06.09.02 · riemann-surfaces / stein

Cartan's Theorems A and B for Stein Riemann surfaces

shipped3 tiersLean: none

Anchor (Master): Cartan-Serre 1953 *Un théorème de finitude concernant les variétés analytiques compactes* (CRAS 237) and Cartan séminaire 1951–53 (originators of Theorems A and B); Forster *Lectures on Riemann Surfaces* §25–§26 and §29; Hörmander *An Introduction to Complex Analysis in Several Variables* Ch. III; Grauert-Remmert *Theory of Stein Spaces* (Grundlehren 236)

Intuition [Beginner]

Take a non-compact Riemann surface $X$ — the complex line, the punctured plane, an annulus, the upper half-plane, any compact curve with a few points removed. The previous unit on Stein Riemann surfaces 06.09.01 explains why every such $X$ carries an abundant supply of global holomorphic functions: enough to separate any two points, enough to keep the holomorphic hull of any compact set compact. Cartan's Theorems A and B are the cohomological pay-off of that abundance. They say two things about the analytic data that lives on $X$ .

The first theorem (Theorem A) says that whatever local analytic gadget you can attach to $X$ — a sheaf of holomorphic sections of a line bundle, a sheaf of meromorphic functions with prescribed poles, an ideal sheaf cutting out a discrete set — is generated by global sections. Pick any point and any local section there. You can write that local section as a finite sum of global sections of the same sheaf, with coefficients that are local holomorphic functions. There is no obstruction to lifting a local datum to global data, point by point.

The second theorem (Theorem B) says the higher cohomology of any such sheaf vanishes: $H^{q} (X, F) = 0$ for every $q \geq 1$ . Cohomology in this sense measures the obstruction to gluing local data into global data along a cover. Theorem B says: on a Stein surface, that obstruction is always zero. Mittag-Leffler problems, Cousin problems, line-bundle classifications — every classical existence question on $X$ becomes a corollary of one identity.

Visual [Beginner]

A schematic of a non-compact Riemann surface $X$ shown as a curving open band, with a sheaf $F$ depicted as a varying-thickness ribbon attached to it. A point $x \in X$ is highlighted, with the stalk $F_{x}$ drawn as a small disc above it. An arrow labelled "evaluate at $x$ " runs from the global-section box $Γ (X, F)$ down to the stalk $F_{x}$ , with the arrow shown as surjective (Theorem A). Beside the surface a vertical sequence of cohomology groups $H^{0}, H^{1}, H^{2}, \dots$ is drawn with $H^{0} = Γ (X, F)$ non-empty and the higher entries crossed out (Theorem B).

Worked example [Beginner]

Take $X = C$ , the complex line, and the structure sheaf $F = O_{X}$ of holomorphic functions. The stalk $O_{X, x}$ at a point $x$ is the ring of convergent power series at $x$ — every germ of a holomorphic function near $x$ . Theorem A says: every convergent power series at $x$ can be written as a finite combination of entire functions, with coefficients that are themselves germs of holomorphic functions at $x$ .

Concrete instance: pick $x = 0$ and the germ $f (z) = 1 + z + z^{2} /2! + z^{3} /3! + \dots = e^{z}$ at $0$ . The germ $f$ is a single global section of $O_{C}$ — the entire function $e^{z}$ itself — so the generation is direct: take the global section $e^{z}$ and the coefficient $1 \in O_{C, 0}$ .

Pick instead a more interesting germ at $0$ , say $g (z) = z + z^{2} + z^{3}$ . Two global sections suffice to generate it: the entire functions $z$ and $z^{2}$ . Write $g = z + z^{2} \cdot (1 + z)$ , so $g = z \cdot 1 + z^{2} \cdot (1 + z)$ uses local coefficients $1, 1 + z \in O_{C, 0}$ and global sections $z, z^{2} \in Γ (C, O_{C})$ .

Theorem B says: $H^{1} (C, O) = 0$ , $H^{2} (C, O) = 0$ , and so on. The first vanishing is the classical Mittag-Leffler theorem: every prescribed pattern of principal parts at a discrete set of points in $C$ is realised by a global meromorphic function. The higher vanishings are forced by real dimension two — there is no $H^{q}$ to talk about for $q \geq 2$ on a Riemann surface to begin with.

What this tells us: on $C$ , the Stein property collapses every classical existence problem about meromorphic functions, holomorphic line bundles, and prescribed local data into a single statement about the structure sheaf, and Theorems A and B promote that statement from the structure sheaf to every analytic sheaf on the surface.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

Theorem A for a Stein Riemann surface $X$ and a coherent analytic sheaf $F$ on $X$ says:

A. Every section of $F$ defined locally near a point extends to a global section of $F$ . B. For every point $x \in X$ , the stalk $F_{x}$ is generated as a module over $O_{X, x}$ by the images of global sections $Γ (X, F)$ . C. The cohomology $H^{0} (X, F)$ is finite-dimensional. D. The sheaf $F$ is the same as its sheaf of global sections.

Hint

The key word is generated as a module: not "every local section equals a global section", but "every local section is a finite combination of global sections with local coefficients".

Answer

B. Feedback-correct: Theorem A says exactly that the stalk $F_{x}$ is generated as a module over $O_{X, x}$ by the images of global sections — every germ at $x$ is a finite combination of global sections with local coefficients. Feedback-wrong: A is too strong (a local section need not equal a global section); C is unrelated and is moreover false on a Stein RS where $H^{0}$ is typically infinite-dimensional; D conflates sections with the sheaf itself.

Formal definition [Intermediate+]

Let $X$ be a Stein manifold (in this unit, almost always a non-compact connected Riemann surface, with $X$ Stein automatic by Behnke-Stein 06.09.01). Let $O_{X}$ be the structure sheaf of holomorphic functions on $X$ .

Definition (coherent analytic sheaf). A sheaf of $O_{X}$ -modules $F$ on $X$ is coherent when it is locally finitely generated (every $x \in X$ has a neighbourhood $U$ and a surjection $O_{U}^{n} ↠ F ∣_{U}$ ) and locally finitely related (the kernel of any local surjection $O_{U}^{n} \to F ∣_{U}$ is itself locally finitely generated). Coherence is the analytic version of "locally finitely presented" in commutative algebra; Oka's coherence theorem (1950) gives that $O_{X}$ is coherent over itself, and standard constructions (kernels, cokernels, finite tensor products of coherent sheaves) preserve coherence.

Theorem A (Cartan-Serre 1953). Let $X$ be a Stein manifold and $F$ a coherent analytic sheaf on $X$ . For every $x \in X$ the stalk $F_{x}$ is generated as an $O_{X, x}$ -module by the images of global sections. Equivalently, the natural evaluation map

Γ (X, F) \otimes_{Γ (X, O_{X})} O_{X, x} ⟶ F_{x}

is surjective for every $x \in X$ .

Theorem B (Cartan-Serre 1953). Let $X$ be a Stein manifold and $F$ a coherent analytic sheaf on $X$ . Then $H^{q} (X, F) = 0$ for every $q \geq 1$ .

Combined consequence. The functor $Γ (X, -)$ from coherent analytic sheaves on $X$ to $Γ (X, O_{X})$ -modules is exact: every short exact sequence of coherent sheaves $0 \to F^{'} \to F \to F^{''} \to 0$ on $X$ gives a short exact sequence of global sections $0 \to Γ (X, F^{'}) \to Γ (X, F) \to Γ (X, F^{''}) \to 0$ . Coherent sheaves on a Stein manifold behave like modules over the ring $Γ (X, O_{X})$ .

