Cousin II (multiplicative)

Intuition [Beginner]

Imagine you live on a non-compact Riemann surface $X$ and you want to build a single global meromorphic function with a prescribed pattern of zeros and poles. At each point you choose, you specify whether the function should vanish or have a pole there, and to what order. Cousin II is the multiplicative patching version: each local recipe is a meromorphic function $f_i$ on a patch $U_i$, and on overlaps the ratios $f_i/f_j$ must be holomorphic and non-vanishing — meaning the two patches agree on which points are zeros, which are poles, and with what orders, and they only differ by a factor that is everywhere finite and non-zero.

The classical case on the complex line is the Weierstrass product theorem from 1876. Choose a sequence $a_n\to \infty$ in $\mathbb{C}$ and a desired multiplicity $m_n$ at each $a_n$. Weierstrass builds a global entire function whose zeros are exactly the $a_n$ with the right multiplicity, by taking an infinite product of factors, one for each $a_n$, where each factor is a polynomial-corrected version of $(1-z/a_n)$ designed to make the product converge. The Cousin II problem on a Riemann surface is the natural generalisation: the local recipes are now meromorphic, the patching condition is multiplicative, and the question is whether the patched data assembles into a single global meromorphic function.

Pierre Cousin posed this question in his 1895 thesis as the second of his two problems. On a non-compact Riemann surface, the answer is always yes — every Cousin II datum is solvable, no matter how the prescribed zeros and poles are arranged. The reason is that two cohomological obstructions both vanish: one analytic, from the Stein structure, and one topological, from the surface having no top-dimensional homology.

Visual [Beginner]

A non-compact Riemann surface $X$ is shown as an open band, covered by overlapping patches $U_1,U_2,U_3$ in alternating tones. On each patch a meromorphic recipe $f_i$ is depicted with circles marking prescribed zeros and starbursts marking prescribed poles. Where two patches overlap, the ratio $f_i/f_j$ is drawn as a smooth non-vanishing curve free of any markings. Beneath the surface a single curve labelled $f$ runs across all of $X$, with circles and starbursts only where the local recipes asked for them. An arrow from the patched data to the global $f$ is labelled "Theorem B + topology: $\mathrm{Pic}(X)=0$", recording the two-fold vanishing that makes the assembly work.

Worked example [Beginner]

Take $X=\mathbb{C}$ and ask for an entire function with simple zeros at every positive integer $n=1,2,3,\dots$ and no other zeros. The naive guess — multiply $(1-z/n)$ over all $n\ge 1$ — does not converge: for fixed $z$, the factors approach $1$ from below, and the logarithms behave like $-z/n$, summing to a divergent harmonic-like series.

Weierstrass's correction is to multiply each factor by an exponential whose Taylor expansion cancels the slow-decay term in the logarithm. For this example, multiplying $(1-z/n)$ by $\exp (z/n)$ already suffices: the corrected factor at index $n$ becomes

$$E_1(z/n)=(1-z/n)\exp (z/n),$$

whose logarithm has Taylor expansion $-z^2/(2n^2)+O(z^3/n^3)$, decaying like $|z{|}^2/n^2$ for fixed $z$ and large $n$. Multiplying these corrected factors over $n\ge 1$ gives a product that converges absolutely and uniformly on every compact subset of $\mathbb{C}$, and the limit $f$ is entire on $\mathbb{C}$ with a simple zero at each $n\ge 1$ and no other zeros.

Check: at $n=1$, the product factor $(1-z)\exp (z)$ vanishes simply at $z=1$, and every other factor is non-zero at $z=1$, so $f$ has a simple zero at $1$. The same argument applies at every positive integer. The exponential correction $\exp (z/n)$ is everywhere non-zero, so it does not change the zero locus of any factor; it only adjusts the everywhere-non-vanishing part, which the problem leaves free.

What this tells us: on the simplest non-compact Riemann surface, a Cousin II datum is solvable by an explicit convergent product with concrete elementary factors. The Weierstrass trick — multiply each factor by an exponential whose Taylor expansion cancels the slow term — is a hands-on instance of the cohomological vanishing $\mathrm{Pic}(\mathbb{C})=0$ that Theorem B and the topology of $\mathbb{C}$ together produce on every non-compact Riemann surface.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a Riemann surface and ${\mathcal{O}}_X$ its structure sheaf. Write ${\mathcal{O}}_X^{\ast }$ for the sheaf of non-vanishing holomorphic functions on $X$ — a sheaf of multiplicative abelian groups whose stalk at $x\in X$ is the multiplicative group of units in the local ring ${\mathcal{O}}_{X,x}$. Write ${\mathcal{M}}_X^{\ast }$ for the sheaf of non-zero meromorphic functions, with stalk the multiplicative group of the field of fractions of ${\mathcal{O}}_{X,x}$ minus zero. The inclusion ${\mathcal{O}}_X^{\ast }\hookrightarrow {\mathcal{M}}_X^{\ast }$ is a sheaf morphism whose quotient

$${\mathcal{D}}_X={\mathcal{M}}_X^{\ast }/{\mathcal{O}}_X^{\ast }$$

is the divisor sheaf. Its stalk at $x$ is the group of locally finite formal divisors ${\sum }_k{m}_k\cdot p_k$ in a neighbourhood of $x$, isomorphic to ${\mathbb{Z}}^{(\text{discrete points near }x)}$ with the order-of-vanishing map identifying a meromorphic germ with its divisor.

Definition (Cousin II datum). A Cousin II datum on $X$ is a pair $(\mathcal{U},\{f_i\})$ consisting of an open cover $\mathcal{U}=\{U_i{\}}_{i\in I}$ of $X$ and non-zero meromorphic functions $f_i\in {\mathcal{M}}^{\ast }(U_i)$ such that the ratios $f_i/f_j$ are non-vanishing and holomorphic on every overlap $U_i\cap U_j$, i.e. $f_i/f_j\in {\mathcal{O}}^{\ast }(U_i\cap U_j)$.

Definition (solution). A Cousin II datum $(\mathcal{U},\{f_i\})$ is solvable when there exists a global meromorphic function $f\in {\mathcal{M}}^{\ast }(X)$ such that $f/f_i\in {\mathcal{O}}^{\ast }(U_i)$ for every $i$. The Cousin II problem on $X$ is the question of whether every Cousin II datum is solvable.

Equivalent reformulations.

