[proofplan] We establish a bijective correspondence between solvability of Cousin I problems on $\Omega$ and the vanishing of the Dolbeault cohomology group $H^{0,1}_{\bar\partial}(\Omega)$. In the forward direction, given a Cousin I datum $\{(U_i, f_i)\}$, we construct a globally defined smooth $(0,1)$-form $\alpha$ via a [partition of unity](/page/Partition%20of%20Unity), show it is $\bar\partial$-closed, and prove that $\bar\partial$-exactness of $\alpha$ is equivalent to solvability of the Cousin I problem. In the converse direction, given any $\bar\partial$-closed $(0,1)$-form $\alpha$, we produce a Cousin I datum whose solvability forces $[\alpha] = 0$ in $H^{0,1}_{\bar\partial}(\Omega)$. [/proofplan]

[step:Reduce a Cousin I datum to a $\bar\partial$-closed $(0,1)$-form via a partition of unity]Let $\{(U_i, f_i)\}_{i \in I}$ be a Cousin I datum on $\Omega$: that is, $\{U_i\}$ is an open cover of $\Omega$ and each $f_i \in \mathcal{M}(U_i)$ is meromorphic on $U_i$, with the transition functions \begin{align*} g_{ij} := f_i - f_j \in \mathcal{O}(U_i \cap U_j) \end{align*} holomorphic on each non-empty overlap $U_i \cap U_j$. Choose a smooth [partition of unity](/page/Partition%20of%20Unity) $\{\chi_i\}_{i \in I}$ subordinate to $\{U_i\}$, so that $\chi_i \in C_c^\infty(U_i)$, $\chi_i \geq 0$, and $\sum_i \chi_i = 1$ on $\Omega$. Define the smooth function \begin{align*} \tilde{f}: \Omega \setminus \bigcup_i \operatorname{Poles}(f_i) &\to \mathbb{C}, \\ z &\mapsto \sum_{i \in I} \chi_i(z)\, f_i(z). \end{align*} On each $U_k$, the difference $\tilde{f} - f_k$ extends to a smooth function on all of $U_k$: \begin{align*} \tilde{f}(z) - f_k(z) = \sum_{i \in I} \chi_i(z)\,(f_i(z) - f_k(z)) = \sum_{i \in I} \chi_i(z)\, g_{ik}(z), \end{align*} which is smooth on $U_k$ because each $g_{ik} = f_i - f_k$ is holomorphic (hence smooth) on $U_i \cap U_k$, and $\chi_i$ is supported in $U_i$. Define the $(0,1)$-form \begin{align*} \alpha := \bar\partial \tilde{f}, \end{align*} interpreted as follows: on each $U_k$, we have $\alpha|_{U_k} = \bar\partial(\tilde{f} - f_k) + \bar\partial f_k = \bar\partial(\tilde{f} - f_k)$, since $f_k$ is meromorphic (holomorphic away from its poles, and $\bar\partial f_k = 0$ there). Since $\tilde{f} - f_k$ is smooth on all of $U_k$, the form $\alpha$ is a well-defined smooth $(0,1)$-form on $\Omega$. On any overlap $U_k \cap U_\ell$, the two local expressions agree: \begin{align*} \bar\partial(\tilde{f} - f_k) - \bar\partial(\tilde{f} - f_\ell) = \bar\partial(f_\ell - f_k) = \bar\partial g_{\ell k} = 0, \end{align*} since $g_{\ell k}$ is holomorphic. The form $\alpha$ is $\bar\partial$-closed because $\bar\partial\alpha = \bar\partial^2 \tilde{f} = 0$ (by the [$\bar\partial^2 = 0$ identity](/theorems/3409)).[/step]