Notation. $Γ (X, F)$ denotes global sections; $F_{x}$ the stalk at $x$ ; $H^{q} (X, F)$ the $q$ -th sheaf cohomology, computed via Čech with respect to a Stein cover or via the Dolbeault fine resolution 06.04.05; $O_{X, x}$ the local ring of holomorphic germs at $x$ .

Counterexamples to common slips

Theorem B fails on compact Riemann surfaces of positive genus. For $X$ compact of genus $g \geq 1$ , $H^{1} (X, O_{X}) = C^{g} \neq = 0$ . Theorem B's hypothesis is the Stein property, which non-compact non-singular complex manifolds may or may not have in dimension $\geq 2$ but have automatically in dimension one once non-compact.
Coherence is required. The vanishing $H^{q} (X, F) = 0$ does not hold for every sheaf — only coherent ones. The constant sheaf $\underline{Z}_{X}$ on a torus minus a point has nonzero $H^{1}$ , for instance, because it is not coherent as an $O_{X}$ -module.
Theorem A is generation, not local equality. The map $Γ (X, F) \to F_{x}$ (evaluation at $x$ ) need not be surjective; the surjective map is the evaluation tensored with the local ring, $Γ (X, F) \otimes O_{X, x} \to F_{x}$ . Local coefficients are essential.
Theorem B does not say $H^{0}$ vanishes. Global sections of $F$ form $H^{0} (X, F)$ , which is in general infinite-dimensional on a non-compact $X$ . The vanishing kicks in only at $q \geq 1$ .

Key theorem with proof [Intermediate+]

Theorem (Cartan-Serre 1953, dimension-one case). Let $X$ be a non-compact connected Riemann surface and $F$ a coherent analytic sheaf on $X$ . Then $H^{q} (X, F) = 0$ for every $q \geq 1$ (Theorem B), and the natural evaluation map $Γ (X, F) \otimes_{Γ (X, O_{X})} O_{X, x} \to F_{x}$ is surjective for every $x \in X$ (Theorem A).

Proof. The argument runs in five steps: a Stein exhaustion, the $\overset{ˉ}{\partial}$ -vanishing on each piece, propagation to coherent sheaves by the structure-theorem reduction, a Mittag-Leffler limit passage to the union, and the deduction of Theorem A from Theorem B.

Step 1 — Stein exhaustion. Behnke-Stein 06.09.01 gives a Runge exhaustion $X_{1} ⋐ X_{2} ⋐ \dots$ with $⋃_{n} X_{n} = X$ and each $X_{n}$ relatively compact and Runge in $X$ . Each $X_{n}$ is itself Stein (open subset of a non-compact RS, or directly verified from the strictly subharmonic exhaustion restricted to $X_{n}$ ).

Step 2 — vanishing for $O_{X}$ on each piece. On each relatively compact pseudoconvex (Stein) $X_{n}$ , Hörmander's weighted $L^{2}$ -existence theorem 06.04.05 solves $\overset{ˉ}{\partial} u = g$ for every smooth $\overset{ˉ}{\partial}$ -closed $g$ on $X_{n}$ . The Dolbeault comparison 06.04.05 identifies $H^{1} (X_{n}, O_{X_{n}})$ with the cokernel of $\overset{ˉ}{\partial} : C^{\infty} (X_{n}) \to C_{(0, 1)}^{\infty} (X_{n})$ , which Hörmander's theorem makes zero. For $q \geq 2$ , real dimension two on a Riemann surface forces $H^{q} (X_{n}, O_{X_{n}}) = 0$ on dimensional grounds. So $H^{q} (X_{n}, O_{X_{n}}) = 0$ for every $q \geq 1$ .

Step 3 — extension to coherent $F$ on $X_{n}$ . Locally on a relatively compact open $U \subset X_{n}$ the coherent sheaf $F$ admits a finite-rank presentation $O_{U}^{n_{1}} \to O_{U}^{n_{0}} \to F ∣_{U} \to 0$ . Iterating the kernel-of-presentation argument and using Hilbert's syzygy theorem in dimension one (the local rings $O_{X_{n}, x}$ are regular of Krull dimension one, hence have global homological dimension one), one obtains a finite resolution $0 \to O_{X_{n}}^{m} \to O_{X_{n}}^{n} \to F \to 0$ on each $X_{n}$ — a length-one resolution because the surface is one-dimensional. Apply the long exact sequence in cohomology to this short exact sequence of sheaves: each term involves only $H^{q} (X_{n}, O_{X_{n}})$ for $q \in {0, 1}$ , and $H^{q}$ vanishes for $q \geq 1$ by Step 2. The connecting maps then force $H^{q} (X_{n}, F) = 0$ for $q \geq 1$ as well.

Step 4 — limit passage. The full surface $X = ⋃_{n} X_{n}$ is the increasing union of the Stein opens. The Čech complex of $F$ on $X$ relative to a refinement of the Stein-cover of each $X_{n}$ is the inverse limit (in cochain degree, direct limit in section degree) of the Čech complexes on each $X_{n}$ . The Mittag-Leffler condition on the inverse system ${H^{0} (X_{n}, F)}$ — each restriction $H^{0} (X_{n + 1}, F) \to H^{0} (X_{n}, F)$ has dense image in the natural Fréchet topology, by the Runge property of $X_{n}$ in $X_{n + 1}$ — is the input that allows the inverse limit to commute with cohomology. Concretely: a cocycle $ξ$ on $X$ restricts to cocycles $ξ_{n}$ on $X_{n}$ , each of which is a coboundary by Step 3, and the Mittag-Leffler condition lets one assemble the local primitives $η_{n}$ into a global primitive $η$ on $X$ by uniformly approximating each correction term $η_{n + 1} - η_{n}$ on $X_{n - 1}$ (Runge) and summing the geometric corrections. The assembled $η$ is a global primitive of $ξ$ , and $H^{q} (X, F) = 0$ for $q \geq 1$ .

Step 5 — Theorem A from Theorem B. Given $x \in X$ and a germ $σ \in F_{x}$ , choose a neighbourhood $U$ of $x$ on which $σ$ extends to a section $\tilde{σ} \in Γ (U, F)$ . Consider the sheaf morphism $O_{X} \to F$ defined locally as $1 \mapsto \tilde{σ}$ on $U$ and zero outside, with kernel $K$ a coherent subsheaf of $O_{X}$ (locally cut out by the annihilator of $\tilde{σ}$ ). Wrap into the short exact sequence $0 \to K \to O_{X} \to O_{X} / K \to 0$ , with $O_{X} / K$ the skyscraper or rank-one piece supported on the locus where $\tilde{σ}$ has a nonzero annihilator. The long exact sequence in cohomology gives

Γ (X, O_{X}) \to Γ (X, O_{X} / K) \to H^{1} (X, K) \to \dots,

and Theorem B (applied to the coherent $K$ ) makes $H^{1} (X, K) = 0$ , so the connecting map is zero and the global-section evaluation surjects onto $O_{X} / K$ , hence onto a generator of $F_{x}$ at $x$ . The germ $σ$ thereby lies in the $O_{X, x}$ -submodule generated by global sections, and Theorem A holds. $□$

The five-step structure follows Forster §29 (exhaustion + Hörmander on each piece + Mittag-Leffler limit + A-from-B deduction) and dovetails with Hörmander Ch. III (the weighted $L^{2}$ -method as the analytic engine, with the same exhaustion and limit-passage logic). The dimension-one specialisation is shorter than the general Cartan-Serre argument: the structure theorem on the local rings $O_{X, x}$ in dimension one gives length-one resolutions automatically, where the higher-dimensional case needs an induction on the resolution length.