• Sheaf-theoretic. A Cousin II datum is the same as a global section of the divisor sheaf, $\sigma \in \Gamma (X,{\mathcal{D}}_X)$: each $f_i$ projects to a local section $\overline{f_i}\in \Gamma (U_i,{\mathcal{D}}_X)$, and the holomorphic-and-non-vanishing-ratio condition $f_i/f_j\in {\mathcal{O}}^{\ast }(U_i\cap U_j)$ is exactly $\overline{f_i}=\overline{f_j}$ on overlaps. A solution $f\in {\mathcal{M}}^{\ast }(X)$ projects to the same global section $\sigma =\overline{f}$. The Cousin II problem asks whether the global-section map $\Gamma (X,{\mathcal{M}}_X^{\ast })\to \Gamma (X,{\mathcal{D}}_X)$ is surjective.

• Cohomological. The short exact sequence $0\to {\mathcal{O}}_X^{\ast }\to {\mathcal{M}}_X^{\ast }\to {\mathcal{D}}_X\to 0$ produces a long exact sequence in cohomology with connecting map

$$\delta :\Gamma (X,{\mathcal{D}}_X)\to H^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X).$$

The image $\delta (\sigma )\in \mathrm{Pic}(X)$ is the Cousin II obstruction: the holomorphic line bundle determined by the cocycle of ratios $f_i/f_j$. The datum is solvable when $\delta (\sigma )=0$, equivalently when this line bundle is holomorphically isomorphic to the structure-sheaf bundle ${\mathcal{O}}_X$. The Cousin II problem on $X$ is solvable for every datum when $\delta$ is the zero map, which happens whenever $\mathrm{Pic}(X)=0$.

• Čech. On a fine enough cover $\mathcal{U}$, the ratios $g_{ij}=f_i/f_j$ on overlaps satisfy the cocycle condition $g_{ij}\cdot g_{jk}\cdot g_{ki}=1$ on triple overlaps automatically and define a Čech 1-cocycle in ${\check{Z}}^1(\mathcal{U},{\mathcal{O}}_X^{\ast })$ — exactly the transition cocycle of a holomorphic line bundle. The datum is solvable when $\{g_{ij}\}$ is a coboundary $g_{ij}=h_i/h_j$ for some $h_i\in {\mathcal{O}}^{\ast }(U_i)$, and the global solution is then $f{|}_{U_i}=f_i/h_i$.

The exponential exact sequence. The bridge between additive and multiplicative cohomology on a complex manifold is the exponential sheaf sequence:

$$0\to 2\pi i\mathbb{Z}\to {\mathcal{O}}_X\stackrel{\exp }{\to }{\mathcal{O}}_X^{\ast }\to 0,$$

with kernel the constant sheaf $2\pi i\mathbb{Z}$ (locally on a contractible patch, $\exp (g)=h$ has a holomorphic logarithm $g$ unique up to addition of $2\pi i\mathbb{Z}$). The associated long exact sequence in cohomology contains the segment

$$H^1(X,{\mathcal{O}}_X)\stackrel{{\exp }_{\ast }}{\to }{H}^1(X,{\mathcal{O}}_X^{\ast })\stackrel{c_1}{\to }{H}^2(X,2\pi i\mathbb{Z})\to H^2(X,{\mathcal{O}}_X),$$

where $c_1$ is the first Chern class map: it sends a holomorphic line bundle to its underlying topological line bundle, classified by $H^2(X,\mathbb{Z})$. Identifying $H^2(X,2\pi i\mathbb{Z})\cong H^2(X,\mathbb{Z})$ by the constant $2\pi i$, the segment reads $H^1(X,{\mathcal{O}}_X)\to \mathrm{Pic}(X)\to H^2(X,\mathbb{Z})$.

Notation. ${\mathcal{O}}_X^{\ast }$ denotes the sheaf of non-vanishing holomorphic functions on $X$; ${\mathcal{M}}_X^{\ast }$ the sheaf of non-zero meromorphic functions; ${\mathcal{D}}_X={\mathcal{M}}_X^{\ast }/{\mathcal{O}}_X^{\ast }$ the divisor sheaf; $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\ast })$ the Picard group of holomorphic line bundles; $c_1:\mathrm{Pic}(X)\to H^2(X,\mathbb{Z})$ the first-Chern-class connecting map; $E_p(z)=(1-z)\exp \left(z+z^2/2+\cdots +z^p/p\right)$ the Weierstrass elementary factor of order $p$.

Counterexamples to common slips

• Cousin II is the multiplicative problem. The ratios $f_i/f_j$ are required to lie in ${\mathcal{O}}_X^{\ast }$ (holomorphic and non-vanishing). The additive Cousin I problem replaces ratios with differences and asks for ${\mathcal{O}}_X$-valued cocycles instead.

• Cousin II is governed by a different sheaf and a different topology. The Cousin I obstruction lives in $H^1(X,{\mathcal{O}}_X)$, an analytic group. The Cousin II obstruction lives in $H^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X)$, which by the exponential sequence sits between $H^1(X,{\mathcal{O}}_X)$ and $H^2(X,\mathbb{Z})$. Cousin II therefore involves a topological ingredient — the second integer cohomology — that Cousin I does not.

• Cousin II is dimension-and-topology conditional. On every non-compact Riemann surface, Cousin II is solvable: $H^1(X,{\mathcal{O}}_X)=0$ by Theorem B, $H^2(X,\mathbb{Z})=0$ by the topology of a non-compact connected real $2$-manifold, so $\mathrm{Pic}(X)=0$. On a compact Riemann surface of genus $g\ge 1$, $\mathrm{Pic}(X)$ is a non-zero group (an extension of $\mathbb{Z}$ by the Jacobian ${\mathrm{Pic}}^0(X)$, a $g$-dimensional complex torus), and a generic Cousin II datum is not solvable. In higher complex dimension, Cousin II can fail even on Stein manifolds when $H^2(X,\mathbb{Z})\ne 0$.

• The divisor sheaf is not the meromorphic-units sheaf. The sheaf ${\mathcal{M}}_X^{\ast }$ records all non-zero meromorphic functions, including their everywhere-non-zero factors; the quotient ${\mathcal{D}}_X={\mathcal{M}}_X^{\ast }/{\mathcal{O}}_X^{\ast }$ retains only the divisor information (zeros and poles with multiplicities). A Cousin II datum is a section of ${\mathcal{D}}_X$, not of ${\mathcal{M}}_X^{\ast }$; the difference is exactly the freedom in the everywhere-non-vanishing factor.

• Cousin II is not Cousin I composed with the exponential. The exponential sequence relates ${\mathcal{O}}_X$ and ${\mathcal{O}}_X^{\ast }$, and on a non-compact RS it makes the two cohomological obstructions vanish together. The relation is not "Cousin I implies Cousin II" — both rest on Theorem B, and Cousin II additionally requires the topological vanishing $H^2(X,\mathbb{Z})=0$. On a compact RS, Cousin I has a $g$-dimensional space of obstructions and Cousin II has a non-zero $\mathrm{Pic}(X)$, with the Chern class map $c_1$ surjecting onto $\mathbb{Z}$.

