MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

So from the $\overline{\partial}$-Poincare lemma, there is a short exact sequence of sheaves on $X = \mathbb{P}^1$

$$0 \to \Omega \to A^{1,0} \to Z^{1,1} \to 0$$

where $\Omega$ is the sheaf of holomophic 1-forms, $A^{1,0}$ is the sheaf of (1,0)-forms, $Z^{1,1}$ is sheaf of closed (1,1)-forms and the surjection is apply $\overline{\partial}$. The higher cohomology of $A^{1,0}$ vanishes which gives an isomorphism:

$$H^1(X,\Omega) \cong \frac{\Gamma(Z^{1,1})}{\overline{\partial}\Gamma(A^{1,0})} =: H_{\overline{\partial}}^{1,1}(X)$$

Both groups are isomorphic to $\mathbb{C}$. The first group can be described explicitly via Cech cohomology and the standard cover of $X$; in suitable coordinates a generator is $\dfrac{dz}{z}$.

What is the image of $\dfrac{dz}{z}$ in $H_{\overline{\partial}}^{1,1}(X)$? More specifically, can you write down a global closed (1,1) form that represents the class of $\dfrac{dz}{z}$?

I know that being able to avoid descriptions such as these is one of the great virtues of Cech Cohomology, but I guess some stubborn part of me would like to see the Dolbeault description.

share|cite|improve this question
Wild guess: Fubini-Study metric. – S. Carnahan Mar 29 '10 at 7:50
up vote 7 down vote accepted

I wrote a blog post about almost exactly this question. I'll give a summary here:

Since $H^{1,1}(X)$ is one dimensional, I could answer your question by giving anythng with the correct integral. However, I'll try to give you the kind of cocycle which actually comes out of the proof of Dolbeaut-Cech equality.

Your cocycle isn't $dz/z$ but, rather, $dz/z$ with a specific choice of open cover of $X$. Lets say your choice is $U_1 \cup U_2$, where $U_1 = \{ z : z \neq \infty \}$ and $U_2 = \{ z : z \neq 0 \}$. Refine your cover to $V_1 \cup V_2$, where $V_1 = \{ z : |z| > r \}$ and $V_2 = \{ z : |z| < r^{-1} \}$ for some $r < 1$.

Let $\theta_1$ and $\theta_2$ be $1$ forms on $V_1$ and $V_2$ such that $\theta_1|_{V_1} - \theta_2|_{V_2} = dz/z$. Then $\overline{\partial} \theta_1$ and $\overline{\partial} \theta_2$ have equal restrictions to $V_1 \cap V_2$. The $(1,1)$-form you are looking for is their common value, which I'll call $\omega$.

Let's first do a fake solution. A real solution would look like a $C^{\infty}$ smearing out of this one.

We'll work in the degenerate case $r=1$, so we are only gluing along a circle, not an annulus. We'll take $\theta_1 = (1/2) \ \overline{z}\ dz$ and $\theta_2 = -(1/2) \ dz / (\overline{z} z^2)$. Notice that both $\theta_1$ and $\theta_2$ restrict to $dz/z$ on the unit circle, but are constructed to extend smoothly to the appropriate discs.

So $\overline{\partial} \theta_1 = (1/2) d \overline{z} d z$ and $\overline{\partial} \theta_2 = (1/2) dz d \overline{z} / (\overline{z}^2 z^2)$. Our $\omega$ is formed by gluing these two differential forms together.

A genuine smooth solution would be like this, but would interpolate smoothly between these two. If you push forward in a brute force manner, you'll get something with bump functions in it.

If you are more clever, you may discover the solution $$\theta_1 = \frac{dz}{z} \left( 1- \frac{1}{1+z \overline{z}} \right)$$ and $$\theta_2 = - \frac{dz}{z} \left( \frac{1}{1+z \overline{z}} \right).$$

You should check that $\theta_1|_{V_1} - \theta_2|_{V_2} = dz/z$ and that $\theta_i$ is smooth and well-defined on $U_i$.

Then $$\overline{\partial} \theta_1 = \overline{\partial} \theta_2 = \frac{dz \ d\overline{z}} {(1+z \overline{z})^2}.$$

This is, as Scott guessed, the Fubini-Study form.

share|cite|improve this answer
There are some minor errors here, although they don't effect the main point. I'll edit when I get a chance later. – David Speyer Mar 29 '10 at 13:34
OK, I think I got them all. – David Speyer Mar 29 '10 at 15:12
Awesome thanks! – solbap Mar 29 '10 at 15:44

The 2-form $\omega$ which is the derivative of $dz/z$ is

$\displaystyle\omega = \bar{\partial}\ \frac{dz}{z}=\frac{\partial}{\partial\bar{z}}\left(\frac{1}{z}\right)d\bar{z}\wedge dz = \pi\delta\ d\bar{z}\wedge dz = 2\pi i\delta\ dx\wedge dy$.

Hence $\omega$ belongs to the class of 2-forms which integrate to $2\pi i$. Another representative of this class is

$\displaystyle\frac{d\bar{z}\wedge dz}{(1+|z|^2)^2}$

corresponding to smearing out the point mass $2\pi i\delta$ evenly over the sphere.

share|cite|improve this answer
Can you explain what you are doing here? This looks like nonsense, but your end result does seem to be morally right in some sense. (I assume that $\delta$ is the $\delta$ measure at $0$.) Here are some of the things which confuse me: – David Speyer Apr 3 '10 at 13:03
(1) your construction should remember not just the form $dz/z$, but the choice of Cech cover. If we switched the order of the open sets $\mathbb{P}^1 \setminus \{ 0 \}$ and $\mathbb{P}^1 \setminus \{ \infty \}$, we should negate the cohomology class. – David Speyer Apr 3 '10 at 13:03
(2) I guess it really is true that $d (dz/z) = \delta_0$ in the sense that $\int_{\partial A} dz/z = \int_A \delta_0$ for any disc $A$. So this is going to be a horrible nitpick but: In what sense am I allowed to do DeRham cohomology with measures? The theorem/definition I know is that $H^2$ is (closed smooth $2$-forms)/d(smooth $1$-forms). Is there an analogous result where "smooth forms" is replaced by measures? – David Speyer Apr 3 '10 at 13:16
(3) Why didn't you pick up a second $\delta$ function at $\infty$, where $dz/z$ also has a pole? Presumably, the answer to this is related to the answer to (1). – David Speyer Apr 3 '10 at 13:17
Hi David, Griffiths and Harris chapter 3 has a section on how to define cohomology using distributions. I would start there. – Clay Cordova Apr 13 '10 at 18:10

The answer is zero. The image of the given 1-form belongs to the class of exact 2-forms and the simplest such 2-form is the zero 2-form.

To actually calculate the derivative of the given 1-form, first lift it to homogeneous coordinates $[z:w]$. The 1-form then reads $$\theta=\frac{w}{z}\ d\left(\frac{z}{w}\right)=\frac{dz}{z}-\frac{dw}{w}.$$ Now choose a new affine coordinate $u=z+1=w-1$ and the 1-form in this coordinate is $$\theta=\frac{du}{u-1}-\frac{du}{u+1},$$ which has the derivative $$\bar{\partial}\theta=\pi\delta_1\ d\bar{u}\wedge du\ -\ \pi\delta_{-1}\ d\bar{u}\wedge du$$ and this integrates to zero like any other exact 2-form.

share|cite|improve this answer
This answer, on the other hand, is wrong. Your answer should be a nontrivial cohomology class. – David Speyer Apr 3 '10 at 13:17

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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