Bridge. The proof binds two pieces of upstream machinery into one statement. From 06.09.01 comes the Stein property: a strictly subharmonic exhaustion exists on every non-compact RS, and the holomorphic-convex hull of every compact set is compact. From 06.04.05 comes the Hilbert-space framework for $\overset{ˉ}{\partial}$ : $\overset{ˉ}{\partial}$ is a closed densely-defined operator on $L^{2}$ -spaces, with a self-adjoint Hodge-Laplace and finite-dimensional kernels on compact pieces, and Hörmander's weighted estimate solves the equation $\overset{ˉ}{\partial} u = g$ with quantitative control on Stein domains. The theorem fuses these inputs: Stein gives the geometry on which the $\overset{ˉ}{\partial}$ -PDE is solvable, the $L^{2}$ -PDE produces the cohomological vanishing on each Stein piece, and the exhaustion glues the local vanishings into a global one. Theorem A is then a one-step exact-sequence consequence of Theorem B. Cartan and Serre's 1953 paper extracted this chain in arbitrary complex dimension, with the length-one syzygy of dimension one replaced by an induction on syzygy length using Hilbert's general theorem on regular local rings; the dimension-one case in Forster §29 is the cleanest expression because the syzygy collapses to one step. The downstream consequence is that every classical existence problem on a non-compact RS — Cousin I, Cousin II, Mittag-Leffler, Runge approximation, Picard-group triviality — collapses to a single corollary of $H^{1} (X, F) = 0$ , with $F$ the appropriate ideal or quotient sheaf encoding the problem's local data.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Let $X$ be a non-compact Riemann surface and $D = p_{1} + \dots + p_{k}$ a divisor of $k$ distinct points. Show that the sequence

0 \to O_{X} (- D) \to O_{X} \to O_{D} \to 0

induces an isomorphism $Γ (X, O_{X}) /Γ (X, O_{X} (- D)) ≅ Γ (X, O_{D}) = C^{k}$ .

Hint

Use the long exact sequence in cohomology and Theorem B to kill the $H^{1}$ -correction term.

Answer

The short exact sequence is the standard "ideal-of-a-divisor" sequence; $O_{X} (- D)$ is the sheaf of holomorphic functions vanishing along $D$ , and $O_{D}$ is the skyscraper sheaf $⨁_{i} C_{p_{i}}$ on the discrete set ${p_{1}, \dots, p_{k}}$ . The long exact sequence in cohomology reads

0 \to Γ (X, O_{X} (- D)) \to Γ (X, O_{X}) \to Γ (X, O_{D}) \to H^{1} (X, O_{X} (- D)) \to H^{1} (X, O_{X}) \to \dots .

By Theorem B applied to the coherent sheaves $O_{X} (- D)$ and $O_{X}$ , both $H^{1}$ -terms vanish. The connecting map's target is therefore zero, and the four-term exact sequence reduces to a short exact

0 \to Γ (X, O_{X} (- D)) \to Γ (X, O_{X}) \to Γ (X, O_{D}) \to 0,

giving $Γ (X, O_{X}) /Γ (X, O_{X} (- D)) ≅ Γ (X, O_{D}) = C^{k}$ . Rubric: full credit for the long exact sequence + double application of Theorem B + identification with $C^{k}$ .

Exercise 4 (medium, short-answer).

Show that the Cousin I problem on a non-compact Riemann surface $X$ — given an open cover ${U_{i}}$ and meromorphic functions $f_{i} \in M (U_{i})$ with $f_{i} - f_{j} \in O (U_{i} \cap U_{j})$ , find a global meromorphic $f \in M (X)$ with $f - f_{i} \in O (U_{i})$ — is always solvable.

Hint

The obstruction is the cohomology class $[{f_{i} - f_{j}}] \in H^{1} (X, O_{X})$ .

Answer

The differences $g_{ij} = f_{i} - f_{j}$ are holomorphic on $U_{i} \cap U_{j}$ by hypothesis, satisfy the cocycle condition $g_{ij} + g_{j k} + g_{k i} = 0$ on triple overlaps automatically, and define a Čech 1-cocycle on the cover ${U_{i}}$ with values in $O_{X}$ . The class $[g_{ij}] \in H^{1} (X, O_{X})$ is the Cousin I obstruction: it is the coboundary of a global $f$ exactly when $g_{ij} = h_{i} - h_{j}$ for some $h_{i} \in O (U_{i})$ , in which case $f$ defined by $f = f_{i} - h_{i}$ on $U_{i}$ is well-defined globally and meromorphic, with $f - f_{i} \in O (U_{i})$ . By Theorem B applied to the coherent sheaf $O_{X}$ on the Stein surface $X$ , $H^{1} (X, O_{X}) = 0$ , so the cocycle is a coboundary and the Cousin I problem is solvable. Rubric: full credit for the cocycle + Theorem B + assembly of global $f$ .

Exercise 5 (medium, short-answer).

Show that the Cousin II problem on a non-compact connected Riemann surface $X$ is always solvable: given a cover ${U_{i}}$ and non-zero meromorphic $f_{i} \in M^{*} (U_{i})$ with $f_{i} / f_{j} \in O^{*} (U_{i} \cap U_{j})$ , there exists a global $f \in M^{*} (X)$ with $f / f_{i} \in O^{*} (U_{i})$ .

Hint

The Cousin II obstruction lives in $H^{1} (X, O_{X}^{*})$ , which the exponential exact sequence relates to $H^{1} (X, O_{X})$ and $H^{2} (X, Z)$ .

Answer

The ratios $g_{ij} = f_{i} / f_{j}$ are holomorphic and non-vanishing on the overlaps, satisfy the multiplicative cocycle condition $g_{ij} \cdot g_{j k} = g_{ik}$ , and define a class in $H^{1} (X, O_{X}^{*}) = Pic (X)$ . The exponential sequence $0 \to Z \to O_{X} \to O_{X}^{*} \to 0$ gives the long exact sequence

H^{1} (X, Z) \to H^{1} (X, O_{X}) \to H^{1} (X, O_{X}^{*}) \to H^{2} (X, Z) .

By Theorem B (Cartan-Serre on the Stein surface $X$ ), $H^{1} (X, O_{X}) = 0$ . By the topology of a non-compact connected real-2-surface, $H^{2} (X, Z) = 0$ . Squeezed between two zeros, $H^{1} (X, O_{X}^{*}) = 0$ , and the multiplicative cocycle is a coboundary $g_{ij} = h_{i} / h_{j}$ with $h_{i} \in O^{*} (U_{i})$ . The function $f$ defined by $f = f_{i} / h_{i}$ on $U_{i}$ glues to a global non-zero meromorphic function on $X$ with $f / f_{i} = 1/ h_{i} \in O^{*} (U_{i})$ . Rubric: full credit for the exponential sequence + Theorem B + topological vanishing + assembly.

Exercise 6 (medium, short-answer).

