4
$\begingroup$

Why do we need compactness to have $ \langle\Delta_\partial\omega,\omega\rangle = || \partial \omega||^2 + ||\partial^*\omega||^2 $?

I think $ \langle\Delta_\partial\omega,\omega\rangle =<\partial\partial^*\omega,\omega>+<\partial^*\partial\omega,\omega>=<\partial^*\omega,\partial^*\omega>+<\partial\omega,\partial\omega>= || \partial \omega||^2 + ||\partial^*\omega||^2 $. Why do we need compactness then?

$\endgroup$
  • $\begingroup$ @peterag I am just confused, it seems that it’s true even without compactness $\endgroup$ – Danny Feb 21 at 6:56
  • $\begingroup$ how did you define $\partial ^*$? $\endgroup$ – klirk Feb 21 at 17:15
5
$\begingroup$

The problem is that the adjoint operator $\partial^*$ in general only exists on compact manifolds.

We want an operator $\partial^*$, such that $\left< \partial \alpha,\beta\right>=\left< \alpha, \partial^*\beta\right>$ for all $\alpha, \beta$.

Let $\alpha$ a $(p-1,q)$-form and $\beta$ a $(p,q)$-form on $M$. We calculate: (see Huybrechts p.125) $$\left<\partial \alpha, \beta\right>= \int_M \partial \alpha\wedge *\overline {\beta} \stackrel{Leibniz~Rule}= \int_M \partial(\alpha \wedge * \overline \beta)-(-1)^{p+q-1}\int_X\alpha\wedge \partial(*\overline \beta)$$ The first summand vanishes if $M$ is compact due to Stokes' theorem as $\alpha \wedge *\overline \beta$ is of type $(n-1,n)$ and hence $\partial(\alpha \wedge *\overline \beta)=d(\alpha \wedge *\overline \beta)$.

For the second summand, note that $\partial(*\overline \beta)$ is of type $(n-p+1,n-q)$. Hence $$\partial(*\overline \beta)= (-1)^{(2n-p-q+1)(p+q-1)}**\partial(*\overline \beta)= (-1)^{-(p+q-1)^2}**\partial(*\overline \beta)=(-1)^{-(p+q-1)}**\partial(*\overline \beta).$$ Inserting this in the second summand we get: $$-(-1)^{p+q-1}\int_X\alpha\wedge \partial(*\overline \beta)= -(-1)^{p+q-1}\int_X\alpha\wedge (-1)^{-(p+q-1)}**\partial(*\overline \beta)\\= -(-1)^{p+q-1-(p+q-1)} \left< \alpha, \overline{*\partial*\overline\beta}\right>=\left< \alpha,- *\overline \partial*\beta\right> $$

We conclude $$\left< \partial \alpha, \beta\right>= \left< \alpha,- *\overline \partial*\beta\right> + \int_M \partial(\alpha \wedge * \overline \beta).$$ If $M$ is compact, then the second summand vanishes and $\partial^*=-*\overline\partial *$.
If $M$ is not compact, then the second summand does in general not vanish and could take any value, so there is no adjoint operator in this case.

$\endgroup$
  • $\begingroup$ you have a typo on "We want an operator $\partial^*$, such that $\left< \partial \alpha,- \partial^*\beta\right>$ for all $\alpha, \beta$" $\endgroup$ – Danny Feb 21 at 23:57
  • $\begingroup$ Also, O'Wells' book, p.169 says $\partial^*=-\bar{*}\partial\bar{*}$ $\endgroup$ – Danny Feb 22 at 0:00
  • $\begingroup$ @Danny: oh, fixed it $\endgroup$ – klirk Feb 22 at 11:06
  • 1
    $\begingroup$ @Danny: $\bar *\neq *$, but $\bar * \partial \bar*$ should still be equal to $*\bar \partial *$. $\endgroup$ – klirk Feb 22 at 11:25

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.