# The Dolbeault complex

*Written by Claude Opus 5.5 (Anthropic), October 2026. Self-checked by the writing AI. Public domain (CC0).*

On a complex manifold, differential forms split by type, and the exterior derivative splits as \(d=\partial+\bar\partial\). The \(\bar\partial\)-operator is the Cauchy–Riemann operator in all degrees: a function is holomorphic exactly when \(\bar\partial f=0\). This lesson proves that \(\bar\partial\)-closed forms are locally \(\bar\partial\)-exact (the Dolbeault–Grothendieck lemma), so that smooth forms of type \((p,\bullet)\) form a resolution of the sheaf of holomorphic \(p\)-forms by sheaves without higher cohomology. The resulting Dolbeault isomorphism computes the cohomology of holomorphic vector bundles by solving \(\bar\partial\)-equations, and it is the bridge to the \(L^2\) methods of the next lessons.

We use Holomorphic functions of several variables and the Cauchy–Pompeiu formula of one variable [Complex Analysis, Theorem 4.6.3] from the core course [Complex Analysis](https://kokunoyumeto.github.io/program-matematika-indonesia/en/#course-C50). Smooth partitions of unity are [Local tools for bundles and transport, Theorem 3.1](course:DG-FND/local-tools-for-bundles-and-transport#3-smooth-weights-with-controlled-support). The homological facts used are: cohomology of sheaves of modules on a ringed space and Čech cohomology [Cohomology of sheaves on ringed spaces](course:AG-QC/cohomology-of-sheaves-on-ringed-spaces); a sheaf whose Čech cohomology vanishes for every open covering of every open set has no higher cohomology on any open set [Stacks, Tag 01EV](https://kokunoyumeto.github.io/stacks-zh-hans-cn/en/cohomology.html#cohomology-lemma-cech-vanish); and a resolution by acyclic sheaves computes cohomology [Stacks, Tag 015E](https://kokunoyumeto.github.io/stacks-zh-hans-cn/en/derived.html#derived-lemma-leray-acyclicity).

Basic references are [Demailly] and [Lebl SCV].

## 1. Forms of type \((p,q)\)

Let \(X\) be a complex manifold of dimension \(n\). In local holomorphic coordinates \(z_j=x_j+iy_j\), the complex-valued \(1\)-forms are spanned by \(dz_j=dx_j+i\,dy_j\) and \(d\bar z_j=dx_j-i\,dy_j\). A smooth complex-valued \(k\)-form is uniquely a sum of forms of **type \((p,q)\)**, \(p+q=k\),

\[
f=\sum_{|I|=p,\ |J|=q}f_{I,J}\,dz_I\wedge d\bar z_J ,
\tag{1.1}
\]

with increasing multi-indices \(I,J\). The type decomposition does not depend on the coordinates, because a holomorphic change of coordinates expresses the new \(dz_j\) in terms of the old \(dz_k\) only. Write \(\mathcal E^{p,q}\) for the sheaf of smooth \((p,q)\)-forms. For a function, \(df=\partial f+\bar\partial f\) with \(\partial f=\sum(\partial f/\partial z_j)\,dz_j\) and \(\bar\partial f=\sum(\partial f/\partial\bar z_j)\,d\bar z_j\). On forms, \(\bar\partial\) acts on coefficients:

\[
\bar\partial\Bigl(\sum f_{I,J}\,dz_I\wedge d\bar z_J\Bigr)=\sum_{I,J}\sum_{l=1}^n\frac{\partial f_{I,J}}{\partial\bar z_l}\,d\bar z_l\wedge dz_I\wedge d\bar z_J ,
\tag{1.2}
\]

a map \(\mathcal E^{p,q}\to\mathcal E^{p,q+1}\). Since the operators \(\partial/\partial\bar z_l\) commute, \(\bar\partial\circ\bar\partial=0\); and \(\bar\partial(f\wedge g)=\bar\partial f\wedge g+(-1)^{\deg f}f\wedge\bar\partial g\). By Holomorphic functions of several variables, Proposition 2.2, a \((p,0)\)-form satisfies \(\bar\partial f=0\) exactly when its coefficients are holomorphic; such forms are the **holomorphic \(p\)-forms**, and they form the sheaf \(\Omega^p\), with \(\Omega^0=\mathcal O\).