Let $X$ be a non-compact Riemann surface, $K \subset X$ compact, and $f \in O (U)$ holomorphic on a neighbourhood $U$ of $K$ . Show, using Theorem B for $O_{X}$ and a cutoff $\overset{ˉ}{\partial}$ argument, that for every $ε > 0$ there exists $F \in O (X)$ with $sup_{K} ∣ F - f ∣ < ε$ (Runge approximation, with the standard topological hypothesis on $X ∖ K$ ).

Hint

Multiply $f$ by a smooth cutoff supported in $U$ , set up a global $\overset{ˉ}{\partial}$ -equation on $X$ , and use Theorem B to solve.

Answer

Choose a smooth cutoff $χ$ with $χ \equiv 1$ on a neighbourhood of $K$ and $supp χ ⋐ U$ . The function $χ f$ extends by zero to a smooth function on $X$ , and $\overset{ˉ}{\partial} (χ f) = (\overset{ˉ}{\partial} χ) f$ is supported in the compact annulus $U ∖ K$ . The class $[\overset{ˉ}{\partial} (χ f)] \in H^{1} (X, O_{X})$ vanishes by Theorem B (alternatively, the Dolbeault comparison 06.04.05 identifies this $H^{1}$ with the cokernel of $\overset{ˉ}{\partial}$ , and the cokernel is zero on the Stein $X$ ). So there exists $u \in C^{\infty} (X)$ with $\overset{ˉ}{\partial} u = \overset{ˉ}{\partial} (χ f)$ . The function $F = χ f - u$ is then holomorphic on $X$ (since $\overset{ˉ}{\partial} F = 0$ globally) and equals $f - u$ on a neighbourhood of $K$ . Hörmander's weighted estimate 06.04.05 makes $∥ u ∥_{L^{2} (ϕ)}$ controllable by $∥ \overset{ˉ}{\partial} χ \cdot f ∥_{L^{2} (ϕ)}$ , which is small in sup norm on $K$ (with the Sobolev embedding from $L^{2}$ to sup, given the topological hypothesis on $X ∖ K$ ). Iterating with shrinking $ε$ produces uniform approximation of $f$ on $K$ by global holomorphic $F$ . Rubric: full credit for the cutoff + $\overset{ˉ}{\partial}$ -solve + Theorem B + sup-norm control.

Exercise 7 (hard, short-answer).

Let $X$ be a Stein Riemann surface and $0 \to F^{'} \to F \to F^{''} \to 0$ a short exact sequence of coherent analytic sheaves on $X$ . Show that the global-section sequence $0 \to Γ (X, F^{'}) \to Γ (X, F) \to Γ (X, F^{''}) \to 0$ is exact, and conclude that $Γ (X, -)$ is an exact functor on coherent sheaves.

Hint

The connecting homomorphism in the long exact sequence lands in $H^{1} (X, F^{'})$ , which Theorem B kills.

Answer

The long exact sequence in sheaf cohomology produced by the short exact sequence reads

0 \to Γ (X, F^{'}) \to Γ (X, F) \to Γ (X, F^{''}) δ H^{1} (X, F^{'}) \to H^{1} (X, F) \to H^{1} (X, F^{''}) \to H^{2} (X, F^{'}) \to \dots

By Theorem B applied to the coherent sheaf $F^{'}$ on the Stein surface $X$ , $H^{1} (X, F^{'}) = 0$ . The connecting map $δ$ has zero target, hence is the zero map. The four-term exact sequence at the left collapses to the desired short exact $0 \to Γ (X, F^{'}) \to Γ (X, F) \to Γ (X, F^{''}) \to 0$ . Since $Γ (X, -)$ is left exact in general (true for any sheaf cohomology computation) and right exactness has just been established for arbitrary short exact sequences of coherent sheaves, $Γ (X, -)$ is exact on the category of coherent analytic sheaves on $X$ . Rubric: full credit for the long exact sequence + Theorem B + exactness deduction.

Exercise 8 (hard, short-answer).

Let $X$ be a Stein Riemann surface and $F$ a coherent analytic sheaf on $X$ . Show that for every finite set ${x_{1}, \dots, x_{k}} \subset X$ and every choice of germs $σ_{i} \in F_{x_{i}}$ , there exists a single global section $s \in Γ (X, F)$ whose germ at each $x_{i}$ equals $σ_{i}$ .

Hint

Apply Theorem A at each point separately, then combine the local-coefficient combinations using Cousin / Mittag-Leffler interpolation.

Answer

By Theorem A at each $x_{i}$ , the stalk $F_{x_{i}}$ is generated by global sections: for each $i$ there exist global sections $s_{i}^{(1)}, \dots, s_{i}^{(n_{i})} \in Γ (X, F)$ and germs $a_{i}^{(j)} \in O_{X, x_{i}}$ such that $σ_{i} = \sum_{j} a_{i}^{(j)} \cdot (s_{i}^{(j)})_{x_{i}}$ . The combined collection ${s_{i}^{(j)}}_{i, j}$ is a finite set of global sections of $F$ . The local-coefficient germs $a_{i}^{(j)} \in O_{X, x_{i}}$ together specify a Mittag-Leffler datum on the discrete divisor $D = \sum_{i} x_{i}$ for the finitely many sheaves $O_{X}$ . By Theorem B applied to the structure sheaf and the ideal sheaves $O_{X} (- D)$ , the Mittag-Leffler datum lifts to global sections $A^{(j)} \in Γ (X, O_{X})$ with $A_{x_{i}}^{(j)} = a_{i}^{(j)}$ for each $i$ (the lift exists because the relevant connecting map lands in $H^{1} (X, O_{X} (- D)) = 0$ ). Set $s = \sum_{i, j} A^{(j)} \cdot s_{i}^{(j)} \in Γ (X, F)$ (finite sum, hence well-defined). At each $x_{i}$ , $s_{x_{i}} = \sum_{j} A_{x_{i}}^{(j)} \cdot (s_{i}^{(j)})_{x_{i}} = \sum_{j} a_{i}^{(j)} \cdot (s_{i}^{(j)})_{x_{i}} = σ_{i}$ , as required. Rubric: full credit for Theorem A at each point + Mittag-Leffler lift via Theorem B + global assembly.

Lean formalization [Intermediate+]

Mathlib does not currently formalise Cartan's Theorems A and B for Stein manifolds, nor the dimension-one Riemann-surface specialisation. A proposed signature, in Lean 4 / Mathlib syntax, sketching the target statements:

[object Promise]

The formalisation depends on names that do not currently exist in Mathlib (the coherent-analytic-sheaf predicate, the global-section evaluation map, sheaf cohomology in the analytic setting on a manifold, Hörmander's $L^{2}$ -existence theorem). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

The Cartan-Serre framework was published in two stages: the 1951–53 Cartan séminaire at the École Normale Supérieure, where Henri Cartan and Jean-Pierre Serre developed Theorems A and B for Stein manifolds in arbitrary complex dimension, and the 1953 Comptes Rendus note Un théorème de finitude concernant les variétés analytiques compactes (CRAS 237) ^{[Cartan-Serre 1953]}, which gave the published companion. The séminaire treatment proves the theorems for any Stein manifold $M$ of complex dimension $n$ , with the analytic engine being Cartan-Thullen's exhaustion-by-pseudoconvex-domains and a Mittag-Leffler / Schwartz finiteness limit-passage argument on the Fréchet space of global sections. The dimension-one specialisation runs faster: the local rings $O_{X, x}$ on a Riemann surface are regular of Krull dimension one, hence have global homological dimension one, so every coherent sheaf admits a length-one resolution by free $O_{X}^{k}$ , and the long exact sequence in cohomology immediately reduces to the structure-sheaf case.

