$ for $d\ge 0$ and zero for $d\le -1$; $\dim H^1({\mathbb{P}}^1,\mathcal{O}(d))=-d-1$ for $d\le -2$ and zero otherwise.

Proof. The standard cover $\mathcal{U}=\{U_0,U_1\}$ has affine pieces $\mathbb{C}$ each, with overlap ${\mathbb{C}}^{\ast }$ and transition function $z\mapsto z^d$ for $\mathcal{O}(d)$. The Čech complex is

$$0\to \mathbb{C}[z]\oplus \mathbb{C}[w]\stackrel{\delta }{\to }\mathbb{C}[z,z^{-1}]\to 0,$$

with $\delta (f(z),g(w))=z^{d}g(z^{-1})-f(z)$ on the overlap.

For $d\ge 0$: the image of $\delta$ contains all polynomials in $z$ (from $-f(z)$) and all monomials $z^{d-k}$ for $k\ge 0$ (from $g(w)=w^k$), i.e. all monomials of degree $\le d$. Together these span the full Laurent polynomial ring, so ${\check{H}}^1=0$. The kernel: pairs with $f(z)=z^{d}g(z^{-1})$ in $\mathbb{C}[z,z^{-1}]$. The right side is a polynomial in $z^{d-k}$ for $k\ge 0$ (i.e. $z^d,z^{d-1},\dots ,z^{d-k},\dots$), and the left is a polynomial in $z$. They agree exactly when $f(z)=a_0{z}^d+a_1{z}^{d-1}+\cdots +a_d{z}^0$, a polynomial of degree $\le d$ in $z$ with $d+1$ coefficients. So ${\check{H}}^0={\mathbb{C}}^{d+1}$.

For $d\le -1$: the kernel argument forces $f(z)=z^{d}g(z^{-1})$ with the right side having only negative-power monomials and the left only non-negative — agreement only at zero. So ${\check{H}}^0=0$. The image: $-f(z)$ contributes non-negative powers, and $z^{d}g(z^{-1})=z^d{\sum }_{k\ge 0}{b}_k{z}^{-k}={\sum }_{k\ge 0}{b}_k{z}^{d-k}$ contributes monomials $z^d,z^{d-1},z^{d-2},\dots$. Together they miss the monomials $z^{-1},z^{-2},\dots ,z^{d+1}$ — exactly $|d|-1=-d-1$ monomials when $d\le -2$, and zero when $d=-1$. So ${\check{H}}^1={\mathbb{C}}^{-d-1}$ for $d\le -2$ and zero for $d=-1$.

The dimensions match the Serre-duality identity $\dim H^1(\mathcal{O}(d))=\dim H^0(\mathcal{O}(-d-2))$, with $K_{{\mathbb{P}}^1}=\mathcal{O}(-2)$. $\square$

Corollary (long exact sequence on a divisor short exact sequence). For a holomorphic line bundle $L$ on a compact Riemann surface $X$ and a point $p\in X$, the short exact sequence $0\to \mathcal{O}(L-p)\to \mathcal{O}(L)\to {\mathbb{C}}_p\to 0$ produces the long exact sequence

$$0\to H^0(L-p)\to H^0(L)\to \mathbb{C}\to H^1(L-p)\to H^1(L)\to 0.$$

Proof. The skyscraper sheaf ${\mathbb{C}}_p$ has $H^0(X,{\mathbb{C}}_p)=\mathbb{C}$ and $H^q(X,{\mathbb{C}}_p)=0$ for $q\ge 1$ (a skyscraper is supported on a point, and global sections of a constant sheaf on a point are the stalk). The associated long exact sequence in cohomology of sheaves cuts off after degree one by dimensional vanishing on the curve, giving the stated five-term sequence. $\square$

Corollary (Mittag-Leffler obstruction class). A Mittag-Leffler datum $\mu \in H^0(X,\mathcal{M}(L)/\mathcal{O}(L))$ for a line bundle $L$ on a compact Riemann surface $X$ is the principal-parts data of a global meromorphic section iff its image $\delta (\mu )\in {\check{H}}^1(X,L)$ under the connecting homomorphism vanishes.