Let \(E\) be a holomorphic vector bundle of rank \(r\) on \(X\), with sheaf of holomorphic sections \(\mathcal O(E)\). Smooth \((p,q)\)-forms with values in \(E\) form a sheaf \(\mathcal E^{p,q}(E)\). In a local holomorphic frame \(e_1,\ldots,e_r\), such a form is \(\sum_\nu f_\nu\otimes e_\nu\) with \(f_\nu\in\mathcal E^{p,q}\), and \(\bar\partial_E(\sum f_\nu\otimes e_\nu)=\sum\bar\partial f_\nu\otimes e_\nu\). This is independent of the frame because the transition matrices between holomorphic frames are holomorphic, hence annihilated by \(\bar\partial\). Again \(\bar\partial_E^2=0\), and \(\Omega^p(E)=\ker(\bar\partial_E:\mathcal E^{p,0}(E)\to\mathcal E^{p,1}(E))\).

## 2. The Dolbeault–Grothendieck lemma

**Lemma 2.1 (the \(\bar\partial\)-equation in one variable).** Let \(W\subset\mathbf R^m\) be open and let \(g(\zeta,w)\) be a smooth function on \(\mathbf C\times W\) whose support meets each set \(\mathbf C\times K\), \(K\subset W\) compact, in a compact set. Put

\[
u(z,w)=-\frac1\pi\int_{\mathbf C}\frac{g(\zeta,w)}{\zeta-z}\,dA(\zeta).
\tag{2.1}
\]

Then \(u\) is smooth and \(\partial u/\partial\bar z=g\). If \(g\) depends holomorphically on further complex parameters, so does \(u\).

**Proof.** Substituting \(\zeta=z+\eta\), \(u(z,w)=-\frac1\pi\int g(z+\eta,w)\,\eta^{-1}\,dA(\eta)\). The kernel \(1/\eta\) is integrable on bounded sets and \(g\) is smooth with locally uniformly compact support, so derivatives of every order in \((z,w)\) may be taken under the integral sign; \(u\) is smooth, and holomorphic in any further parameters in which \(g\) is. Moreover

\[
\frac{\partial u}{\partial\bar z}(z,w)=-\frac1\pi\int\frac{\partial g}{\partial\bar\zeta}(z+\eta,w)\,\frac{dA(\eta)}{\eta}=-\frac1\pi\int\frac{\partial g/\partial\bar\zeta\,(\zeta,w)}{\zeta-z}\,dA(\zeta)=g(z,w),
\]

the last equality by the Cauchy–Pompeiu formula [Complex Analysis, Theorem 4.6.3] applied to \(g(\cdot,w)\) on a disc containing its support, where the boundary integral vanishes. \(\square\)

For radii \(R=(R_1,\ldots,R_n)\) with \(0<R_j\leq\infty\) let \(\Delta(R)=\{|z_j|<R_j\}\).

**Theorem 2.2 (Dolbeault–Grothendieck lemma).** Let \(f\) be a smooth \(\bar\partial\)-closed \((p,q)\)-form on \(\Delta(R)\), with \(q\geq1\). For every \(r=(r_1,\ldots,r_n)\) with \(r_j<R_j\) there is a smooth \((p,q-1)\)-form \(u\) on \(\Delta(r)\) with \(\bar\partial u=f\) there.