Hörmander's $L^{2}$ -PDE proof (1965). The modern analytic engine for Theorem B is Hörmander's 1965 $L^{2}$ -estimates and existence theorems for the $\overset{ˉ}{\partial}$ -operator ^{[Hörmander 1965]} (Acta Math. 113). On a Stein manifold $M$ with strictly plurisubharmonic exhaustion $ϕ$ , the weighted estimate

∥ u ∥_{L^{2} (e^{- ϕ})}^{2} \leq C ∥ g ∥_{L^{2} (e^{- ϕ})}^{2}

for the canonical solution $u$ of $\overset{ˉ}{\partial} u = g$ on $\overset{ˉ}{\partial}$ -closed $g$ is the input that produces the cohomological vanishing in arbitrary complex dimension. The Bochner-Kodaira-Nakano identity bounds the weighted Dolbeault Laplacian below by a positive constant times the curvature of the chosen Hermitian weight; the Riesz representation theorem on the resulting weighted Hilbert space gives a unique $L^{2}$ -bounded solution. In dimension one, the curvature reduces to the Levi form $\partial \overset{ˉ}{\partial} ϕ$ (positive by hypothesis), and the estimate specialises to Hörmander Ch. III's elementary version. The Hörmander route makes the proof fully analytical and avoids the Schwartz-finiteness limit-passage of the Cartan séminaire.

Forster's elementary proof in dimension one. Forster Lectures on Riemann Surfaces §29 ^[Forster] gives an alternative route in dimension one that avoids the $L^{2}$ -method. Forster's setup uses the Runge exhaustion of 06.09.01 and the elementary $\overset{ˉ}{\partial}$ -solvability on each $X_{n}$ (proved by direct integral kernels: the Cauchy-Pompeiu formula on each disc, glued by partitions of unity). The Mittag-Leffler / Schwartz finiteness limit-passage on the Fréchet space of holomorphic functions is the input that promotes vanishing on each $X_{n}$ to vanishing on $X = ⋃_{n} X_{n}$ . Forster's route is cleaner pedagogically; Hörmander's $L^{2}$ -route generalises to higher dimension and to twisted sheaves with curvature constraints (Kodaira vanishing, Nakano vanishing, Demailly's analytic continuation theorems). On a Riemann surface the two routes prove the same theorem.

Cousin problems and the corollary list. Theorem B for the structure sheaf $O_{X}$ on a non-compact Riemann surface is the cohomological reformulation of the classical Cousin I problem (additive principal parts), originally posed by Pierre Cousin in 1895. Theorem B for $O_{X}^{*}$ (via the exponential sequence and the topology of a non-compact connected surface) is the reformulation of Cousin II (multiplicative divisors). The Mittag-Leffler theorem of 1884, proved on the plane by Magnus Mittag-Leffler, is the statement that prescribed-principal-parts data on $C$ is realised by a global meromorphic function — and is the special case $X = C$ of the Cousin I corollary. The classical statement of Runge's theorem (1885), that holomorphic functions on a compact set with simply connected complement are uniformly approximable by polynomials, generalises to Stein Riemann surfaces as: holomorphic functions on a relatively compact pseudoconvex $K \subset X$ are uniformly approximable on $K$ by global $O (X)$ functions, with the standard topological hypothesis " $X ∖ K$ has no relatively compact components" providing the obstruction-free regime.

Higher-dimensional Theorem A / B and GAGA. The Cartan-Serre theorems on Stein manifolds of arbitrary complex dimension are the analytic analogue of Serre's Faisceaux algébriques cohérents (1955) ^{[Serre 1955]} vanishing on affine schemes: for an affine scheme $X = Spec A$ and a quasi-coherent $F$ , $H^{q} (X_{Zar}, F) = 0$ for $q \geq 1$ . The two vanishings — analytic on Stein, algebraic on affine — are mirror statements and the foundation of the dictionary between analytic and algebraic geometry. Serre's GAGA (1956) Géométrie algébrique et géométrie analytique gives the comparison theorem on the projective side: on a smooth projective variety, coherent algebraic sheaves and coherent analytic sheaves have the same cohomology, and the global-section functors agree. The Stein-versus-affine pairing on the open side is more delicate (Stein implies affine for embedded varieties; the converse fails for Serre's 1953 example of a Stein manifold not algebraisable as an affine variety).

The Oka principle. A homotopy-theoretic refinement of Theorem A, due to Kiyoshi Oka 1939 and Hans Grauert 1957, says that on a Stein manifold $X$ every continuous section of a holomorphic fibre bundle is homotopic to a holomorphic section, and the holomorphic sections form a deformation retract of the continuous ones. For line bundles on a Stein RS, the Oka principle reduces to the topological classification ( $Pic (X) = H^{2} (X, Z)$ topologically, which is zero for a non-compact connected RS), recovering the Picard-group triviality. The Oka principle is the "soft" / homotopy-theoretic shadow of Theorem A and remains an active topic in modern complex analysis (Forstnerič 2011 Stein Manifolds and Holomorphic Mappings gives the contemporary monograph treatment).

Stein theory and finite-presentation. On a non-compact RS the Stein property has a third equivalent characterisation: the Fréchet algebra $Γ (X, O_{X})$ separates points and the structure morphism $X \to Specm Γ (X, O_{X})$ to the maximal-ideal spectrum (with the Gelfand topology) is a homeomorphism. The Cartan-Serre theorems convert this Fréchet-algebra picture into a category-equivalence statement: coherent analytic sheaves on the Stein $X$ correspond to finitely-presented modules over $Γ (X, O_{X})$ , with global sections as the equivalence functor. The exactness of $Γ (X, -)$ promoted by Theorem B is exactly the property that distinguishes a "tame" geometric category (Stein, affine) from a "wild" one (compact Kähler, projective).

Synthesis. Cartan's Theorems A and B are the cohomological pay-off of the Stein hypothesis: the structure of a complex manifold on $X$ guaranteeing enough global holomorphic functions for $\overset{ˉ}{\partial}$ -PDE solvability, exhaustion by relatively compact pseudoconvex pieces, and a strictly plurisubharmonic Morse function. Once that geometric hypothesis is in place, Theorem B converts every coherent-sheaf cohomological obstruction in positive degree to zero, and Theorem A converts every local-section datum to a global one. On a non-compact Riemann surface — where Behnke-Stein makes the Stein hypothesis automatic — the two theorems consolidate the classical Mittag-Leffler, Cousin I, Cousin II, Runge, and Picard-group questions into a single corollary $H^{q} (X, F) = 0$ . The dimension-one specialisation runs in five steps, with the length-one syzygy collapsing the Cartan-Serre induction and the Hörmander / Forster $\overset{ˉ}{\partial}$ -solvability furnishing the analytic engine. In higher dimension the proof retains the same skeleton — Stein exhaustion, $L^{2}$ -solvability on each piece, limit passage — with a longer syzygy induction and the Bochner-Kodaira-Nakano curvature estimate doing the heavy lifting. The bridge to algebraic geometry runs through Serre's affine-scheme vanishing and GAGA on projective varieties; the bridge to symplectic topology runs through the Cieliebak-Eliashberg Stein-Weinstein duality (recorded in 06.09.01). On a Riemann surface, the entire analytic spine of complex curve theory is captured by one identity: $H^{1} (X, F) = 0$ for every coherent $F$ , with everything else flowing as a corollary.

