Why is a Kähler form not always holomorphic?

1. Mar 6, 2014

nonequilibrium

Any Kähler form (?) can be written in local coordinates as $\omega = \frac{i}{2} \sum h_{ij} dz^i \wedge d z^j$ with $h_{ij}(z) = \delta_{ij} + \mathcal O(z^2)$.

But that would mean $\bar \partial \omega = \frac{i}{2} \sum \frac{\partial h_{ij}}{\partial \bar z_k} d \bar z^k \wedge dz^i \wedge d z^j = 0$ since $\frac{\partial h_{ij}}{\partial \bar z_k} = \mathcal O(z)$ is zero at our point $z = 0$.

I must be making a stupid mistake. What is it?

2. Mar 6, 2014

Ben Niehoff

Your mistake is that the Kahler form is a (1,1) form:

$$\omega = \frac{i}{2} \, h_{i \bar \jmath} \, dz^i \wedge d \bar z^j$$

3. Mar 6, 2014

Ben Niehoff

Also, this reasoning only applies at a point. The property of holomorphicity requires an open set around a point.

4. Mar 7, 2014

nonequilibrium

Thanks for pointing out my typo with the bar, but that doesn't change the argument though.

As for your second post: the reasoning is as follows: I check that $\bar \partial \omega$ is zero at every point (but to calculate it at a point, I indeed need a neighbourhood around that point). So fix a point z = 0. Then around that point we can write $\omega$ as I did with the expansion of $h$. Then I act with $\bar \partial$ and afterwards I put $z = 0$ to get that at that point $\bar \partial \omega = 0$. Since our original point was arbitrary, this is true for every point!

5. Mar 7, 2014

quasar987

The Kähler form is (1,1) and $(dz^i\wedge d\overline{z}^j)$ does not define a holomorphic structure on the bundle of (1,1) forms. Only on the bundle of (p,0) forms do the complex coordinates on M define a holomorphic structure in this way.

Last edited: Mar 7, 2014
6. Mar 7, 2014

nonequilibrium

So is my mistake in interpreting $\bar \partial \omega = 0$ to mean that my form is holomorphic? I thought that was the definition of holomorphic. What is the correct meaning?

7. Mar 7, 2014

quasar987

To speak of holomorphicity of a section of a complex vector bundle you need first to put a holomorphic structure on your bundle. That is, you need to specify trivializing local frames (e_i) for it such that the transition maps between these frames are holomorphic. Then, and only then, will it be well-defined to say a section $s=\sum_i s_ie_i$ is holomorphic if $\overline{\partial}s_i=0$. Otherwise, we may have $\overline{\partial}s_i=0$ but $\overline{\partial}s_i'\neq 0$ wrt another local frame (e_i').

Also, you seem to think your argument shows the functions h_ij to be holomorphic but that is not so:

Fix a point p in your manifold and coordinates (z) in a nbhd of p. As Ben said, holomorphicity of h_ij at p means $\overline{\partial}h_{ij}=0$ in a nbhd of p. What your argument shows is that for any point p', you can always find a coordinate system (z') wrt which $\overline{\partial}h'_{ij}=0$ at p'. But this does not imply that $\overline{\partial}h_{ij}=0$ at p' wrt (z) because the transition function btw $(d\overline{z})$ and $(d\overline{z'})$ (which appear when you write h_ij in terms of h'_ij) are not holomorphic. (Same reason why (dzi∧dz¯j) does not define a holomorphic structure on the bundle of (1,1) forms.)

Last edited: Mar 7, 2014
8. Mar 7, 2014

nonequilibrium

I'm confused, I'm not using coordinate expressions in my definition, I'm just using the fact that the coordinate independent notion $\bar \partial \eta$ (defined in terms of the exterior derivative $d$ etc) with $\eta = \eta_{IJ} dz^I \wedge d \bar z^J$ can be written in coordinates as $\bar \partial \eta = \frac{\partial \eta_{IJ}}{\partial \bar z^k} d \bar z^k \wedge dz^I \wedge d\bar z^J$

9. Mar 7, 2014

quasar987

I should have said explicitely that your definition of a form being holomorphic iff $\overline{\partial}\eta=0$ is wrong. That is not how holomorphicity is defined. See paragraph 1 of post #7.

10. Mar 7, 2014

quasar987

Ah and also it is the bundle of (p,0) forms that admit a natural holomorphic structure, not the (0,p) forms as I said in post #5, so I edited that post as well as #7 accordingly.

11. Mar 7, 2014

nonequilibrium

Ah okay, thanks. So I see what I did wrong, but still trying to understand what the correct notion of holomorphicity of (p,q)-forms is. I think I get what you are saying, but that would seem to suggest that we need to make an extra choice before we have that notion. Is there no canonical notion of holomorphicity? E.g. when people say "a Calabi-Yau mfd is a compact Kähler mfd with holomorphic top-form", how am I to interpret that?

EDIT: I just realized another way to see my original thought was wrong, since if "holomorphic top-form" meant "$\bar \partial \omega = 0$", then that's trivial because $\bar \partial$ of a top-form is always zero! (since $\Omega^{2n+1} = 0$))

12. Mar 7, 2014

quasar987

There is no natural/canonical notion of holomorphic (p,q)-form if q>0. On the other hand, on $\Omega^{(p,0)}$, there is a natural holomorphic structure given by the local frames $dz^{i_1}\wedge\ldots dz^{i_p}$ for every local coordinate system (z) on M (check that the transition functions between these local frames are indeed holomorphic). A holomorphic p-form is then one of type (p,0) such that $\overline{\partial}\eta=0$.

A Calabi-Yau manifold is (depending on your definition...) a compact Kähler manifold with trivial canonical bundle. The canonical bundle of a complex n-manifold is the bundle $\Omega^{(n,0)}$ with its natural holomorphic structure. Triviality of this bundle then means that there exists a global nowhere vanishing holomorphic n-form. Note that $\Omega^{(n,0)}$ may be trivial as a complex line bundle but not as a holomorphic line bundle in which case we don't have Calabi-Yau.

Last edited: Mar 7, 2014
13. Mar 7, 2014

quasar987

Note that a holomorphic top form is an element of $\Omega^{(n,0)}\subset \Omega^n$, and not of $\Omega^{2n}$. That is because $\Omega^{(p,0)} = \wedge^p\Omega^{(1,0)}$ and $\Omega^{(1,0)}$ has complex dimension 1, so we can wedge it with itself only n times before it vanishes. So while $\partial \eta=0$ for a (n,0)-form, $\bar\partial \eta\in\Omega^{(n,1)}$ does not vanish in general.

Last edited: Mar 7, 2014
14. Mar 7, 2014

quasar987

A last remark: While it is true that for (p,0)-forms, holomorphicity in the sense of post #7 is the same as the condition $\bar \partial \eta=0$, for a general (p,q)-form, the condition $\bar \partial \eta=0$ has nothing to do with holomorphicity of anything.

As an illustration, consider a (0,1)-form on a cx 2-manifold: $\eta=\eta_1d\bar z_1 + \eta_2 d\bar z_2$. Then
$$\bar\partial \eta = \left(\frac{\partial \eta_1}{\partial \bar z_1} - \frac{\partial \eta_1}{\partial \bar z_2}\right)d\bar z_1\wedge d\bar z_2$$
What does the vanishing of this has to do with holomorphicity ?

15. Mar 8, 2014

nonequilibrium

Thank you very much!