**Proof.** Since \(\bar\partial\) acts on the coefficients of \(dz_I\), it suffices to treat \(p=0\). We prove by induction on \(k\) the statement: *if a \(\bar\partial\)-closed \((0,q)\)-form \(f\) on \(\Delta(R)\), \(q\geq1\), involves only \(d\bar z_1,\ldots,d\bar z_k\), then for all \(r<R\) it is \(\bar\partial\)-exact on \(\Delta(r)\).* For \(k=0\) such a form is zero. Let \(k\geq1\) and write \(f=d\bar z_k\wedge g+h\), where \(g\) and \(h\) involve only \(d\bar z_1,\ldots,d\bar z_{k-1}\). For \(l>k\), the coefficient of \(d\bar z_l\wedge d\bar z_k\wedge d\bar z_J\) in \(\bar\partial f\) is \(\partial g_J/\partial\bar z_l\), and similarly for \(h\); since \(\bar\partial f=0\), the coefficients of \(g\) and \(h\) are holomorphic in \(z_{k+1},\ldots,z_n\).

Choose \(r'\) with \(r<r'<R\) and a smooth function \(\psi\) of \(z_k\) alone, equal to \(1\) for \(|z_k|\leq r'_k\) and with compact support in \(|z_k|<R_k\). For each coefficient \(g_J\) put

\[
G_J(z)=-\frac1\pi\int_{\mathbf C}\frac{\psi(\zeta)\,g_J(z_1,\ldots,z_{k-1},\zeta,z_{k+1},\ldots,z_n)}{\zeta-z_k}\,dA(\zeta),
\qquad G=\sum_JG_J\,d\bar z_J .
\]

By Lemma 2.1, with \(z_k\) as the variable and the other coordinates as parameters, \(G\) is smooth on \(\Delta(R)\) with \(|z_j|<R_j\) for \(j\neq k\) (and \(z_k\in\mathbf C\)), holomorphic in \(z_{k+1},\ldots,z_n\), and \(\partial G_J/\partial\bar z_k=\psi g_J=g_J\) where \(|z_k|<r'_k\). Hence on \(\Delta(r')\),

\[
\bar\partial G=\sum_J\sum_{l\leq k}\frac{\partial G_J}{\partial\bar z_l}\,d\bar z_l\wedge d\bar z_J=d\bar z_k\wedge g+(\text{terms involving only }d\bar z_1,\ldots,d\bar z_{k-1}),
\]

so \(f-\bar\partial G\) is \(\bar\partial\)-closed on \(\Delta(r')\) and involves only \(d\bar z_1,\ldots,d\bar z_{k-1}\). By the induction hypothesis, applied on \(\Delta(r')\), there is \(v\) on \(\Delta(r)\) with \(\bar\partial v=f-\bar\partial G\), and \(u=v+G\) solves \(\bar\partial u=f\) on \(\Delta(r)\). \(\square\)

*Reference:* the induction is due to Grothendieck; [Demailly] and [Lebl SCV] give the same argument.

**Corollary 2.3.** For every holomorphic vector bundle \(E\) on \(X\) and every \(p\), the sequence of sheaves

\[
0\longrightarrow\Omega^p(E)\longrightarrow\mathcal E^{p,0}(E)\xrightarrow{\ \bar\partial_E\ }\mathcal E^{p,1}(E)\xrightarrow{\ \bar\partial_E\ }\cdots\xrightarrow{\ \bar\partial_E\ }\mathcal E^{p,n}(E)\longrightarrow0
\tag{2.2}
\]

is exact.

**Proof.** Exactness at \(\mathcal E^{p,0}(E)\) is the description of \(\Omega^p(E)\) as a kernel. Exactness at \(\mathcal E^{p,q}(E)\), \(q\geq1\), is a statement about germs: near a point, use a holomorphic frame and a coordinate polydisc, and apply Theorem 2.2 componentwise. \(\square\)

## 3. Fine sheaves and the Dolbeault isomorphism

**Lemma 3.1.** Let \(\mathcal F\) be a sheaf of modules over the sheaf \(\mathcal C^\infty_X\) of smooth complex functions on a manifold \(X\). Then \(H^k(U,\mathcal F)=0\) for every open set \(U\subset X\) and every \(k\geq1\). In particular this holds for every \(\mathcal E^{p,q}(E)\).