Key theorem with proof [Intermediate+]

Theorem (Cousin II on a non-compact Riemann surface). Let $X$ be a connected non-compact Riemann surface. Every Cousin II datum on $X$ is solvable: for every open cover $\mathcal{U}=\{U_i\}$ of $X$ and every collection of non-zero meromorphic functions $f_i \in \mathcal{M}^(U_i)$with$f_i / f_j \in \mathcal{O}^(U_i \cap U_j)$oneveryoverlap,thereexistsaglobalmeromorphicfunction$f \in \mathcal{M}^(X)$with$f / f_i \in \mathcal{O}^(U_i)$forevery$i$.

Proof. The argument runs in five steps: package the datum as a multiplicative Čech 1-cocycle in ${\mathcal{O}}_X^{\ast }$, identify the Cousin II obstruction with a class in $H^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X)$, route this group through the exponential sequence into $H^2(X,\mathbb{Z})$ and $H^1(X,{\mathcal{O}}_X)$, invoke the topological and analytic vanishing of those two groups on a non-compact RS, and assemble the global meromorphic solution from the resulting coboundary trivialisation.

Step 1 — multiplicative cocycle assembly. For each pair $(i,j)$ with $U_i\cap U_j\ne \varnothing$, set $g_{ij}=f_i/f_j$ on $U_i\cap U_j$. By hypothesis $g_{ij}\in {\mathcal{O}}^{\ast }(U_i\cap U_j)$. On every triple overlap $U_i\cap U_j\cap U_k$, the algebraic identity $$ g_{ij} \cdot g_{jk} \cdot g_{ki} = (f_i / f_j) (f_j / f_k) (f_k / f_i) = 1 $$ holds pointwise, so $\{g_{ij}\}$ is a Čech 1-cocycle in the alternating Čech complex ${\check{C}}^{\ast }(\mathcal{U},{\mathcal{O}}_X^{\ast })$ for the cover $\mathcal{U}$ with values in the multiplicative-units sheaf. The cocycle condition reads multiplicatively (with inverse instead of negative for antisymmetry: $g_{ij}=g_{ji}^{-1}$); both conditions are read off the algebraic definition of $g_{ij}$ as a ratio.

Step 2 — obstruction class in $\mathrm{Pic}(X)$. The cohomology class $[\{g_{ij}\}]\in {\check{H}}^1(\mathcal{U},{\mathcal{O}}_X^{\ast })$ refines through any common refinement and passes to ${\check{H}}^1(X,{\mathcal{O}}_X^{\ast })$. The Cartan-Leray comparison 06.04.02 identifies this with the derived-functor cohomology $H^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X)$, the Picard group of holomorphic line bundles. Sheaf-theoretically the class is the image of the Cousin II datum's divisor-sheaf section $\sigma \in \Gamma (X,{\mathcal{D}}_X)$ under the connecting map $\delta :\Gamma (X,{\mathcal{D}}_X)\to \mathrm{Pic}(X)$ in the long exact sequence attached to $0\to {\mathcal{O}}_X^{\ast }\to {\mathcal{M}}_X^{\ast }\to {\mathcal{D}}_X\to 0$; the Čech and connecting-map descriptions agree by the standard zigzag chase. The line bundle determined by $\{g_{ij}\}$ is the holomorphic bundle $L\to X$ with transition functions $g_{ij}$ on $U_i\cap U_j$, and the Cousin II datum is solvable exactly when $L$ is holomorphically isomorphic to ${\mathcal{O}}_X$.

Step 3 — exponential-sequence routing. The exponential sheaf sequence $0\to 2\pi i\mathbb{Z}\to {\mathcal{O}}_X\to {\mathcal{O}}_X^{\ast }\to 0$ produces the long exact sequence in cohomology with the segment

$$H^1(X,{\mathcal{O}}_X)\stackrel{{\exp }_{\ast }}{\to }\mathrm{Pic}(X)\stackrel{c_1}{\to }{H}^2(X,2\pi i\mathbb{Z})\to H^2(X,{\mathcal{O}}_X).$$

The Picard group sits between an analytic group on the left and a topological group on the right: a class in $\mathrm{Pic}(X)$ is killed iff its image in $H^2(X,\mathbb{Z})$ is zero (the topological obstruction vanishes) and its preimage in $H^1(X,{\mathcal{O}}_X)$ exists (the analytic obstruction also vanishes). The two conditions correspond to the two ingredients of a holomorphic line bundle: its underlying topological type and the holomorphic-trivialisation freedom within that type.

Step 4 — vanishing of both obstructions. The Behnke-Stein theorem 06.09.03 makes $X$ a Stein Riemann surface. Cartan's Theorem B 06.09.02 gives $H^q(X,\mathcal{F})=0$ for every coherent analytic sheaf $\mathcal{F}$ and every $q\ge 1$. Specialising $\mathcal{F}={\mathcal{O}}_X$ at $q=1$ produces $H^1(X,{\mathcal{O}}_X)=0$, the analytic vanishing. Independently, $X$ is a connected non-compact orientable real $2$-manifold, and the top-degree integer cohomology of a connected non-compact $n$-manifold vanishes: $H^n(X,\mathbb{Z})=0$. Specialising $n=2$ produces $H^2(X,\mathbb{Z})=0$, the topological vanishing. By the exponential long exact sequence, $\mathrm{Pic}(X)$ is sandwiched between two zero groups, so $\mathrm{Pic}(X)=0$.

Step 5 — global assembly. The cohomology class $[\{g_{ij}\}]\in \mathrm{Pic}(X)$ is therefore the zero class, and the cocycle $\{g_{ij}\}$ is a Čech coboundary: there exist non-vanishing holomorphic $h_i\in {\mathcal{O}}^{\ast }(U_i)$ with $g_{ij}=h_i/h_j$ on every overlap. Define $f$ on $X$ piecewise by $f{|}_{U_i}=f_i/h_i$. On every overlap $U_i\cap U_j$, $$ \frac{f_i / h_i}{f_j / h_j} = \frac{f_i / f_j}{h_i / h_j} = \frac{g_{ij}}{g_{ij}} = 1, $$ so the local definitions agree and $f$ is a well-defined function on $X$. On each $U_i$, $f{|}_{U_i}=f_i/h_i$ is a ratio of a non-zero meromorphic function and a non-vanishing holomorphic function, hence non-zero meromorphic on $U_i$, with the same divisor as $f_i$ (since $h_i$ is everywhere non-zero and holomorphic). The global $f$ is therefore in ${\mathcal{M}}^{\ast }(X)$ with $f/f_i=1/h_i\in {\mathcal{O}}^{\ast }(U_i)$ for every $i$, the Cousin II conclusion. $\square$

The five-step structure follows Forster §26 (multiplicative cocycle assembly + exponential sequence + Theorem B + topological vanishing + coboundary trivialisation) and dovetails with Hörmander Ch. III (the higher-dimensional Cousin II solution on a Stein manifold runs the same skeleton, with the topological vanishing replaced by the hypothesis that the underlying topological line bundle has zero first Chern class — the Oka-Grauert principle).