Full proof set [Master]

Lemma (Oka coherence). The structure sheaf $O_{X}$ of a complex manifold $X$ is coherent over itself.

Proof. Oka 1950 proves coherence of $O_{X}$ over itself by an explicit local-presentation argument: on a polydisc, the kernel of any surjection $O^{k} \to O$ is finitely generated by an explicit Weierstrass-preparation calculation. The Riemann-surface case is an immediate specialisation (each local ring $O_{X, x}$ is a discrete valuation ring, hence Noetherian and of homological dimension one). $□$

Lemma ( $\overset{ˉ}{\partial}$ -solvability on Stein). Let $X$ be a Stein Riemann surface and $ϕ : X \to R$ a strictly subharmonic exhaustion. For every $g \in C_{(0, 1)}^{\infty} (X)$ that is $\overset{ˉ}{\partial}$ -closed (which is automatic on a 1-dimensional manifold) and has $∥ g ∥_{L^{2} (e^{- ϕ})} < \infty$ , there exists $u \in C^{\infty} (X)$ with $\overset{ˉ}{\partial} u = g$ and $∥ u ∥_{L^{2} (e^{- ϕ})} \leq ∥ g ∥_{L^{2} (e^{- ϕ})}$ .

Proof. Hörmander's $L^{2}$ -method 06.04.05 in dimension one. Equip $X$ with a Hermitian metric whose Kähler form is $ω = i \partial \overset{ˉ}{\partial} ϕ$ (positive on $X$ by strict subharmonicity). Form the weighted $L^{2}$ -space $L_{(0, q)}^{2} (X, e^{- ϕ})$ . The unbounded operator $\overset{ˉ}{\partial} : L_{(0, 0)}^{2} \to L_{(0, 1)}^{2}$ is closed and densely defined, with Hilbert adjoint $\overset{ˉ}{\partial}^{*}$ . The Bochner-Kodaira-Nakano identity in dimension one reads $⟨ \overset{ˉ}{\partial}^{*} g, \overset{ˉ}{\partial}^{*} g ⟩ \geq ∥ g ∥^{2}$ for $g \in dom (\overset{ˉ}{\partial}^{*}) \cap ker \overset{ˉ}{\partial}$ , the lower-bound estimate that promotes $\overset{ˉ}{\partial}$ to a bijection from a closed subspace of $L_{(0, 0)}^{2}$ onto $ker \overset{ˉ}{\partial} \subset L_{(0, 1)}^{2}$ . The Riesz representation theorem on the linear functional $L_{g} : v \mapsto ⟨ v, g ⟩$ on $dom (\overset{ˉ}{\partial}^{*})$ produces $u \in L^{2}$ with $\overset{ˉ}{\partial} u = g$ and $∥ u ∥ \leq ∥ g ∥$ . Elliptic regularity on the Cauchy-Riemann operator (an elliptic first-order operator on a Riemann surface) bootstraps the $L^{2}$ -solution to a $C^{\infty}$ -solution via Sobolev embedding. The Forster §28 alternative (Cauchy-Pompeiu kernels glued by Runge approximation) is an elementary route to the same statement. $□$

Lemma (Vanishing of $H^{q} (X_{n}, O_{X_{n}})$ ). For each relatively compact Stein open $X_{n} ⋐ X$ in the exhaustion, $H^{q} (X_{n}, O_{X_{n}}) = 0$ for every $q \geq 1$ .

Proof. The Dolbeault comparison 06.04.05 identifies $H^{1} (X_{n}, O_{X_{n}})$ with the cokernel $C_{(0, 1)}^{\infty} (X_{n}) / \overset{ˉ}{\partial} C^{\infty} (X_{n})$ . The previous lemma makes $\overset{ˉ}{\partial}$ surjective on $X_{n}$ , so the cokernel is zero. For $q \geq 2$ , the Riemann surface has real dimension two, and the Dolbeault complex is concentrated in cohomological degrees $0$ and $1$ , so $H^{q} = 0$ on dimensional grounds. $□$

Lemma (Vanishing of $H^{q} (X_{n}, F)$ for coherent $F$ ). For every coherent analytic sheaf $F$ on a relatively compact Stein open $X_{n}$ , $H^{q} (X_{n}, F) = 0$ for every $q \geq 1$ .

Proof. Locally on $X_{n}$ , $F$ admits a finite presentation $O_{X_{n}}^{a} \to O_{X_{n}}^{b} \to F \to 0$ . By the Hilbert-syzygy theorem in dimension one (the local rings are regular of Krull dimension one, hence have global homological dimension one), the kernel of the right-hand map is itself locally free of finite rank, giving a length-one resolution

0 \to O_{X_{n}}^{a} \to O_{X_{n}}^{b} \to F \to 0.

The long exact sequence in cohomology and the previous lemma (vanishing of $H^{q} (X_{n}, O_{X_{n}})$ for $q \geq 1$ ) force $H^{q} (X_{n}, F) = 0$ for every $q \geq 1$ — the connecting maps have zero source and zero target. $□$

Lemma (Mittag-Leffler condition on the Stein exhaustion). The inverse system ${H^{0} (X_{n}, F)}$ of global sections of a coherent analytic sheaf $F$ along a Stein exhaustion $X_{1} ⋐ X_{2} ⋐ \dots$ satisfies the Mittag-Leffler condition: each restriction $H^{0} (X_{n + 1}, F) \to H^{0} (X_{n}, F)$ has dense image in the natural Fréchet topology.

Proof. The Runge property of $X_{n}$ in $X_{n + 1}$ (Behnke-Stein lemma in 06.09.01) gives uniform approximability of every $f \in Γ (X_{n}, F)$ on every compact $K \subset X_{n}$ by elements of $Γ (X_{n + 1}, F)$ . The natural Fréchet topology on $Γ (X_{n}, F)$ is the projective limit over compact $K \subset X_{n}$ of sup-norm topologies on $Γ (K, F)$ , and density on each $K$ is precisely Runge approximation. The Mittag-Leffler condition is the abstract formulation of "every term in the inverse system can be uniformly approximated by terms in the next term" — and is what makes the inverse limit commute with cohomology in a Fréchet category. $□$

Theorem (Cartan's Theorem B for Stein RS, full statement). Let $X$ be a non-compact connected Riemann surface and $F$ a coherent analytic sheaf on $X$ . Then $H^{q} (X, F) = 0$ for every $q \geq 1$ .