**Proof.** Let \(\mathcal U=(U_i)\) be an open covering of an open set \(U\), and choose a smooth partition of unity \((\chi_i)\) on \(U\) subordinate to it, with locally finite supports. For a Čech cochain \(c=(c_{i_0\ldots i_k})\), \(k\geq1\), define \((hc)_{i_0\ldots i_{k-1}}=\sum_i\chi_i\,c_{i\,i_0\ldots i_{k-1}}\), where \(\chi_ic_{i\,i_0\ldots i_{k-1}}\), defined on \(U_{i\,i_0\ldots i_{k-1}}\), is extended by zero to \(U_{i_0\ldots i_{k-1}}\); this extension is a section because the support of \(\chi_i\) is closed in \(U\) and contained in \(U_i\). The usual computation gives \(dh+hd=\mathrm{id}\) on cochains of degree \(k\geq1\), so the Čech cohomology \(\check H^k(\mathcal U,\mathcal F)\) vanishes for \(k\geq1\). By [Stacks, Tag 01EV](https://kokunoyumeto.github.io/stacks-zh-hans-cn/en/cohomology.html#cohomology-lemma-cech-vanish), \(H^k(U,\mathcal F)=0\) for every open \(U\) and \(k\geq1\). \(\square\)

**Theorem 3.2 (Dolbeault isomorphism).** For every holomorphic vector bundle \(E\) on a complex manifold \(X\), all \(p\) and \(q\), and every open \(U\subset X\),

\[
H^q(U,\Omega^p(E))\ \cong\ \frac{\ker\bigl(\bar\partial_E:\mathcal E^{p,q}(U,E)\to\mathcal E^{p,q+1}(U,E)\bigr)}{\bar\partial_E\,\mathcal E^{p,q-1}(U,E)} .
\tag{3.1}
\]

**Proof.** By Corollary 2.3 and Lemma 3.1, (2.2) restricted to \(U\) is a resolution of \(\Omega^p(E)\) by sheaves without higher cohomology, and such a resolution computes cohomology [Stacks, Tag 015E](https://kokunoyumeto.github.io/stacks-zh-hans-cn/en/derived.html#derived-lemma-leray-acyclicity). \(\square\)

**Corollary 3.3.** For every holomorphic vector bundle \(E\) on an \(n\)-dimensional complex manifold, \(H^q(X,\mathcal O(E))=0\) for \(q>n\).

**Proof.** There are no nonzero \((0,q)\)-forms for \(q>n\). \(\square\)

**Corollary 3.4.** On every polydisc \(\Delta(R)\), \(0<R_j\leq\infty\), and for \(q\geq1\), every class in \(H^q(\Delta(R),\mathcal O)\) restricts to zero on every smaller polydisc \(\Delta(r)\), \(r<R\).

**Proof.** Represent the class by a \(\bar\partial\)-closed \((0,q)\)-form by Theorem 3.2 and apply Theorem 2.2. \(\square\)

The full vanishing \(H^q(\Delta(R),\mathcal O)=0\), and much more, is proved in the lesson on Theorems A and B, where polydiscs appear as examples of Stein manifolds.

<span id="smooth-module-acyclicity-programme-proof"></span>

**Exact programme proof of smooth-module acyclicity and the resolution step.** The Čech contraction in Lemma 3.1 and its Stacks reference remain above. The following direct proof supplies an additional route through existing programme results, without assuming that every fixed-cover Čech complex computes sheaf cohomology. We use the Hausdorff, second-countable smooth manifolds of this course; every open subset has the same properties. Smooth subordinate partitions, with supports closed in that open subset and lying inside members of the chosen cover, are proved in [Local tools, Theorem 3.1](course:DG-FND/local-tools-for-bundles-and-transport#3-smooth-weights-with-controlled-support).