Bridge. The proof binds three pieces of upstream machinery into one statement. From 06.09.04 comes the additive Cousin I theorem: the analytic engine that drives the analytic vanishing $H^1(X,{\mathcal{O}}_X)=0$. From 06.09.02 comes Cartan's Theorem B: the cohomological vanishing principle on a Stein manifold. From the topology of non-compact surfaces comes $H^2(X,\mathbb{Z})=0$. The Cousin II theorem here is exactly the corollary of routing these three vanishings through the exponential sequence. The foundational reason Cousin II is unconditional on a non-compact Riemann surface is that its obstruction lies in a Picard group that is sandwiched between an analytic group (Stein-killed) and a topological group (dimension-killed), and the two-fold geometric hypothesis "non-compact + dimension one" is generous enough to kill both.

This builds toward the higher-dimensional Cartan-Serre and Oka-Grauert framework and appears again in 06.07.01 when the same skeleton runs on Stein manifolds in arbitrary complex dimension. The Cousin I theorem 06.09.04 is the additive sibling of the present theorem; together they form the classical existence-problem package on a Stein space. The classical Weierstrass product theorem on $\mathbb{C}$, with its convergence-improving elementary factors, is one explicit instantiation of the same chain on the simplest non-compact Riemann surface. The downstream pattern recurs: putting these together, every classical existence problem on a Stein space (Cousin I, Cousin II, Mittag-Leffler, Picard-group identity, Weierstrass product, Runge approximation) becomes a corollary of one or two cohomological vanishings, and Cousin II is the example that shows how both an analytic and a topological obstruction interact through a single exact sequence. The central insight is that the multiplicative problem is genuinely richer than the additive one — it sees topology — and the dimension-one Stein case is the only setting where both obstructions vanish for free.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib does not currently formalise the Cousin II problem, the meromorphic-units sheaf, the divisor quotient sheaf, or the exponential sheaf sequence on a Riemann surface. 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 meromorphic-units sheaf, the divisor quotient sheaf, the exponential sheaf sequence, the Picard group as a sheaf cohomology group, sheaf cohomology in the analytic setting on a manifold, Cartan's Theorem B, the Oka-Grauert principle). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

The Cousin II problem is the multiplicative sibling of Cousin I and the cleanest illustration of how analytic and topological obstructions interact on a complex manifold. A single short exact sequence of sheaves — the exponential sequence — sandwiches the Picard group between an analytic obstruction (cohomology of the structure sheaf) and a topological obstruction (integer second cohomology), and the Cousin II solvability question reduces to the joint vanishing of both. On a non-compact Riemann surface the two obstructions vanish for distinct reasons: Theorem B 06.09.02 kills the analytic part via Behnke-Stein 06.09.03, and the dimension of the underlying real manifold kills the topological part. This makes the dimension-one Stein case the unique setting where Cousin II is unconditional on geometric grounds alone.

The exponential sheaf sequence on a complex manifold. The cleanest packaging of Cousin II uses the exponential sequence

$$0\to 2\pi i\mathbb{Z}\to {\mathcal{O}}_X\stackrel{\exp }{\to }{\mathcal{O}}_X^{\ast }\to 0,$$

valid on every complex manifold $X$. The kernel is the constant sheaf $2\pi i\mathbb{Z}$ (locally any non-vanishing holomorphic function $h$ admits a holomorphic logarithm $g$ with $\exp (g)=h$, unique modulo $2\pi i\mathbb{Z}$ on a contractible patch); the surjectivity holds at the sheaf level because every point has a contractible neighbourhood on which logarithms exist. The associated long exact sequence in cohomology

$$\cdots \to H^1(X,2\pi i\mathbb{Z})\to H^1(X,{\mathcal{O}}_X)\stackrel{{\exp }_{\ast }}{\to }{H}^1(X,{\mathcal{O}}_X^{\ast })\stackrel{c_1}{\to }{H}^2(X,2\pi i\mathbb{Z})\to H^2(X,{\mathcal{O}}_X)\to \cdots$$

identifies the Picard group $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\ast })$ as fitting in an exact sequence between $H^1(X,{\mathcal{O}}_X)$ on the left (analytic) and $H^2(X,\mathbb{Z})$ on the right (topological). The connecting map $c_1$ is the first Chern class: it sends a holomorphic line bundle to the underlying topological line bundle, which is classified by $H^2(X,\mathbb{Z})$ via the obstruction theory of $U(1)$-principal bundles on a topological space.

Weierstrass product theorem on $\mathbb{C}$ as the planar case. The 1876 Weierstrass product theorem is the special case $X=\mathbb{C}$ of the Cousin II theorem. Karl Weierstrass proved ^{[Weierstrass 1876]} that for any sequence $a_n\to \infty$ in $\mathbb{C}$ and any choice of multiplicities $m_n\in {\mathbb{Z}}_{\ge 1}$, the convergence-improved product

$$f(z)=z^{m_0}\prod _{n\ge 1}{E}_{p_n}(z/a_n{)}^{m_n}$$

with elementary factors $E_p(z)=(1-z)\exp (z+z^2/2+\cdots +z^p/p)$ converges to an entire function on $\mathbb{C}$ with the prescribed zero divisor, where $p_n$ is chosen large enough that ${\sum }_n|a_n{|}^{-(p_n+1)}<\infty$. The elementary factors $E_{p_n}$ are the explicit Čech-coboundary trivialisation of the Cousin II cocycle on the cover $\{U_n,U_0\}$ with $U_n$ a small disc around $a_n$ and $U_0=\mathbb{C}\setminus \{a_n\}$. The 1876 theorem predates the Cousin II formulation (1895) by nineteen years and the cohomological language (1951–53) by seventy-five years; it is the empirical instance from which the abstraction generalises.