Proof. The short exact sequence $0\to \mathcal{O}(L)\to \mathcal{M}(L)\to \mathcal{M}(L)/\mathcal{O}(L)\to 0$ gives the connecting map $\delta :H^0(\mathcal{M}(L)/\mathcal{O}(L))\to H^1(\mathcal{O}(L))$. By exactness, $\mu$ comes from a global meromorphic section iff $\delta (\mu )=0$. The class $\delta (\mu )$ is the Mittag-Leffler obstruction. On a non-compact (Stein) Riemann surface, $H^1(X,\mathcal{O}(L))=0$ by Theorem B (Cartan-Serre, applied via Behnke-Stein 1949), so every Mittag-Leffler datum lifts. On a compact $X$, the obstruction is finite-dimensional and genuine. $\square$

Connections [Master]

• Riemann-Roch theorem for compact Riemann surfaces 06.04.01. Riemann-Roch reads $\chi (L)=\dim H^0(L)-\dim H^1(L)=\mathrm{deg}L+1-g$ for a line bundle $L$ on $X$. Both cohomology groups are computed by the Čech construction of the present unit; the divisor-bumping inductive proof of Riemann-Roch in Donaldson §11 is a sequence of long exact sequences in Čech cohomology associated to single-point divisor twists, with the Čech machinery as the load-bearing input.

• Holomorphic line bundle on a Riemann surface 06.05.02. The objects of this unit are line bundles on a compact $X$; the transition cocycle $g_{ij}\in {\mathcal{O}}^{\times }(U_i\cap U_j)$ that defines $L$ is itself a Čech one-cocycle (multiplicatively) classifying $L$ in $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\times })$. The sheaf $\mathcal{O}(L)$ of holomorphic sections of $L$ is the additive Čech-cohomology object whose dimensions encode the linear-system data of $L$.

• Hodge decomposition on a compact Riemann surface 06.04.03. The Čech-Dolbeault comparison for $L={\mathcal{O}}_X$ gives ${\check{H}}^1(X,{\mathcal{O}}_X)\cong H^{0,1}(X)$, the anti-holomorphic summand of the Hodge decomposition. The genus identity $\dim H^1(X,{\mathcal{O}}_X)=g$ is read off from the Hodge symmetry $h^{0,1}=h^{1,0}=g$. The harmonic-representative formulation of ${\check{H}}^1$ on a compact Riemann surface is precisely the input the Hodge decomposition provides.

• Serre duality on a curve 06.04.04. The Serre-duality pairing ${\check{H}}^1(X,L)\times H^0(X,K_X\otimes L^{-1})\to \mathbb{C}$ uses the Čech-cohomology cup product on ${\check{H}}^1(X,L)$ and the trace map ${\check{H}}^1(X,K_X)\to \mathbb{C}$. The unit ships with the Čech construction as the source side of the pairing; the Serre-duality unit shows the pairing is non-degenerate, identifying ${\check{H}}^1(X,L)$ with the global sections of the canonical-twisted dual.

• Holomorphic 1-form 06.06.01. The canonical bundle $K_X$ of holomorphic 1-forms is the Serre-duality dual side; $\dim H^0(X,K_X)=g$ by definition of the genus, and $\dim {\check{H}}^1(X,K_X)=1$ via the trace isomorphism. The Čech construction of the present unit specialised to $L=K_X$ is the input data for the Serre-duality computation.

• Picard group of a Riemann surface. $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\times })$ is the Čech-cohomology group of the multiplicative unit sheaf, classifying holomorphic line bundles up to isomorphism. The exponential sheaf sequence $0\to \mathbb{Z}\to {\mathcal{O}}_X\to {\mathcal{O}}_X^{\times }\to 0$ together with the present unit's ${\check{H}}^1(X,{\mathcal{O}}_X)\cong {\mathbb{C}}^g$ gives the structural identification ${\mathrm{Pic}}^0(X)\cong {\mathbb{C}}^g/\Lambda$ — the Jacobian.

• Sheaf cohomology of schemes 04.03.03. The algebraic-side counterpart: Čech cohomology of quasi-coherent sheaves on a separated scheme, with Cartan's comparison theorem replacing Cartan-Leray for paracompact spaces. The two pictures agree on a smooth projective curve over $\mathbb{C}$ via the GAGA theorem, identifying analytic Čech of $\mathcal{O}(L)$ with algebraic Čech of the corresponding invertible sheaf.

