4
$\begingroup$

Suppose $\omega$ is a nonvanishing $1$-form on a $3$-manifold $M$. It's known that $\ker\omega$ is an integral distribution iff $\omega\wedge d\omega=0$.

I'm trying to understand this, but I don't get why $\ker\omega$ integrable implies $\omega\wedge d\omega=0$.

I worked out so far: suppose $\omega\wedge d\omega=0$. Let $X,Y$ be vector fields in $\ker\omega$. Then $$ d\omega(X,Y)=X(\omega(Y))-Y(\omega(X))-\omega([X,Y]) $$ so $\omega([X,Y])=-d\omega(X,Y)$. Since $\omega$ is nonvanishing, we can find $Z$ such that $\omega(Z)\neq 0$. Then $$ 0=(\omega\wedge d\omega)(X,Y,Z)=c\omega(Z)d\omega(X,Y) $$ for some $c\neq 0$. Thus $d\omega(X,Y)=0$, so $\omega([X,Y])=0$, so $[X,Y]\in\ker\omega$. Then $\ker\omega$ is involutive, so by Frobenius, it is integrable.

On the other hand, if $\ker\omega$ is integrable, then the annihilator ideal $I(\ker\omega)$ is closed under $d$. Clearly $\omega\in I(\ker\omega)$, so by assumption $d\omega\in I(\ker\omega)$. I'm trying to show $\omega\wedge d\omega(X,Y,Z)=0$ for any three vector fields, but it looks like a dead end. Is it possible to assume $X,Y\in\ker\omega$ without loss of generality some how? Because that would make the calculation work out.

$\endgroup$
2
$\begingroup$

Yes, you can assume $X,Y\in\ker(\omega)$. Since $M$ is three dimensional $L=\bigwedge^3TM$ is a line bundle over $M$, and $\omega\wedge d\omega$ is naturally defined on (local) sections of this line bundle. Take a local basis $(X,Y)$ of vector fields parallel to $\ker(\omega)$, and a smooth local extension $Z$ of some vector $Z_m\in T_mM\setminus\ker(\omega_m)$, then $Z\wedge X\wedge Y$ is a local basis of the line bundle $L$, and evaluating $\omega\wedge d\omega$ on it yields zero (through your calcuation), so that this form is actually locally zero, and hence globally zero.

$\endgroup$
  • $\begingroup$ Thanks. So would it make sense to say $\omega\wedge d\omega(X,Y,Z)=c\omega(Z)d\omega(X,Y)=0$ since $d\omega\in I(\ker\omega)$? $\endgroup$ – Clara Mar 20 '15 at 6:47
  • 1
    $\begingroup$ Yes, with $c=1$. $\endgroup$ – Olivier Bégassat Mar 20 '15 at 6:53

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.