Higher-dimensional Cousin II — Cousin, Oka, Grauert. Cousin himself posed the problem in arbitrary complex dimension in 1895 ^{[Cousin 1895]} (Acta Math. 19, 1–61). On a polydisc in ${\mathbb{C}}^n$, Cousin solved Cousin II directly using the same iterated Cauchy integrals as for Cousin I, since a polydisc is contractible and $H^2({\Delta }^n,\mathbb{Z})=0$. On a general domain in ${\mathbb{C}}^n$, the Cousin II problem is governed by the exponential sequence and solvability is equivalent to the holomorphic line bundle determined by the cocycle being holomorphically isomorphic to the structure sheaf. Kiyoshi Oka's 1939 paper ^{[Oka 1939]} Sur les fonctions analytiques de plusieurs variables III (J. Sci. Hiroshima Univ. Ser. A, 9) gave the first counterexample: a domain in ${\mathbb{C}}^2$ with non-zero second integer cohomology where Cousin II fails for explicit divisor data, even though Cousin I succeeds. The same paper proved the Oka principle in its first form: on a Stein manifold, every continuous Cousin II datum has a holomorphic solution iff the topological obstruction vanishes. Hans Grauert's 1958 paper ^{[Grauert 1958]} Analytische Faserungen über holomorph-vollständigen Räumen (Math. Ann. 135) extended the principle to vector bundles of arbitrary rank: on a Stein manifold, the natural map from holomorphic to topological complex vector bundle isomorphism classes is a bijection. The Oka-Grauert principle is the modern unified statement and the reason Cousin II on a Stein manifold reduces entirely to topology.

Failure on non-Stein manifolds and on compact spaces. On a compact Riemann surface $X$ of genus $g\ge 1$, $H^2(X,\mathbb{Z})=\mathbb{Z}$ and the first Chern class map $c_1:\mathrm{Pic}(X)\to \mathbb{Z}$ is surjective (achieved by the line bundles $\mathcal{O}(np)$ for any base point $p$). The kernel ${\mathrm{Pic}}^0(X)$ is the Jacobian of $X$, a $g$-dimensional complex torus ${\mathbb{C}}^g/\Lambda$ where $\Lambda$ is the period lattice of the holomorphic differentials; ${\mathrm{Pic}}^0(X)$ is itself an abelian variety in algebraic-geometry language. On ${\mathbb{C}}^2\setminus \{0\}$ — non-Stein in dimension two — the punctured plane retracts onto $S^3$ whose $H^2=0$, so $H^2({\mathbb{C}}^2\setminus \{0\},\mathbb{Z})=0$ and the topological obstruction vanishes; but $H^1({\mathbb{C}}^2\setminus \{0\},\mathcal{O})\ne 0$ supplies an analytic obstruction, and Cousin II can fail on this space via the analytic part of the exponential sequence even though the topological part vanishes. The richer counterexamples in higher dimension involve compact Kähler manifolds with non-zero $H^2(X,\mathbb{Z})$, where the topological obstruction is the dominant failure mode.

Compact case — Picard variety and the Jacobian. On a compact Riemann surface $X$ of genus $g$, the Picard group sits in the exact sequence

$$0\to {\mathrm{Pic}}^0(X)\to \mathrm{Pic}(X)\stackrel{\mathrm{deg}}{\to }\mathbb{Z}\to 0,$$

with the degree map $\mathrm{deg}=c_1$ surjecting onto $\mathbb{Z}=H^2(X,\mathbb{Z})$, and the kernel ${\mathrm{Pic}}^0(X)\cong H^1(X,{\mathcal{O}}_X)/H^1(X,\mathbb{Z})\cong {\mathbb{C}}^g/\Lambda$ a $g$-dimensional complex torus. The Cousin II problem on a compact RS therefore has a Picard-group of obstructions, dimension-counted by the genus on the analytic side and rank-counted by $\mathbb{Z}$ on the topological side. A Cousin II datum is solvable iff its associated line bundle has degree zero (topological condition) and represents the identity element of ${\mathrm{Pic}}^0(X)$ (analytic condition); generically, both fail. The compact-versus-non-compact dichotomy on a Riemann surface, which Behnke-Stein exposes as the Stein-versus-not dichotomy, also exposes itself here as the obstructed-Cousin-II-versus-unobstructed dichotomy.

Cousin I and Cousin II compared. The Cousin I obstruction lives in $H^1(X,{\mathcal{O}}_X)$, an analytic group with no topological component. The Cousin II obstruction lives in $H^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X)$, sandwiched between $H^1(X,{\mathcal{O}}_X)$ (analytic) and $H^2(X,\mathbb{Z})$ (topological). On a non-compact RS, both Cousin I and Cousin II are unconditional, with vanishings from different sources: Cousin I needs only the Stein hypothesis (Behnke-Stein), while Cousin II additionally needs the topological vanishing $H^2(X,\mathbb{Z})=0$. In higher complex dimension, Cousin I is solvable on every Stein manifold (Cartan-Serre), but Cousin II can fail on a Stein manifold with non-zero $H^2(X,\mathbb{Z})$ — Cousin's original 1895 problem on a non-contractible Stein domain is exactly this failure mode. The Cousin II problem is therefore strictly richer than Cousin I in higher dimension, and the Oka-Grauert principle is the modern statement of how rich.

Approximation and density. The Cousin II theorem combines with the Cousin I theorem and Runge approximation to give a powerful structural result: on a non-compact Riemann surface $X$, every locally consistent zero-and-pole datum is realised by a global meromorphic function, with the Weierstrass-product / Mittag-Leffler / Cousin-coboundary machinery supplying explicit convergent constructions. The analytic structure of ${\mathcal{M}}^{\ast }(X)$ — its divisor map, its kernel, its surjective image in $\Gamma (X,{\mathcal{D}}_X)$ — is fully described by the two Cousin theorems and their cohomological packaging.

Synthesis. The Cousin II problem on a non-compact Riemann surface is the multiplicative half of the classical Stein-theory existence package: a divisor datum is realised by a global non-zero meromorphic function exactly when the holomorphic line bundle determined by its ratio cocycle is isomorphic to the structure sheaf, and on a non-compact RS this happens for every datum because the analytic and topological obstructions both vanish. This is exactly the principle Pierre Cousin formulated in 1895, the analogue of his additive problem from the same paper, and the bridge is short: multiplicative cocycle assembly, exponential sequence routing, Theorem B vanishing, topological vanishing, coboundary trivialisation, global lift. The 1876 Weierstrass product theorem on $\mathbb{C}$ is the planar case; the higher-dimensional Oka-Grauert principle generalises to every Stein manifold; the Hopf-bundle counterexample on ${\mathbb{C}}^2\setminus \{0\}$ traces the failure to the failure of the Stein hypothesis on the analytic side. Putting these together, the central insight is that Cousin II is the cleanest of the four classical existence problems on a Stein space at which both analytic and topological data interact, and the dimension-one Stein case is the one where both obstructions vanish for free. The pattern recurs in 06.07.01 when the same skeleton runs on Stein manifolds in arbitrary complex dimension and builds toward the full Oka-Grauert framework. The bridge is from the 1876 planar Weierstrass-product elementary factors to a uniform exponential-sequence vanishing principle that subsumes every classical multiplicative existence problem on a non-compact curve.