Let \(p:\mathcal B\to\mathcal C\) be a surjection of sheaves of \(\mathcal C^\infty_U\)-modules, and let \(s\in\mathcal C(U)\). Surjectivity on stalks gives an open cover \((W_i)\) and local lifts \(t_i\in\mathcal B(W_i)\) with \(p(t_i)=s|_{W_i}\): choose a preimage germ and shrink until its image agrees with \(s\). Choose a locally finite subordinate smooth partition \((\chi_i)\). A refined cover can be used, with each local lift restricted to the corresponding refinement member. On \(W_i\) the section \(\chi_i t_i\) is zero outside \(\operatorname{supp}\chi_i\). Because that support is closed in \(U\) and lies inside \(W_i\), this section glues with zero on \(U\setminus\operatorname{supp}\chi_i\), giving a section \(\widetilde{\chi_i t_i}\) on all of \(U\). The locally finite sum is a sheaf section, and
\[
t=\sum_i\widetilde{\chi_i t_i}\in\mathcal B(U),\qquad
p(t)=\sum_i\chi_i s=s.
\]
Thus taking sections on \(U\) preserves surjections of smooth-module sheaves. It already preserves kernels, which are computed on sections, so \(\Gamma(U,-)\) is exact on this module category. This assertion does not say that an arbitrary smooth module is flasque: it lifts through a surjection of sheaves, not through restriction to an arbitrary smaller open set.

The existence of injective resolutions in modules over any sheaf of rings, in particular \(\mathcal C^\infty_U\), is proved in [Injective modules, Theorem 2.3](course:derived-categories-and-sheaf-operations/injective-modules-and-bounded-below-derived-functors#2-functorial-injective-embeddings). In such a resolution of \(\mathcal F|_U\), write
\[
0\longrightarrow Q^{r-1}\longrightarrow I^r\longrightarrow Q^r\longrightarrow0,
\qquad Q^{-1}=\mathcal F|_U.
\]
The preceding lifting proof makes every section sequence exact. Consequently \(\Gamma(U,I^\bullet)\) is exact in positive degrees, so its positive cohomology vanishes. [Independence of coefficients, Theorem 3.1](course:AG-QC/cohomology-of-sheaves-on-ringed-spaces#3-coefficients-and-higher-direct-images) identifies these groups with the cohomology of the underlying abelian sheaf. Its proof uses [injectives are flasque, Lemma 1.1](course:derived-categories-and-sheaf-operations/injective-modules-and-bounded-below-derived-functors#1-extending-maps-and-extending-sections), and does not assume that forgetting coefficients preserves injectivity. This proves all of Lemma 3.1 on every open \(U\).

For Theorem 3.2 the exact acyclic-resolution input is [Bounded-below derived functors, Theorem 4.1(4)–(5)](course:derived-categories-and-sheaf-operations/injective-modules-and-bounded-below-derived-functors#4-right-derived-functors-on-bounded-below-complexes), with its full proof under [Why acyclic terms suffice](course:derived-categories-and-sheaf-operations/injective-modules-and-bounded-below-derived-functors#why-acyclic-terms-suffice-in-the-bounded-below-case). Apply it to the left exact section functor on abelian sheaves and the finite, hence bounded-below, resolution (2.2) restricted to \(U\). Each term is acyclic by the preceding argument. The differential \(\bar\partial_E\) is additive and holomorphic-function-linear; it is generally **not** smooth-function-linear. Acyclicity of its individual terms therefore comes from their smooth-module structures, whereas the complex and its cohomology comparison are taken in abelian sheaves (or equivalently in holomorphic modules by coefficient independence). These are different uses of the coefficient structures, not an assertion of smooth linearity for the differential.

The comparison identifies the cohomology of \(\Gamma(U,\mathcal E^{p,\bullet}(E))\) with \(H^q(U,\Omega^p(E))\), exactly as stated in Theorem 3.2. In degree zero the term of degree \(-1\) is zero. The full bundle statement, every open \(U\), the earlier Čech contraction, and the original source citations are retained.