[guided]The idea is to "average" the meromorphic functions $f_i$ into a single smooth function $\tilde{f}$ using a [partition of unity](/page/Partition%20of%20Unity). The cost is that $\tilde{f}$ is no longer holomorphic -- it is only smooth. The failure of holomorphicity is measured by the $(0,1)$-form $\alpha = \bar\partial\tilde{f}$. Why is $\alpha$ globally well-defined? On each $U_k$, the [meromorphic function](/page/Meromorphic%20Function) $f_k$ satisfies $\bar\partial f_k = 0$ (away from its poles, and the poles of all $f_i$ that meet $U_k$ cancel in $\tilde{f} - f_k$ because the transition functions $g_{ik} = f_i - f_k$ are holomorphic). So we can write $\alpha|_{U_k} = \bar\partial(\tilde{f} - f_k)$, and since $\tilde{f} - f_k = \sum_i \chi_i g_{ik}$ is smooth on all of $U_k$, there are no poles to worry about. Why is $\alpha$ $\bar\partial$-closed? The identity $\bar\partial^2 = 0$ (proved in the [$\bar\partial^2 = 0$ theorem](/theorems/3409)) applies: $\bar\partial\alpha = \bar\partial(\bar\partial\tilde{f}) = \bar\partial^2\tilde{f} = 0$. This computation is valid on each $U_k$ because the expression $\bar\partial(\tilde{f} - f_k)$ involves only smooth functions.[/guided]

[step:Show that $\bar\partial$-exactness of $\alpha$ is equivalent to solvability of the Cousin I datum]Suppose $\alpha = \bar\partial u$ for some $u \in C^\infty(\Omega)$. Define \begin{align*} f := \tilde{f} - u. \end{align*} On each $U_k$, we compute \begin{align*} \bar\partial(f - f_k) = \bar\partial(\tilde{f} - u - f_k) = \bar\partial(\tilde{f} - f_k) - \bar\partial u = \alpha - \alpha = 0. \end{align*} Since $f - f_k$ is smooth on $U_k$ and satisfies $\bar\partial(f - f_k) = 0$, it is holomorphic on $U_k$. Therefore $f$ is meromorphic on $\Omega$ with $f - f_i \in \mathcal{O}(U_i)$ for every $i$, which is the definition of a solution to the Cousin I problem. Conversely, suppose the Cousin I problem is solvable: there exists a meromorphic $f$ on $\Omega$ with $h_i := f - f_i \in \mathcal{O}(U_i)$ for all $i$. Define $u := \tilde{f} - f$, which is smooth on $\Omega$ (since on each $U_k$, $u = (\tilde{f} - f_k) - (f - f_k) = (\tilde{f} - f_k) - h_k$, a difference of smooth functions). Then $\bar\partial u = \bar\partial\tilde{f} - \bar\partial f = \alpha - 0 = \alpha$, so $\alpha$ is $\bar\partial$-exact.[/step]

[guided]The forward direction is the key construction: if we can solve $\bar\partial u = \alpha$, then $f = \tilde{f} - u$ "corrects" the smooth average $\tilde{f}$ back to a [meromorphic function](/page/Meromorphic%20Function). On each $U_k$, the correction $u$ absorbs exactly the non-holomorphic part of $\tilde{f}$, so that $f - f_k = (\tilde{f} - f_k) - u$ is holomorphic. The converse is the consistency check: if the Cousin I problem already has a solution $f$, then $u = \tilde{f} - f$ is the smooth function whose $\bar\partial$ equals $\alpha$. The equivalence can be summarised as: the Cousin I datum $\{(U_i, f_i)\}$ is solvable if and only if the cohomology class $[\alpha] \in H^{0,1}_{\bar\partial}(\Omega)$ vanishes.[/guided]

[step:Prove the forward implication: $H^{0,1}_{\bar\partial}(\Omega) = 0$ implies every Cousin I datum is solvable] Suppose $H^{0,1}_{\bar\partial}(\Omega) = 0$. Let $\{(U_i, f_i)\}$ be any Cousin I datum on $\Omega$. By the first step, the associated $(0,1)$-form $\alpha$ is $\bar\partial$-closed. Since $H^{0,1}_{\bar\partial}(\Omega) = 0$, every $\bar\partial$-closed $(0,1)$-form is $\bar\partial$-exact, so there exists $u \in C^\infty(\Omega)$ with $\bar\partial u = \alpha$. By the second step, the Cousin I datum is solvable. [/step]