Full proof set [Master]

Lemma (multiplicative cocycle assembly). Given a Cousin II datum $(\mathcal{U},\{f_i\})$ on a Riemann surface $X$, the ratios $g_{ij}=f_i/f_j$ on overlaps $U_i\cap U_j$ form a Čech 1-cocycle in $\check Z^1(\mathcal{U}, \mathcal{O}_X^)$.*

Proof. Holomorphy and non-vanishing of $g_{ij}$ on $U_i\cap U_j$ is the hypothesis of the Cousin II datum. Multiplicative antisymmetry $g_{ij}=g_{ji}^{-1}$ is immediate from the algebraic definition. The cocycle condition on triple overlaps is the algebraic identity $(f_i/f_j)(f_j/f_k)(f_k/f_i)=1$, which holds pointwise. The collection $\{g_{ij}\}$ therefore lies in ${\check{Z}}^1(\mathcal{U},{\mathcal{O}}_X^{\ast })$ for the alternating Čech complex with values in the multiplicative-units sheaf. $\square$

Lemma (connecting-map identification for Cousin II). The Čech-cocycle class $[{g_{ij}}] \in \check H^1(X, \mathcal{O}_X^) = \mathrm{Pic}(X)$equals$\delta(\sigma)$,where$\sigma \in \Gamma(X, \mathcal{D}_X)$istheglobaldivisor-sheafsectiondeterminedbytheCousinIIdatumand$\delta$istheconnectinghomomorphisminthelongexactsequenceof$0 \to \mathcal{O}_X^* \to \mathcal{M}_X^* \to \mathcal{D}_X \to 0$.*

Proof. Standard zigzag in the snake lemma applied to the short exact sequence at the level of Čech cochains. Choose lifts: each $\sigma {|}_{U_i}\in \Gamma (U_i,{\mathcal{D}}_X)$ lifts to $f_i\in \Gamma (U_i,{\mathcal{M}}_X^{\ast })$ by the surjectivity of ${\mathcal{M}}_X^{\ast }\to {\mathcal{D}}_X$. Apply the multiplicative Čech differential: $(d\{f_i\}{)}_{ij}=f_j/f_i$ on $U_i\cap U_j$, which lies in $\Gamma (U_i\cap U_j,{\mathcal{O}}_X^{\ast })$ because $\sigma$ is a global section of ${\mathcal{D}}_X$. The connecting-map definition reads $\delta (\sigma )=[(d\{f_i\}{)}^{-1}]=[\{g_{ij}\}]$, with the inverse-versus-sign convention. The Cartan-Leray comparison 06.04.02 identifies Čech with derived-functor cohomology on the paracompact $X$. The line bundle $L\to X$ with transition functions $g_{ij}$ is the geometric realisation of the cohomology class. $\square$

Lemma (analytic vanishing). On a connected non-compact Riemann surface $X$, $H^1(X,{\mathcal{O}}_X)=0$.

Proof. By Behnke-Stein 06.09.03, $X$ is Stein. By Cartan's Theorem B 06.09.02 applied to the coherent structure sheaf ${\mathcal{O}}_X$, $H^q(X,{\mathcal{O}}_X)=0$ for every $q\ge 1$. Specialise $q=1$. $\square$

Lemma (topological vanishing). On a connected non-compact orientable real $2$-manifold $X$, $H^2(X,\mathbb{Z})=0$.

Proof. For a connected non-compact $n$-manifold $M$, the top-dimensional integer cohomology vanishes: $H^n(M,\mathbb{Z})=0$. This is a standard consequence of Poincaré-Lefschetz duality for non-compact manifolds, $H_c^n(M,\mathbb{Z})\cong H_0(M,\mathbb{Z})$ where compact-support cohomology gives the dual to homology, and the ordinary-vs-compact-support distinction shifts the dimension by one — concretely, every $n$-cocycle on a connected non-compact $n$-manifold has compact support pulled to a non-zero class only when the manifold is compact. Apply with $n=2$ and $M=X$. $\square$

Lemma (exponential-sequence vanishing of $\mathrm{Pic}(X)$). On a connected non-compact Riemann surface $X$, $\mathrm{Pic}(X) = H^1(X, \mathcal{O}_X^) = 0$.*

Proof. The exponential sheaf sequence $0\to 2\pi i\mathbb{Z}\to {\mathcal{O}}_X\to {\mathcal{O}}_X^{\ast }\to 0$ has long exact sequence segment

$$H^1(X,{\mathcal{O}}_X)\stackrel{{\exp }_{\ast }}{\to }{H}^1(X,{\mathcal{O}}_X^{\ast })\stackrel{c_1}{\to }{H}^2(X,2\pi i\mathbb{Z}).$$

The left group vanishes by the analytic-vanishing lemma. The right group vanishes by the topological-vanishing lemma (with the constant sheaf $2\pi i\mathbb{Z}$ identified with $\mathbb{Z}$ via the constant $2\pi i$). Exactness in the middle forces $H^1(X,{\mathcal{O}}_X^{\ast })=0$. $\square$

Theorem (Cousin II on a non-compact Riemann surface, full statement). Every Cousin II datum on a connected non-compact Riemann surface $X$ is solvable.

Proof. Combine the four lemmas. The Čech-cocycle $\{g_{ij}\}$ from Lemma 1 has cohomology class equal to the connecting-map image $\delta (\sigma )\in \mathrm{Pic}(X)$ by Lemma 2. By Lemma 5, this class is zero, so $\{g_{ij}\}$ is a multiplicative Čech coboundary: there exist non-vanishing holomorphic $h_i\in {\mathcal{O}}^{\ast }(U_i)$ with $g_{ij}=h_i/h_j$ on every overlap. Set $f{|}_{U_i}=f_i/h_i$. On overlaps, $$ \frac{f_i / h_i}{f_j / h_j} = \frac{g_{ij}}{h_i / h_j} = \frac{g_{ij}}{g_{ij}} = 1, $$ so $f$ is a well-defined function on $X$. On each $U_i$, $f{|}_{U_i}$ is a ratio of a non-zero meromorphic and a non-vanishing holomorphic function, hence non-zero meromorphic, with $f/f_i=1/h_i\in {\mathcal{O}}^{\ast }(U_i)$. The global $f\in {\mathcal{M}}^{\ast }(X)$ realises the Cousin II datum. $\square$

Corollary (Weierstrass product theorem on a non-compact Riemann surface). Let $X$ be a non-compact connected Riemann surface, $\{p_n\}$ a discrete (locally finite) subset, and $m_n\in \mathbb{Z}$ a prescribed multiplicity at each $p_n$. Then there exists $f \in \mathcal{M}^(X)$whosedivisorisexactly$\sum_n m_n \cdot p_n$,andwhichiseverywherenon-zeroandholomorphicon$X \setminus {p_n}$.*