Proof. Combine the four lemmas. The Stein exhaustion $X_{n} ⋐ X$ exists by Behnke-Stein 06.09.01. On each $X_{n}$ , $H^{q} (X_{n}, F ∣_{X_{n}}) = 0$ for $q \geq 1$ by Lemma 4. The Mittag-Leffler condition on ${H^{0} (X_{n}, F)}$ from Lemma 5 makes the inverse limit commute with cohomology: $H^{q} (X, F) = lim_{n} H^{q} (X_{n}, F ∣_{X_{n}}) = 0$ for $q \geq 1$ . Concretely, given a global Čech 1-cocycle $ξ$ on $X$ , restrict to a 1-cocycle $ξ_{n}$ on each $X_{n}$ ; by Lemma 4 each $ξ_{n}$ is a coboundary $ξ_{n} = d η_{n}$ ; the Mittag-Leffler condition lets one modify each $η_{n}$ by a constant cocycle (uniformly approximable on $X_{n - 1}$ ) to make the sequence ${η_{n}}$ converge in the Fréchet topology to a global $η$ with $d η = ξ$ on every $X_{n}$ , hence on $X$ . The cocycle $ξ$ is a coboundary, $H^{1} (X, F) = 0$ . Higher $H^{q}$ vanish on dimensional grounds. $□$

Corollary (Cartan's Theorem A for Stein RS). Let $X$ be a non-compact connected Riemann surface and $F$ a coherent analytic sheaf on $X$ . For every $x \in X$ , the natural evaluation map $Γ (X, F) \otimes_{Γ (X, O_{X})} O_{X, x} \to F_{x}$ is surjective.

Proof. Given a germ $σ \in F_{x}$ , the local section extension to a neighbourhood $U$ of $x$ is $\tilde{σ} \in Γ (U, F)$ . The kernel sheaf $K$ of the morphism $O_{X} \to F$ defined by $1 \mapsto \tilde{σ}$ on $U$ and zero outside is coherent (kernels of sheaf maps between coherent sheaves are coherent). The short exact sequence $0 \to K \to O_{X} \to O_{X} / K \to 0$ produces

Γ (X, O_{X}) \to Γ (X, O_{X} / K) \to H^{1} (X, K) \to \dots .

By Theorem B applied to the coherent $K$ , $H^{1} (X, K) = 0$ , the connecting map is zero, and the global-section map is surjective onto $O_{X} / K$ . Pulling back along the natural map $O_{X} / K \to F$ at $x$ , the germ $σ$ lies in the image of the evaluation map $Γ (X, F) \otimes O_{X, x} \to F_{x}$ , as required. $□$

Corollary (Exactness of $Γ (X, -)$ ). The functor $Γ (X, -)$ from coherent analytic sheaves on a Stein Riemann surface $X$ to $Γ (X, O_{X})$ -modules is exact.

Proof. Left-exactness holds for any sheaf cohomology computation. Right-exactness on a short exact sequence $0 \to F^{'} \to F \to F^{''} \to 0$ of coherent sheaves follows from Theorem B applied to $F^{'}$ : the connecting map $Γ (X, F^{''}) \to H^{1} (X, F^{'})$ in the long exact sequence has zero target. $□$

Corollary (Cousin I, Cousin II, Mittag-Leffler). On a non-compact connected Riemann surface $X$ , the Cousin I problem (additive principal parts) and the Cousin II problem (multiplicative divisors) are always solvable; in particular the classical Mittag-Leffler theorem holds in the form: for any prescribed principal-parts datum at a discrete set of points of $X$ , there exists a global meromorphic function realising it.

Proof. Cousin I obstruction lies in $H^{1} (X, O_{X}) = 0$ by Theorem B applied to the coherent $O_{X}$ . Cousin II obstruction lies in $H^{1} (X, O_{X}^{*})$ , and the exponential sequence $0 \to Z \to O_{X} \to O_{X}^{*} \to 0$ together with $H^{2} (X, Z) = 0$ (non-compact connected real-2-surface) and Theorem B for $O_{X}$ squeezes $H^{1} (X, O_{X}^{*}) = 0$ between two zeros. Mittag-Leffler is the special case Cousin I + the principal-parts sheaf $M_{X} / O_{X}$ , identified with a coherent skyscraper supported on a discrete divisor. $□$

Connections [Master]

Stein Riemann surfaces 06.09.01. The geometric input to Theorems A and B: the existence of a strictly subharmonic exhaustion, holomorphic-convex hulls, and holomorphic separability. Behnke-Stein 1949 makes the hypothesis automatic on every non-compact RS; the unit 06.09.01 establishes this and equips the theorem environment of 06.09.02 with the analytic geometry needed to run the Cartan-Serre proof.
Hilbert-space PDE for $\overset{ˉ}{\partial}$ 06.04.05. The analytic engine of Theorem B in the Hörmander route: on a Stein manifold with strictly plurisubharmonic exhaustion, the weighted $L^{2}$ -existence theorem solves $\overset{ˉ}{\partial} u = g$ for every $\overset{ˉ}{\partial}$ -closed $g$ , with quantitative norm control. The Dolbeault comparison from 06.04.05 then identifies the cokernel of $\overset{ˉ}{\partial}$ with $H^{1} (X, O_{X})$ , and the $L^{2}$ -existence makes that cokernel zero. The same machinery on each piece $X_{n}$ of a Stein exhaustion produces the local vanishings used in Step 2 of the proof above.
Čech cohomology of holomorphic line bundles 06.04.02. The cohomological language in which Theorem B is stated: $H^{q} (X, F)$ is computed via Čech with respect to a Stein cover, or equivalently via the Dolbeault fine resolution. The Mittag-Leffler / Cousin / Picard corollaries of Theorems A and B are reformulations of the cohomological vanishing of specific Čech cohomology groups attached to the structure sheaf and the multiplicative-units sheaf. The non-compact case here is the negation of the compact-case generic non-vanishing recorded in 06.04.02 (where $H^{1}$ on a compact RS of genus $g$ has dimension $g$ ).
Holomorphic line bundle on a Riemann surface 06.05.02. The Picard group $Pic (X) = H^{1} (X, O_{X}^{*})$ vanishes on a non-compact RS by Theorem B applied through the exponential sequence, recovering the holomorphic-triviality of every line bundle. The line-bundle classification on a Riemann surface is therefore concentrated in the compact case (06.04.01, 06.04.04); the non-compact Stein case trivialises the entire structure.
Riemann-Roch theorem for compact Riemann surfaces 06.04.01. The compact-case Riemann-Roch identity $dim H^{0} (L) - dim H^{1} (L) = de g L + 1 - g$ has the cohomological correction $H^{1}$ explicitly because the Stein property fails. The non-compact Stein case kills the right-hand cohomology entirely. The dichotomy compact-vs-Stein on a Riemann surface is the dichotomy substantive-Riemann-Roch-correction-vs-no-correction.
Serre duality on a curve 06.04.04. Serre duality $H^{1} (X, L) ≅ H^{0} (X, K_{X} \otimes L^{- 1})^{*}$ on a compact RS pairs the $H^{1}$ that survives compact regimes with a dual $H^{0}$ . On a non-compact Stein RS, Theorem B kills the left-hand side and the "duality" degenerates: both sides are zero or infinite-dimensional with no canonical pairing. Theorems A and B replace the duality pairing with the simpler statement that the cohomological obstruction vanishes outright.
Meromorphic function 06.01.05. The Mittag-Leffler theorem (1884 plane case, Behnke-Stein 1949 RS case) is the corollary of Theorem B saying every prescribed principal-parts datum on a discrete divisor of a non-compact RS is realised by a global meromorphic function. The Cousin I framework records the same statement in cohomological language.
Riemann mapping theorem 06.01.06. The Riemann mapping theorem identifies every simply connected proper open subset of $C$ as biholomorphic to the unit disc. Combined with Theorem B, every such region admits a strictly subharmonic exhaustion, $H^{1} (D, F) = 0$ for coherent $F$ , and the structure sheaf's global-section algebra is the complete metric space $O (D)$ . The Stein-rigidity of simply connected non-compact Riemann surfaces feeds directly into Theorem B's universal vanishing.
Hartogs phenomenon 06.07.02. Hartogs' theorem on automatic extension of holomorphic functions across a punctured polydisc in $C^{n}$ (for $n \geq 2$ ) is the obstruction to non-Stein-ness in higher dimension: the punctured polydisc fails Theorem B precisely because Hartogs-extension forces too many global sections, breaking the holomorphic-convex-hull condition. On a Riemann surface this phenomenon does not occur, and the higher-dimensional analytic-extension question is replaced by the universal vanishing of Theorem B.
Holomorphic functions of several complex variables 06.07.01. The higher-dimensional Cartan-Serre statement of Theorems A and B, valid on every Stein manifold of dimension $n$ , is the analytic engine of every multi-variable function-theoretic existence problem. The dimension-one specialisation collapses to the curve case treated here; the higher-dimensional case is the substantive content of the Cartan séminaire and Hörmander's monograph.

