Ok so I am following a set of notes on complex differential geometry and there is a theorem that says the following:

If $E$ is a complex vector bundle over a complex manifold and $\overline{\partial^E}:\Omega^{p,q}E \to \Omega^{p,q+1}E$ is a linear operator that satisfies the Leibniz rule. Then $\overline{\partial^E}$ is induced from the structure of a holomorphic vector bundle on $E$ iff $(\overline{\partial^E})^2=0$

My first question is on the wording of the above theorem, is this theorem saying that $E$ is a holomorphic vector bundle iff $(\overline{\partial^E})^2=0$?

The second question is about proving the $\impliedby$ implication. My notes suggest using the Newlander-Nirenberg theorem. So I have a couple of ideas about proving it but I am not sure how to formalise it.


The Newlander-Nirenberg (NN) gives a criterion for an almost complex manifold to be a complex manifold. In particular it says that $E$ will be a complex manifold iff $[X,Y]_p\in T_p^{(1,0)}E$ whenever $X,Y\in T_p^{(1,0)}E$

If i can show that $E$ is a complex manifold (via NN) then i can use it's charts to construct bi-holomorphic trivialisations of $E$ then the result follows. Here are the issues I am having:

I need to first equip $E$ with the structure of an almost complex manifold in order to use the NN. So what type of linear map am I meant to construct on its tangent spaces? I think it would need to be one related to the $\overline{\partial^E}$ operator in order for the $(\overline{\partial^E})^2=0$ condition to imply the criteria needed to use the NN theorem.

If someone could give me advice on how to complete the proof, construct the almost complex structure on $E$ which would allow me to use NN or even point me to a reference where the proof is established I'd be very very grateful.

Thanks for reading.


To answer your first question, a more precise statement should be:

If $\bar{\partial}^{E}:\Omega^{p,q}\to \Omega^{p,q+1}$ is an operator on comple vector bundle $E$ satisfying Leibniz rule and $(\bar{\partial}^{E})^2=0$, then $\bar{\partial}^{E}$ defines a holomorphic structure on $E$.

For your second question, you need to relate $\bar{\partial}^2=0$ with the Newlander-Nirenberg condition.

This works in general (forget about the vevtor bundle structure), so let me start with an almost complex structure $J$ on a smooth manifold $M$, i.e., $J:TM\to TM$ is an operator such that $J^2=-\text{Id}$. One has decomposition of $TM\otimes \mathbb C=TM^{1,0}\oplus TM^{0,1}$ with respect to the eigenspaces $\pm 1$ of $J$. Let $v_1,...,v_n$ be a local basis on $T^{0,1}M$ and $v_1^*,...,v_n^*$ the dual sections in $\Omega^{0,1}M$, then the $\bar{\partial}$ operator can be written as $$\bar{\partial}=\sum_iv_i\otimes v_i^*:\Omega^{p,q}\to \Omega^{p,q+1}$$ For example, if $f$ is a smooth function, $\bar{\partial}(f)=\sum_iv_i(f)v_i^*$. Therefore $$\bar{\partial}^2=\bar{\partial}(\sum_iv_i\otimes v_i^*)=\sum_{i<j}[v_j,v_i]v_i^*\wedge v_j^*\tag{1}\label{1}$$

where $[v_j,v_i]=v_j\circ v_i-v_i\circ v_j$ is the Lie bracket of vector fields.

Based on the discussion above, we claim that:

Claim: $J$ is integrable (Newlander-Nirenberg condition holds) if and only if $\bar{\partial}^2=0$.

$\textit{Proof.}$ It follows from $(\ref{1})$ that $\bar{\partial}^2=0$ implies that $v_1,...,v_n$ are $\textit{mutually commutative}$, so any $[\sum_if_iv_i,\sum_jg_jv_j]$ is a linear combination of $v_i$, so $[T^{0,1},T^{0,1}]\subset T^{0,1}$ (It's easy to check this is equivalent to the NN condition on $T^{1,0}$).

Conversely, assuming $[T^{0,1},T^{0,1}]\subset T^{0,1}$, to make sure $\bar{\partial}^2=0$, according to $(\ref{1})$, we need to find a mutually commutative basis $v_1,..,v_n$. This essentially is the Frobenius theorem. One can imitate the proof to give a complex version of that.$\tag*{$\blacksquare$}$

  • $\begingroup$ I know it's not particularly important, but don't you get $\sum\limits_{i<j}[v_j, v_i]v_j^*\wedge v_i^*$? $\endgroup$ Feb 5 '21 at 22:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.