3

Let $M$ be a manifold and $\alpha \in \Omega^1(M)$ is a nowhere-vanishing one-form. I have to show that $\ker \alpha$ is involutive if and only if $\alpha \wedge d\alpha = 0$ but I'm having some trouble to prove this. I found this question - Why is $\ker\omega$ integrable iff $\omega\wedge d\omega=0$? - but I'm having a hard time proving this in the case when $M$ is a manifold of an arbitrary dimension.

Mee98
  • 1,133

1 Answers1

2

Let's think for a moment about what it means to say $\omega\wedge d\omega = 0$ when $\omega\in\Omega^1(M)$. If you consider extending $\omega$ to a basis $\{\omega=\omega_1,\omega_2,\dots,\omega_n\}$ for the $1$-forms locally, you can easily convince yourself that $d\omega = \omega\wedge\eta$ for some $1$-form $\eta$, again locally. Then (locally) it is clear that $d\omega(X,Y) = 0$ whenever $\omega(X)=\omega(Y)=0$. Thus, $\omega\wedge d\omega = 0$ implies involutivity.

On the other hand, if the distribution is involutive, then you already know that $d\omega(X,Y) = 0$ whenever $\omega(X)=\omega(Y)=0$. But then this tells us that $d\omega = \omega\wedge\eta$ for some $1$-form $\eta$; for, if not, we would have $$d\omega=\sum\limits_{i=2\\i<j}^n f_{ij}\omega_i\wedge\omega_j$$ for some functions $f_{ij}$, with some coefficient $f_{i'j'}$ nonzero (at a particular point). Choosing $X,Y$ so that $\omega_{i'}(X)=1$ and $\omega_k(X)=0$ for all $k\ne i'$ (including $k=1$) and $\omega_{j'}(Y)=1$ and $\omega_k(Y)=0$ for all $k\ne j'$ (likewise) shows that $d\omega(X,Y)\ne 0$.

Ted Shifrin
  • 115,160
  • The second part of my exercise is proving that for all $\omega \in \Omega^2(M)$ it holds that $\alpha \wedge \omega = 0$ iff there is a $\beta \in \Omega(M)$ such that $\omega = \alpha \wedge \beta$ so I don't think that I can use that here. However, I only have a problem with the proof that assumes that $\ker \alpha$ is involutive. I see why $\alpha \wedge d\alpha = 0$ when you take two vector field in $\ker \alpha$ but not why it holds for random vector fields. – Mee98 Jan 26 '20 at 19:18
  • 1
    You can use a permutation calculation for $(\alpha\wedge d\alpha)(X,Y,Z)$. Remember that you can reduce by multilinearity to the case where two of them lie in your distribution. – Ted Shifrin Jan 26 '20 at 19:44
  • 1
    Nearly 4 years later, but for the second part, why do such vector fields $X$ and $Y$ exist? – A Name Feb 15 '24 at 01:23
  • 1
    @AName I’m not completely sure which part of the second part you’re referring to. But this is all local, so take the basis ${X_i}$ for the tangent bundle dual to the basis we have for the cotangent bundle. Then take $X=X_{i’}$ and $Y=X_{j’}$. – Ted Shifrin Feb 15 '24 at 02:20
  • Yes thank you that's exactly it. Sorry I wasn't clear. – A Name Feb 15 '24 at 02:24