I'm struggling a little bit thinking about this apparenlty innocent exercise. I'll provide an incomplete solution.


Let $M$ be an $n$-dim manifold and let $S\subset M$ be an embedded $(n-1)$-dim submanifold. For every $P\in M$ let's identify the tangent space $T_PS$ with its image in $T_PM$ through the differential of the inclusion of $S$ in $M$.

$(i)$ Let $\alpha\in A^1(M)$ be a $1$-form on $M$ such that $T_PS\subseteq \ker(\alpha_P)$, for every $P\in S$. Prove that $(\alpha\wedge d\alpha)_P=0$, for every $P\in S$.

Solution (i)

Now, if $\alpha$ is a $1$-form on $M$ such that $T_PS\subseteq \ker(\alpha_P)$ for every $P\in S$, then, for every $P\in S$ there exist a local chart $\phi=(x^1,\ldots,x^n)$ in $P$ such that, with respect to that coordinates,


i.e. all the coefficients of $\alpha$ are zero except for the $n$-th (but I'm not at all sure about it!). So we get

$$d\alpha=\sum_{i=1}^{n-1}\frac{\partial\alpha_n(x)}{\partial x^i}dx^i\wedge dx^n$$ and

$$(\alpha\wedge d\alpha)=\alpha_n(x)dx^n\wedge\sum_{i=1}^{n-1}\frac{\partial\alpha_n(x)}{\partial x^i}dx^i\wedge dx^n=\sum_{i=1}^{n-1}\alpha_n(x)\frac{\partial\alpha_n(x)}{\partial x^i}dx^n\wedge dx^i\wedge dx^n=0$$

$(ii)$ Let's now suppose that $M$ is oriented and let $\omega\in A^n(M)$ be a volume form associated to the fixed orientation. Prove that for every $(n-1)$-form $\phi\in A^{n-1}(M)$ there exists a unique vector field $X^\phi\in\mathcal{T}(M)$ such that


for every $P\in M$ and every $v_1,\ldots,v_{n-1}\in T_PM$. Besides, prove that $\Theta\colon A^{n-1}(M)\to\mathcal{T}(M)$ defined by $\Theta(\phi)=X^\phi$ is an isomorphism.

For this second problem I really don't know how to begin.

  • $\begingroup$ Yes, for the first part, you're choosing local coordinates on $M$ in which $S$ is given by $x^n=0$. For the second part, maybe you should think about the inverse map $\Theta^{-1}$ (remembering that $\Lambda^{n-1} (T^*_PM)$ and $T_PM$ are each $n$-dimensional). $\endgroup$ – Ted Shifrin Sep 19 '16 at 22:32
  • $\begingroup$ @TedShifrin: Shouldn't there also be a comment about the fact that $\alpha_p = \sum_{i=1}^{n-1} a_n(p) \ (dx^i)_p$? (i.e a sum, not just $a_n dx^n$). $\endgroup$ – Faraad Armwood Sep 19 '16 at 22:37
  • $\begingroup$ @Faraad: No, since $S$ is locally given by $x^n=0$, $dx^n$ is the $1$-form that annihilates its tangent spaces. So $\alpha = f dx^n$ for some nowhere-zero smooth function $f$. $\endgroup$ – Ted Shifrin Sep 19 '16 at 22:38
  • $\begingroup$ @TedShifrin: Hmm, I need to read some more then because if anything the sum should be up to $n-1$ since $(x^1,...,x^{n-1},0)$ locally defines $S$. $\endgroup$ – Faraad Armwood Sep 19 '16 at 22:40
  • $\begingroup$ Nope, @Faraad. Think about $1$-forms $\omega$ on $\Bbb R^2$ whose restriction to the $x$-axis is $0$. (Officially, pull back by the inclusion map.) Do they look like $f\,dx$ or $f\,dy$? $\endgroup$ – Ted Shifrin Sep 19 '16 at 22:42

(1) I am still patching together a solution but thanks to @Ted Shifrin, I can clarify your first remark about the expression $\alpha$. Let $\iota: S \to M$ be the inclusion map then $\iota^*: T_pM \to T_pS$. Now take $(U,\phi) = (U, x^1,...,x^n)$ to be a chart on $M$ then $\omega \in \Omega^1(M) \Rightarrow \omega_p = \sum_{i=1}^n a^i(p)\ dx^i$. Consider;

$$\iota^*\omega = \sum_1^n (a^i \circ \iota) \ d(x^i \circ \iota) = \sum_1^{n-1} (a^i \circ \iota) \ d(x^i \circ \iota)$$

i.e if you want a 1-form on $M$ that has your property, you need to dispose of $dx^j$ for $1 \leq j \leq n-1$. Hence you have $\alpha = f dx^n$ where $f$ is non-vanishing.


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.