Historical & philosophical context [Master]

Henri Cartan and Jean-Pierre Serre proved Theorems A and B in the 1951–53 Cartan séminaire at the École Normale Supérieure ^{[Cartan 1951–53]}, with the published companion appearing in 1953 in the Comptes Rendus de l'Académie des Sciences as Cartan-Serre 1953 Un théorème de finitude concernant les variétés analytiques compactes ^{[Cartan-Serre 1953]} (CRAS 237, 128–130). The séminaire treatment defined a Stein manifold of arbitrary complex dimension as a holomorphically convex and holomorphically separable complex manifold (extending Stein 1951's foundational definition), and proved that on such a manifold every coherent analytic sheaf has all positive-degree cohomology vanishing (Theorem B) and is generated by global sections at every point (Theorem A). The proof in the séminaire used Cartan-Thullen's exhaustion-by-pseudoconvex-domains and a Schwartz-finiteness limit-passage on the Fréchet space of global sections; the dimension-one specialisation, recorded in Forster Lectures on Riemann Surfaces §29 ^[Forster], runs faster because the local rings on a Riemann surface are regular of Krull dimension one.

Lars Hörmander's 1965 $L^{2}$ -estimates and existence theorems for the $\overset{ˉ}{\partial}$ -operator ^{[Hörmander 1965]} (Acta Math. 113, 89–152) supplied the modern analytic engine for Theorem B. The weighted $L^{2}$ -method on a strictly plurisubharmonic exhaustion produces the cohomological vanishing as a direct corollary of the Bochner-Kodaira-Nakano curvature estimate and the Riesz representation theorem, replacing the séminaire's Schwartz-finiteness argument with a self-contained PDE proof. Hörmander's monograph An Introduction to Complex Analysis in Several Variables ^{[Hörmander HSCV]} (North-Holland 1973, Ch. III) gives the textbook treatment.

Hans Grauert and Reinhold Remmert's Theory of Stein Spaces ^{[Grauert-Remmert]} (Springer Grundlehren 236, 1979) is the canonical reference for Stein theory in the analytic-spaces setting (extending the manifold framework to allow analytic singularities). Theorems A and B in the Grauert-Remmert framework apply to coherent sheaves on Stein analytic spaces, where the local rings need no longer be regular but are Noetherian by Oka coherence. The Riemann-surface case here is the smooth specialisation in dimension one.

Pierre Cousin's 1895 paper Sur les fonctions de $n$ variables complexes (Acta Math. 19, 1–61) posed the Cousin I and Cousin II problems on bidiscs in $C^{n}$ , providing the classical motivation for the cohomological framework that Cartan-Serre and Stein later formalised. Mittag-Leffler's 1884 theorem on $C$ (Acta Math. 4, 1–79) is the planar special case of Cousin I, which Theorems A and B generalise to every non-compact Riemann surface and to every Stein manifold. Behnke and Stein's 1949 Entwicklung analytischer Funktionen auf Riemannschen Flächen ^{[Behnke-Stein 1949]} (Math. Ann. 120, 430–461) closed the dimension-one half of the problem by establishing that every non-compact Riemann surface is Stein, which made the Cartan-Serre theorems unconditional in that setting.

Jean-Pierre Serre's Faisceaux algébriques cohérents ^{[Serre 1955]} (Ann. of Math. 61, 1955, 197–278) developed the algebraic-side counterpart on schemes: for an affine scheme $X = Spec A$ and a quasi-coherent sheaf $F$ , $H^{q} (X_{Zar}, F) = 0$ for $q \geq 1$ . The pairing of Cartan-Serre 1953 (analytic, on Stein manifolds) with Serre 1955 (algebraic, on affine schemes) established the coherent-cohomology vanishing as a uniform principle across the analytic and algebraic categories, and Serre's GAGA in 1956 (Géométrie algébrique et géométrie analytique) made the comparison theorem precise on the projective side.

Bibliography [Master]

[object Promise]

Prerequisites

06.09.01
06.04.05

Tier anchors

beginner: Forster *Lectures on Riemann Surfaces* §25–§26 informal
intermediate: Forster *Lectures on Riemann Surfaces* §25–§26; Hörmander *An Introduction to Complex Analysis in Several Variables* §3
master: Cartan-Serre 1953 *Un théorème de finitude concernant les variétés analytiques compactes* (CRAS 237) and Cartan séminaire 1951–53 (originators of Theorems A and B); Forster *Lectures on Riemann Surfaces* §25–§26 and §29; Hörmander *An Introduction to Complex Analysis in Several Variables* Ch. III; Grauert-Remmert *Theory of Stein Spaces* (Grundlehren 236)

References

TODO_REF
Cartan 1951–53 séminaire — Théorèmes A et B · Séminaire Cartan, ENS, exposés 1951–53 (originator of Theorems A and B for Stein manifolds)
TODO_REF
Cartan-Serre 1953 — Un théorème de finitude concernant les variétés analytiques compactes · C. R. Acad. Sci. Paris 237, 128–130 (published companion to the Cartan séminaire; originator)
TODO_REF
Forster — Lectures on Riemann Surfaces · Springer GTM 81, §25–§26 (Cousin problems and Theorem B in dimension one) and §29 (Stein theory)
TODO_REF
Hörmander — An Introduction to Complex Analysis in Several Variables · North-Holland 1973, Ch. III (the $L^2$-method for $\bar\partial$ and Theorem B on Stein manifolds)
TODO_REF
Hörmander 1965 — $L^2$-estimates and existence theorems for the $\bar\partial$-operator · Acta Math. 113, 89–152 (modern $L^2$-PDE proof of Theorem B)
TODO_REF
Grauert-Remmert — Theory of Stein Spaces · Springer Grundlehren 236, 1979 (canonical reference for Stein theory and coherent analytic sheaves)
TODO_REF
Serre 1955 — Faisceaux algébriques cohérents · Ann. of Math. 61, 197–278 (algebraic-side counterpart: vanishing on affine schemes)
TODO_REF
Behnke-Stein 1949 — Entwicklung analytischer Funktionen auf Riemannschen Flächen · Math. Ann. 120, 430–461 (every non-compact RS is Stein; foundational input)

Reviewer

TBD

Estimated time

beginner: 14m
intermediate: 40m
master: 66m