*Smooth-module and resolution dependency integration: GPT-6 Astra (OpenAI), Ultra, 13 October 2026; author self-check, CC0. This is a scoped proof supplement, not an independent whole-course review.*

## 4. Exercises

**Exercise 4.1.** On \(\mathbf C\), find a smooth solution of \(\partial u/\partial\bar z=z\bar z\), and describe all solutions on a connected open set.

*Solution.* \(u_0=z\bar z^2/2\) works, since \(\partial(\bar z^2)/\partial\bar z=2\bar z\). Two solutions differ by a function with \(\partial/\partial\bar z\) equal to zero, that is, by a holomorphic function; so the solutions on a connected open set are \(u_0+h\) with \(h\) holomorphic.

**Exercise 4.2.** Let \(f=g\,d\bar z_1\) be a smooth \((0,1)\)-form on \(\mathbf C^2\). Show that \(\bar\partial f=0\) exactly when \(g\) is holomorphic in \(z_2\), and in that case write down a solution of \(\bar\partial u=f\) on \(\{|z_1|<1\}\times\mathbf C\).

*Solution.* \(\bar\partial f=\frac{\partial g}{\partial\bar z_2}\,d\bar z_2\wedge d\bar z_1\), which vanishes exactly when \(\partial g/\partial\bar z_2=0\). Then, with \(\psi\) a smooth function of \(z_1\) equal to \(1\) on \(|z_1|\leq1\) and with compact support, \(u(z)=-\frac1\pi\int_{\mathbf C}\psi(\zeta)g(\zeta,z_2)(\zeta-z_1)^{-1}\,dA(\zeta)\) is smooth and holomorphic in \(z_2\) by Lemma 2.1, so \(\bar\partial u=\frac{\partial u}{\partial\bar z_1}\,d\bar z_1=\psi g\,d\bar z_1\), which equals \(f\) where \(|z_1|<1\). This is the step \(k=1\) of the proof of Theorem 2.2.

**Exercise 4.3.** Explain why Lemma 3.1 does not apply to \(\mathcal O\), and give an open set \(U\) with \(H^1(U,\mathcal O)\neq0\).

*Solution.* \(\mathcal O\) is not a \(\mathcal C^\infty\)-module: a smooth multiple of a holomorphic function is not holomorphic in general, so the partition of unity argument leaves the sheaf. For \(U=\mathbf C^2\setminus\{0\}\), cover by \(U_1=\{z_1\neq0\}\) and \(U_2=\{z_2\neq0\}\). The function \(1/(z_1z_2)\) on \(U_1\cap U_2\) is a Čech \(1\)-cocycle. If it were a coboundary, \(1/(z_1z_2)=h_2-h_1\) with \(h_i\in\mathcal O(U_i)\); expanding in Laurent series on \(U_1\cap U_2\), \(h_1\) has only terms \(z_1^az_2^b\) with \(b\geq0\), and \(h_2\) only terms with \(a\geq0\), so neither produces the term \(z_1^{-1}z_2^{-1}\). Since the Čech \(H^1\) of a covering injects into \(H^1(U,\mathcal O)\) [Stacks, Tag 0B8R](https://kokunoyumeto.github.io/stacks-zh-hans-cn/en/cohomology.html#cohomology-lemma-cech-h1), the class is nonzero.

**Exercise 4.4.** Let \(L\) be a holomorphic line bundle on a compact Riemann surface \(X\). Show that \(H^q(X,\mathcal O(L))=0\) for \(q\geq2\).

*Solution.* This is Corollary 3.3 with \(n=1\).

## References

- [Demailly] J.-P. Demailly, *Complex Analytic and Differential Geometry*, version of 21 June 2012, freely available from the author with permission to copy, modify and redistribute with credit. <https://www-fourier.univ-grenoble-alpes.fr/~demailly/manuscripts/agbook.pdf>
- [Lebl SCV] J. Lebl, *Tasty Bits of Several Complex Variables*, version 4.4 (2026). <https://www.jirka.org/scv/>
- [Stacks] The Stacks project, cited by tag; each tag links to the same result in the AI Integrated Stacks Project. <https://stacks.math.columbia.edu/>