2
$\begingroup$

A class I am current taking makes some use of differential forms. In particular, we are asked to show that a certain differential form is smooth. I know in general that a differential $k$-form, $\omega$ is smooth if for smooth vector fields $X_1,\ldots,X_k$, $\omega_p(X_1(p),\ldots,X_k(p))$ is smooth in $p$. Is there a good way to visualize this in local coordinates, and is there a general easy way to prove that a differential form is smooth?

Also, and I believe related, is the question of how these vector fields interact with the differential form. ie is it true that if we have a $2$-manifold (in local coordinates) $X_1(p)=a_{11}(p)\frac{\partial}{\partial x_1}+a_{12}(p)\frac{\partial}{\partial x_2}$ and $X_2(p)=a_{21}(p)\frac{\partial}{\partial x_1}+a_{22}(p)\frac{\partial}{\partial x_2}$, does if $\omega_p=f(p) dx_1\wedge dx_2$, what would the formula for be for $\omega_p(X_1(p),X_2(p))$. I think this question is simple yet would illustrate to me how these things work.

I have looked up references for differential forms, but none seem to answer the questions about them that I seem to have. Any reference however would also be greatly appreciated.

Thank you for any help.

$\endgroup$
3
$\begingroup$

Smoothness of a form is the same as the smoothness of its coefficients in some (equivalently, any) system of coordinates. Let's verify this on the $2$-form example in your post. First, suppose that $f$ is a smooth function. Then for any smooth vector fields $X_1,X_2$ we have $$\omega_p(X_1(p),X_2(p)) = f(p)(a_{11}(p)a_{22}(p)-a_{12}(p)a_{21}(p))\tag 1$$ which is smooth. Note that (1) is calculated using the duality relations
$$(dx_1\wedge dx_2)\left(\frac{\partial}{\partial x_1},\frac{\partial}{\partial x_2}\right)=1$$ $$(dx_1\wedge dx_2)\left(\frac{\partial}{\partial x_2},\frac{\partial}{\partial x_1}\right)=-1$$ $$(dx_1\wedge dx_2)\left(\frac{\partial}{\partial x_1},\frac{\partial}{\partial x_1}\right)=0$$ $$(dx_1\wedge dx_2)\left(\frac{\partial}{\partial x_2},\frac{\partial}{\partial x_2}\right)=0$$ That is, the result of plugging basis vectors into a basis form is $1$ if the vector indices match the form indices exactly; it is $(-1)^\sigma$ if they match up to $\sigma$ transpositions, and is $0$ in other cases.

Conversely, suppose that the form $\omega_p$ is smooth. Plug in the coordinate vector fields to obtain $$\omega_p\left(\frac{\partial}{\partial x_1},\frac{\partial}{\partial x_2}\right)= f(p)$$ and conclude that the coefficient $f$ is smooth.

For general $k$-forms the argument is exactly the same.


How one proves that a form is smooth depends on how the form was constructed. The usual way to construct a form involves some of the following steps:

  • taking a scalar function ($0$-form)
  • differentiating a form
  • multiplying two forms
  • pulling a form back by a smooth map

All of the above operation preserve $C^\infty$ smoothness: if your ingredients are smooth, so is the result.

Sometimes the most nontrivial part of proving that a form is smooth is to prove that it is well-defined, that is, different choices of local coordinates give the same form. The Fubini-Study form is an example of such situation. For simplicity take $n=1$. The Fubini-Study form is the 2-form defined on $\mathbb C\mathbb P^1=\mathbb C\cup\{\infty\}$ by means of two charts: $$\omega = \partial\bar\partial \ln(1+|z|^2),\quad z\ne \infty$$ $$\omega = \partial\bar\partial \ln(1+|1/z|^2),\quad z\ne 0$$ One then has to check that the definitions are consistent on the overlap $z\ne 0,\infty$. Indeed, the difference of Kähler potentials on the overlap is $$\ln(1+|z|^2)-\ln(1+|1/z|^2)= \ln (|z|^2) $$ which is a harmonic function. Since $\partial\bar\partial\ln (|z|^2)=0$, the form is well-defined. Smoothness is not an issue: it's a local property, to be checked on each coordinate chart separately. And it's clear that on each chart we differentiate a $C^\infty$ function.

$\endgroup$

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.