Proof. Form the cover $\{U_n{\}}_{n\ge 0}$ where $U_0=X\setminus \{p_n\}$ and $U_n$ is a chart neighbourhood of $p_n$ chosen so that no two $U_n$ for $n\ge 1$ overlap. On each $U_n$ for $n\ge 1$, choose a local non-zero meromorphic function $f_n=z^{m_n}$ in a local coordinate $z$ at $p_n$, realising the prescribed multiplicity. On $U_0$, set $f_0=1$. The ratios $f_n/f_0=f_n$ on $U_n\cap U_0=U_n\setminus \{p_n\}$ are non-vanishing-holomorphic by construction; ratios across $U_n\cap U_m=\varnothing$ for $n,m\ge 1$ with $n\ne m$ are formally $1$. The Cousin II theorem gives a global $f\in {\mathcal{M}}^{\ast }(X)$ with $f/f_n\in {\mathcal{O}}^{\ast }(U_n)$ for every $n$, i.e. multiplicity $m_n$ at $p_n$ and everywhere non-zero holomorphic away from $\{p_n\}$. $\square$

Corollary (Weierstrass product on $\mathbb{C}$, classical form). For every sequence $a_n\to \infty$ in $\mathbb{C}$ and every choice of multiplicities $m_n\in {\mathbb{Z}}_{\ge 1}$, there exist integers $p_n\ge 0$ such that the product ${\prod }_n{E}_{p_n}(z/a_n{)}^{m_n}$ converges uniformly on compact subsets of $\mathbb{C}$ to an entire function with prescribed zero divisor ${\sum }_n{m}_n\cdot a_n$, where $E_p(z)=(1-z)\exp (z+z^2/2+\cdots +z^p/p)$ are the Weierstrass elementary factors.

Proof. The previous corollary applied to $X=\mathbb{C}$ with the discrete divisor ${\sum }_n{m}_n\cdot a_n$ produces $f\in {\mathcal{M}}^{\ast }(\mathbb{C})$ with the prescribed divisor and no other zeros or poles. The convergence-improving factors $\exp (z+z^2/2+\cdots +z^p/p)$ are the multiplicative Čech-coboundary terms $h_n$ from the proof of the Cousin II theorem, expressed concretely as exponentials of partial Taylor expansions of $-\log (1-z/a_n)$ chosen so that the resulting product converges uniformly on compact subsets — Weierstrass's 1876 explicit construction. The integer $p_n$ is chosen large enough that ${\sum }_n|a_n{|}^{-(p_n+1)}<\infty$, which guarantees convergence. $\square$

Corollary (Cousin II solvable on every Stein manifold with $H^2(X,\mathbb{Z})=0$). Let $X$ be a Stein complex manifold of arbitrary complex dimension with $H^2(X,\mathbb{Z})=0$. Every Cousin II datum on $X$ is solvable.

Proof. Cartan's Theorem B 06.09.02 on a Stein manifold gives $H^1(X,{\mathcal{O}}_X)=0$ in arbitrary dimension. The hypothesis gives $H^2(X,\mathbb{Z})=0$. The exponential-sequence routing of Lemma 5 produces $\mathrm{Pic}(X)=0$. The multiplicative cocycle assembly + connecting-map identification + coboundary lift argument is dimension-independent. $\square$

Corollary (Oka-Grauert principle for line bundles, statement). On a Stein manifold $X$, the natural map ${\mathrm{Pic}}_{\mathrm{hol}}(X)\to H^2(X,\mathbb{Z})$ given by the first Chern class is a bijection of abelian groups; equivalently, every continuous complex line bundle on $X$ admits a unique holomorphic structure up to holomorphic isomorphism.

Stated; for proof see Grauert 1958. The Cartan-Serre Theorem B provides $H^1(X,{\mathcal{O}}_X)=0$, so the exponential sequence segment $0=H^1(X,{\mathcal{O}}_X)\to \mathrm{Pic}(X)\to H^2(X,\mathbb{Z})\to H^2(X,{\mathcal{O}}_X)=0$ exhibits $c_1$ as an isomorphism. Grauert's 1958 paper ^{[Grauert 1958]} extended the principle to vector bundles of arbitrary rank using the same exponential / matrix-exponential technique. $\square$

Connections [Master]

• Cousin I additive 06.09.04. The additive sibling of the present unit. Cousin I has obstruction in $H^1(X,{\mathcal{O}}_X)$, an analytic group; Cousin II has obstruction in $H^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X)$, sandwiched between the analytic group and the topological group $H^2(X,\mathbb{Z})$. On a non-compact RS, both are unconditional, but the vanishing chains are different: Cousin I is killed by Behnke-Stein + Theorem B alone, while Cousin II additionally requires the topological dimension argument $H^2(X,\mathbb{Z})=0$. The two units form the classical Stein-theory existence-problem package on a non-compact Riemann surface.

• Cartan's Theorems A and B for Stein Riemann surfaces 06.09.02. Theorem B applied to the structure sheaf ${\mathcal{O}}_X$ at degree one is exactly the cohomological vanishing $H^1(X,{\mathcal{O}}_X)=0$ that makes the analytic part of the Cousin II obstruction vanish via the exponential sequence. Without the Theorem B step, $\mathrm{Pic}(X)$ would not be sandwiched purely by the topological group on the right, and Cousin II would not be solvable in general even on a topologically simple non-compact RS.

• Behnke-Stein theorem 06.09.03. The geometric input that supplies the Stein hypothesis on a non-compact Riemann surface, making Theorem B applicable. Without Behnke-Stein, the Cousin II theorem here would be conditional on the Stein property; with it, every non-compact Riemann surface satisfies the hypothesis automatically.

• Holomorphic line bundle on a Riemann surface 06.05.02. The geometric language for the Cousin II obstruction: a Cousin II datum's cocycle $\{g_{ij}\}$ defines a holomorphic line bundle, and the datum is solvable iff that line bundle is holomorphically isomorphic to ${\mathcal{O}}_X$. The unit 06.05.02 establishes the cocycle-to-bundle equivalence and the Picard group $\mathrm{Pic}(X)=H^1(X,{\mathcal{O}}_X^{\ast })$ as the classifying group, on which the Cousin II problem rests directly.

• Čech cohomology of holomorphic line bundles 06.04.02. The cohomological language in which the Cousin II obstruction is stated: ${\check{H}}^1(X,{\mathcal{O}}_X^{\ast })=\mathrm{Pic}(X)$ via the Cartan-Leray comparison on a paracompact $X$. The non-compact Stein case here makes this group vanish; on a compact RS of genus $g$ the same group has a $\mathbb{Z}$-valued degree map onto $H^2(X,\mathbb{Z})$ and a $g$-dimensional kernel ${\mathrm{Pic}}^0(X)$.