[step:Prove the converse: universal Cousin I solvability implies $H^{0,1}_{\bar\partial}(\Omega) = 0$]Suppose every Cousin I datum on $\Omega$ is solvable. Let $\alpha$ be any $\bar\partial$-closed $(0,1)$-form on $\Omega$. We must show $[\alpha] = 0$ in $H^{0,1}_{\bar\partial}(\Omega)$. Choose a cover $\{U_i\}_{i \in I}$ of $\Omega$ by open polydiscs. On each polydisc $U_i$, the [$\bar\partial$-Poincare Lemma](/theorems/3410) guarantees the existence of $h_i \in C^\infty(U_i)$ with $\bar\partial h_i = \alpha|_{U_i}$: the lemma applies because $U_i$ is a polydisc and $\alpha|_{U_i}$ is a $\bar\partial$-closed $(0,1)$-form on $U_i$. On each non-empty overlap $U_i \cap U_j$, define \begin{align*} g_{ij} := h_i - h_j: U_i \cap U_j \to \mathbb{C}. \end{align*} This function is holomorphic: $\bar\partial g_{ij} = \bar\partial h_i - \bar\partial h_j = \alpha - \alpha = 0$ on $U_i \cap U_j$. The collection $\{(U_i, h_i)\}$ forms a Cousin I datum with transition functions $g_{ij}$. We verify the cocycle condition: on $U_i \cap U_j \cap U_k$, \begin{align*} g_{ij} + g_{jk} = (h_i - h_j) + (h_j - h_k) = h_i - h_k = g_{ik}, \end{align*} confirming that $\{g_{ij}\}$ is a valid Cousin I datum. By assumption, this Cousin I problem is solvable: there exists a [meromorphic function](/page/Meromorphic%20Function) $f$ on $\Omega$ (in fact holomorphic, since all the $h_i$ are smooth) with $f - h_i \in \mathcal{O}(U_i)$ for every $i$. Since $h_i$ is already smooth and $f - h_i$ is holomorphic, $f$ is smooth on $\Omega$. Setting $u := f$, we have on each $U_i$: \begin{align*} \bar\partial u = \bar\partial f = \bar\partial h_i + \bar\partial(f - h_i) = \alpha|_{U_i} + 0 = \alpha|_{U_i}. \end{align*} Since this holds on every $U_i$ and $\{U_i\}$ covers $\Omega$, we conclude $\bar\partial u = \alpha$ on all of $\Omega$. Therefore $[\alpha] = 0$ in $H^{0,1}_{\bar\partial}(\Omega)$.[/step]