• Mittag-Leffler problem and Cousin I/II. The connecting homomorphism $\delta :H^0(\mathcal{M}(L)/\mathcal{O}(L))\to {\check{H}}^1(X,L)$ records the Mittag-Leffler obstruction, recovering Cousin I (additive) and Cousin II (multiplicative) as cohomological-obstruction problems. On a non-compact Riemann surface (Behnke-Stein 1949), ${\check{H}}^1$ vanishes for every line bundle, so every Mittag-Leffler datum lifts; on a compact $X$, the obstruction is finite-dimensional and explicitly computable.

• Jacobian variety 06.06.03. The Jacobian $\mathrm{Jac}(X)=H^0(X,{\Omega }_X^1{)}^{\ast }/H_1(X,\mathbb{Z})$ has tangent space $H^1(X,{\mathcal{O}}_X)$ at the origin; the Čech computation of $H^1(X,{\mathcal{O}}_X)$ on a fine-enough cover is the input for this identification, with the period-matrix integration giving the lattice structure.

Historical & philosophical context [Master]

Eduard Čech introduced the construction now bearing his name in Théorie générale de l'homologie dans un espace quelconque (Fundamenta Mathematicae 19 (1932) 149-183) ^{[Čech 1932]}, originally as a homology theory for general topological spaces defined combinatorially via nerves of open covers. The cohomological dressing with the alternating-sum-of-restriction differential on $\prod \mathcal{F}(U_{i_0\cdots i_p})$ is the modern formulation, due to Henri Cartan in the 1950s.

The application to holomorphic-bundle data on Riemann surfaces was developed by Heinrich Behnke and Karl Stein in Entwicklung analytischer Funktionen auf Riemannschen Flächen (Mathematische Annalen 120 (1949) 430-461) ^{[Behnke-Stein 1949]}: every non-compact Riemann surface is a Stein manifold (the Behnke-Stein theorem), and on a Stein manifold every Mittag-Leffler problem and every Cousin problem is solvable through the vanishing of the relevant Čech-$H^1$ obstruction. The 1949 paper systematised the Čech-cohomological treatment of holomorphic-bundle data on a Riemann surface and established the framework that Forster's Lectures on Riemann Surfaces (Springer GTM 81, 1981) ^[Forster] §12-§17 codifies as the standard analytic textbook treatment.

The algebraic-side parallel was developed by Jean-Pierre Serre in Faisceaux Algébriques Cohérents (Annals of Mathematics 61 (1955) 197-278) ^[Serre], the founding paper of sheaf-cohomological algebraic geometry. Serre's framing — Čech cohomology of coherent sheaves on a complex algebraic variety, computed on a finite affine cover and matched with derived-functor cohomology — became the template for every subsequent cohomology computation in algebraic geometry. The 1955 paper proved affine vanishing, computed $H^i({\mathbb{P}}^n,\mathcal{O}(d))$, and established Serre duality, all through Čech on standard covers; the holomorphic-line-bundle case on a compact Riemann surface is the Čech-1932 + Behnke-Stein-1949 + Serre-1955 synthesis.

Donaldson's Riemann Surfaces (Oxford GTM 22, 2011) §10-§12 ^[Donaldson] presents the Čech construction together with the Dolbeault and harmonic-projection identifications, integrating the analytic and topological perspectives into the proof of Riemann-Roch in §11 via divisor-bumping. The Hodge-theoretic input — that ${\check{H}}^1(X,{\mathcal{O}}_X)$ is identified with anti-holomorphic 1-forms, and via Serre duality with $H^0(X,K_X{)}^{\ast }$ — is the bridge to the period theory of Chapter 11. Griffiths-Harris Principles of Algebraic Geometry (Wiley 1978) ^{[Griffiths-Harris]} §0.4 + §1.2 develops the same Čech machinery for line bundles on a complex manifold of arbitrary dimension, with the curve case as the dimension-one specialisation.

Bibliography [Master]

[object Promise]