• Meromorphic function 06.01.05. The classical Weierstrass product theorem on $\mathbb{C}$ is the planar case of the Cousin II corollary on a non-compact Riemann surface. The unit 06.01.05 introduces meromorphic functions and their divisor; the Cousin II theorem promotes the local divisor data to a global existence statement on every non-compact RS.

• Hartogs phenomenon 06.07.02. The failure mode of the Stein hypothesis in higher dimension, which feeds counterexamples to Cousin II on non-Stein manifolds. The Hartogs-extension counterexample on ${\mathbb{C}}^2\setminus \{0\}$ supplies a non-zero $H^1(X,{\mathcal{O}}_X)$, the analytic obstruction part of the Cousin II exponential-sequence routing.

• Holomorphic functions of several complex variables 06.07.01. The higher-dimensional analogue of Cousin II on Stein manifolds — Oka 1939, Grauert 1958 — generalises the dimension-one statement here. The unit 06.07.01 supplies the multivariable analytic framework on which the Oka-Grauert principle runs; the Cousin II theorem here is the dimension-one specialisation where the topological obstruction vanishes for free.

• Riemann-Roch theorem for compact Riemann surfaces 06.04.01. On a compact RS of genus $g$, the Cousin II obstruction $\mathrm{Pic}(X)$ has a $\mathbb{Z}$-valued degree map onto $H^2(X,\mathbb{Z})=\mathbb{Z}$ and a $g$-dimensional kernel ${\mathrm{Pic}}^0(X)$; the Riemann-Roch identity $\dim H^0(L)-\dim H^1(L)=\mathrm{deg}L+1-g$ records the failure mode quantitatively. The dichotomy compact-vs-non-compact on a Riemann surface — which Behnke-Stein exposes as Stein-vs-not — appears here as obstructed-Cousin-II-vs-unobstructed.

• Cartan-Serre Theorem B in higher dimension 06.07.01. The cohomological vanishing $H^1(X,{\mathcal{O}}_X)=0$ on a Stein manifold of arbitrary dimension is the content of Theorem B, and the Cousin II problem on such a manifold reduces (via the exponential sequence) to the topological obstruction in $H^2(X,\mathbb{Z})$. The Oka-Grauert principle is the modern packaging of this reduction.

Historical & philosophical context [Master]

Pierre Cousin posed the multiplicative Cousin II problem and the additive Cousin I problem in his 1895 doctoral thesis Sur les fonctions de $n$ variables complexes ^{[Cousin 1895]} (Acta Math. 19, 1–61). The thesis worked in ${\mathbb{C}}^n$ on bidiscs and polydiscs, with iterated Cauchy integrals as the analytic tool. Cousin solved the polydisc case for both problems directly: on a polydisc ${\Delta }^n\subset {\mathbb{C}}^n$, contractibility forces $H^2({\Delta }^n,\mathbb{Z})=0$, and the iterated-Cauchy-integral construction handles the analytic side. The solution on a general domain in ${\mathbb{C}}^n$ remained open for both problems until the work of Oka in 1937 (Cousin I) and 1939 (Cousin II). The 1895 paper explicitly distinguished the two problems and noted that they behave differently in higher dimension, anticipating by half a century the cohomological reformulation that would explain the difference.

Karl Weierstrass had proved the planar special case of Cousin II in 1876 in Zur Theorie der eindeutigen analytischen Functionen ^{[Weierstrass 1876]} (Berlin Akademie). Weierstrass's construction is the convergence-improved product: for a sequence $a_n\to \infty$ in $\mathbb{C}$ and prescribed multiplicities $m_n$, the product ${\prod }_n{E}_{p_n}(z/a_n{)}^{m_n}$ with elementary factors $E_p(z)=(1-z)\exp (z+z^2/2+\cdots +z^p/p)$ converges to an entire function with the prescribed zero divisor when $p_n$ is chosen large enough. The 1876 theorem predates the Cousin formulation by nineteen years and the cohomological reframing by seventy-five years; it is the empirical instance from which the abstraction generalises.

Kiyoshi Oka's 1939 paper Sur les fonctions analytiques de plusieurs variables III: Deuxième problème de Cousin ^{[Oka 1939]} (J. Sci. Hiroshima Univ. Ser. A, 9, 7–19) gave the first counterexample to Cousin II on a domain in ${\mathbb{C}}^2$ with non-zero second integer cohomology, showing that Cousin II is genuinely topological in higher dimension whereas Cousin I is purely analytic. The same paper formulated the Oka principle in its first form: on a Stein manifold, every continuous Cousin II datum has a holomorphic solution iff the topological obstruction vanishes. Hans Grauert's 1958 paper Analytische Faserungen über holomorph-vollständigen Räumen ^{[Grauert 1958]} (Math. Ann. 135, 263–273) extended the principle to complex vector bundles of arbitrary rank: on a Stein manifold, the natural map from holomorphic to topological complex vector bundle isomorphism classes is a bijection. The Oka-Grauert principle is the modern statement, and it makes Cousin II on a Stein manifold reduce entirely to topology — solvability is equivalent to the underlying topological line bundle having zero first Chern class.

Henri Cartan and Jean-Pierre Serre's 1951–53 séminaire at the École Normale Supérieure ^{[Cartan 1951–53]} reformulated both Cousin problems in cohomological language: Cousin I as $H^1(X,{\mathcal{O}}_X)=0$ via the principal-parts sheaf, Cousin II as $H^1(X,{\mathcal{O}}_X^{\ast })=0$ via the divisor sheaf and the exponential sequence. The séminaire framework made explicit the bridge between the two problems through the exponential sheaf sequence — the same Cartan-Serre framework that earlier produced Theorem B for coherent analytic sheaves now produced Theorem B for the multiplicative units sheaf via the long exact sequence and the topological obstruction in $H^2(X,\mathbb{Z})$. Behnke and Stein's 1949 paper Entwicklung analytischer Funktionen auf Riemannschen Flächen ^{[Behnke-Stein 1949]} (Math. Ann. 120, 430–461) established that every non-compact Riemann surface is Stein; combined with the topological vanishing $H^2(X,\mathbb{Z})=0$ on a non-compact connected real $2$-manifold, this made the dimension-one Cousin II problem unconditional on geometric grounds.

Lars Hörmander's 1973 monograph An Introduction to Complex Analysis in Several Variables ^{[Hörmander HSCV]} (North-Holland, Ch. III) presents the textbook treatment, with Cousin II as a corollary of the exponential sequence and the $L^2$-existence theorem on a Stein manifold. Hans Grauert and Reinhold Remmert's Theory of Stein Spaces ^{[Grauert-Remmert]} (Springer Grundlehren 236, 1979) is the canonical reference for Cousin theory on Stein analytic spaces, extending the manifold framework to allow analytic singularities and recording the full Oka-Grauert principle for vector bundles.

Bibliography [Master]

[object Promise]