[guided]The strategy is to reverse-engineer a Cousin I datum from an arbitrary $\bar\partial$-closed $(0,1)$-form $\alpha$, then appeal to the hypothesis that every Cousin I datum on $\Omega$ is solvable. The key tool is the $\bar\partial$-Poincare Lemma, which provides local solutions on polydiscs. Suppose every Cousin I datum on $\Omega$ is solvable. Let $\alpha$ be any $\bar\partial$-closed $(0,1)$-form on $\Omega$. We must show $[\alpha] = 0$ in $H^{0,1}_{\bar\partial}(\Omega)$, i.e., we must find $u \in C^\infty(\Omega)$ with $\bar\partial u = \alpha$. Choose a cover $\{U_i\}_{i \in I}$ of $\Omega$ by open polydiscs. Why polydiscs specifically? Because the [$\bar\partial$-Poincare Lemma](/theorems/3410) requires a domain on which $\bar\partial$-closed forms are automatically $\bar\partial$-exact. Polydiscs satisfy this by the explicit Cauchy-Green integral construction in the $\bar\partial$-Poincare Lemma. Any cover by Stein domains would also work, but polydiscs are the most concrete choice. On each polydisc $U_i$, the $\bar\partial$-Poincare Lemma guarantees the existence of $h_i \in C^\infty(U_i)$ with $\bar\partial h_i = \alpha|_{U_i}$. The lemma applies because $U_i$ is a polydisc and $\alpha|_{U_i}$ is a $\bar\partial$-closed $(0,1)$-form on $U_i$. Now we build the Cousin I datum from the local solutions. On each non-empty overlap $U_i \cap U_j$, define \begin{align*} g_{ij} := h_i - h_j: U_i \cap U_j \to \mathbb{C}. \end{align*} Why is $g_{ij}$ holomorphic? Both $h_i$ and $h_j$ solve $\bar\partial h = \alpha$ on $U_i \cap U_j$, so their difference satisfies $\bar\partial g_{ij} = \bar\partial h_i - \bar\partial h_j = \alpha - \alpha = 0$ on $U_i \cap U_j$, which means $g_{ij}$ is holomorphic. The collection $\{(U_i, h_i)\}$ thus forms a Cousin I datum with holomorphic transition functions $g_{ij}$. We verify the cocycle condition: on $U_i \cap U_j \cap U_k$, \begin{align*} g_{ij} + g_{jk} = (h_i - h_j) + (h_j - h_k) = h_i - h_k = g_{ik}, \end{align*} confirming that $\{g_{ij}\}$ is a valid Cousin I datum. This is the step that converts the analytic problem (is $\alpha$ exact?) into the algebraic problem (is the Cousin I datum $\{g_{ij}\}$ solvable?). By assumption, this Cousin I problem is solvable: there exists a [meromorphic function](/page/Meromorphic%20Function) $f$ on $\Omega$ with $f - h_i \in \mathcal{O}(U_i)$ for every $i$. Note that the resulting Cousin I datum involves only smooth (not meromorphic) functions $h_i$, so the solution $f$ is actually a smooth function, not merely meromorphic. To see this: since $h_i$ is smooth and $f - h_i$ is holomorphic (hence smooth), $f = h_i + (f - h_i)$ is smooth on each $U_i$, and since $\{U_i\}$ covers $\Omega$, $f$ is smooth on $\Omega$. Setting $u := f$, we compute on each $U_i$: \begin{align*} \bar\partial u = \bar\partial f = \bar\partial h_i + \bar\partial(f - h_i) = \alpha|_{U_i} + 0 = \alpha|_{U_i}. \end{align*} The first equality is by definition; the second uses $\bar\partial h_i = \alpha|_{U_i}$ (from the local solving step) and $\bar\partial(f - h_i) = 0$ (since $f - h_i$ is holomorphic on $U_i$). Since this holds on every $U_i$ and $\{U_i\}$ covers $\Omega$, we conclude $\bar\partial u = \alpha$ on all of $\Omega$. Therefore $[\alpha] = 0$ in $H^{0,1}_{\bar\partial}(\Omega)$.[/guided]

[step:Conclude the bijective correspondence] Combining the forward and converse directions: every Cousin I datum on $\Omega$ is solvable if and only if $H^{0,1}_{\bar\partial}(\Omega) = 0$. More precisely, the construction establishes a bijection between the set of Cousin I cohomology classes (equivalence classes of Cousin I data modulo solvable data) and the Dolbeault cohomology group $H^{0,1}_{\bar\partial}(\Omega)$. In the forward direction, a Cousin I datum maps to the class $[\alpha]$ of the associated $\bar\partial$-closed $(0,1)$-form; in the converse direction, a class $[\alpha]$ maps to the Cousin I datum obtained by local $\bar\partial$-solving on polydiscs. The datum is solvable if and only if $[\alpha] = 0$. [/step]