It is trivial to prove that the integration of a $(n-1)$ exact form on the boundary of a $n$-manifold is 0. What about the contraposative ? If the integration of a $(n-1)$-form on the boundary of a $n$-manifold is 0, is this form exact ? If not, are there particular conditions to satisfy for this to be the case ?

For example, in the case of $S^1$, any $1$-form can be written as $f(\theta)d\theta=c d\theta+dg(\theta)$, $c$ being the integral around $S^1$, and $g$ a differentiable function on $S^1$. Now in this case if the integral is 0, it implies that the form is exact. I wondered in the general case if such decomposition is always possible, because if so, it can be proven that the integral being 0 implies that the form is exact. Is this correct ?

  • 1
    $\begingroup$ Won't any closed $(n-1)$-form on an $n$-manifold integrate to 0 on the boundary of that manifold? $\endgroup$ – Julian Rosen Jul 20 '13 at 0:42
  • $\begingroup$ The question did not hypothesize it being closed, I think. $\endgroup$ – Eric Auld Jul 20 '13 at 2:02
  • $\begingroup$ Since beign closed is a necessary condition to be exact in any case, maybe I could rephrase my question as "If the integration of a (n−1)-form closed form on the boundary of a n-manifold is 0, is this form exact ?" $\endgroup$ – vkubicki Jul 20 '13 at 8:57
  • $\begingroup$ If the form is closed and the domain is contractible, that implies that the integral is 0 (by the Poincarré lemma), but I think I am looking at the implication in the other direction, from the value of the integral to the characterization of the form. $\endgroup$ – vkubicki Jul 20 '13 at 9:20

I think the result you are looking for is this: Top deRham cohomology group of a compact orientable manifold is 1-dimensional, with the isomorphism $H^n_{dR}(M^n)\to \mathbb R$ given by $\omega\mapsto \int_M \omega$. (One should add "connected" to the assumptions.) In your question, $n$ is actually $n-1$.

The proof of $H^n_{dR}(M^n)\approx \mathbb R$ is not hard but lengthy: see section 8.1, titled de Rham cohomology in the top dimension, of Nigel Hitchin's notes.

See also Poincaré Duality with de Rham Cohomology. And for the non-compact case, De Rham cohomology for non-compact manifolds.

  • 1
    $\begingroup$ Thank you for your answer. May I ask a complementary question ? In the case of a non-top cohomology group, say for the torus as in topospaces.subwiki.org/wiki/Torus , is it possible to link the rank $\binom {n} {k}$ to the values of various integrals on the domain ? $\endgroup$ – vkubicki Jul 27 '13 at 10:36
  • $\begingroup$ @vkubicki Interesting, but a new question should be posted as separate question, not as a comment.. $\endgroup$ – 40 votes Jul 27 '13 at 14:41
  • $\begingroup$ You are right, the scope is a bit larger than a comment. So there it is, waiting at math.stackexchange.com/questions/453644/… . $\endgroup$ – vkubicki Jul 27 '13 at 19:29

Stokes' theorem works both ways, but you have to be careful. In particular, a $n$-form $\omega$ is closed if and only if its integral over every $n$-boundaries is zero, not just one particular!

For every $(n+1)$-dimensional region $\Omega$ we have: $$ \int_{\partial\Omega}\omega=\int_\Omega d\omega. $$

If the expression on the left is zero for every $\Omega$, so is the one on the right, and it means that $d\omega=0$.

If we want $\omega$ to be exact, then we don't have to limit ourselves to $n$-boundaries, but we want its integral to be zero on every closed region! That is, any region $A$ with $\partial A=0$.

This because what cause a closed form not to be exact are the very "holes".

I don't know if from this requirement you can explicitly find that $\omega=d\eta$ for some $\eta$ (that is, I don't know if this reasoning provides you such an $\eta$).

What is immediate to check (just use Stokes), though, is that exact forms yield a zero integral over general closed regions, not just boundaries.


No, and my limited knowledge of the topic is that it depends heavily on the topology of a space. For instance, in a multiply connected 2-manifold, there are closed forms which are not exact (they are not the boundary of any submanifold, since they "go around they hole").

  • $\begingroup$ OK. I edited my question regarding the impact of the holes in a manifold. $\endgroup$ – vkubicki Jul 20 '13 at 9:19

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.