0
$\begingroup$

This question has been crossposted from MSE since there it received no attention. Please notify me if questions like these are not appropriate for this platform.

The question

Let $ M $ be a smooth manifold. Let $ \mathscr C_M^\infty $ be the sheaf of $ \mathscr C^\infty $-functions on $ M $, and let $ \mathscr X $ be the $ \mathscr C_M^\infty $-module of vector fields on $ M $, i.e. $$ \mathscr X(U) = \Gamma(U,\mathrm TM) = \{\text{sections $ X\colon U\to \mathrm TM $}\} $$ for any $ U\subset M $ open, where of course $ \mathrm TM $ is the tangent bundle of $ M $.

The dual $ \mathscr C_M^\infty $-module $ \mathscr X^* $ of $ \mathscr X $ is defined by $$ \mathscr X^*(U) = \underline\hom_{\mathscr C_M^\infty}(\mathscr X,\mathscr C_M^\infty)(U) = \{\text{morphisms of $ \mathscr C_M^\infty $-modules $ \mathscr X{\restriction_U}\to \mathscr C_U^\infty $}\} $$ for any $ U\subset M $ open, where $ \mathscr X{\restriction_U} $ is the restriction of the sheaf $ \mathscr X $ to the open subsets of $ U $.

I then took $ U = M $ and I tried to show that $$ \mathscr X^*(M) \cong \mathscr X(M)^* $$ where $ \mathscr X(M)^* $ is the dual $ \mathscr C_M^\infty(M) $-module of the $ \mathscr C_M^\infty(M) $-module $ \mathscr X(M) $, but I didn't succeeded.

An obvious map $$ \mathscr X^*(M) \rightarrow \mathscr X(M)^* $$ is the one that takes $ \omega = (\omega_U)_{\text{$ U\subset M $ open}}\in \mathscr X^*(M) $ to $ \omega_M\in \mathscr X(M)^* $.

I thought that to define a map in the opposite direction $$ \mathscr X^*(M) \leftarrow \mathscr X(M)^* $$ one could start with an $ \omega\colon \mathscr X(M)\to \mathscr C_M^\infty(M) $ and define $$ \omega_U(X) = \omega(\tilde X) $$ for all $ X\in \mathscr X(U) $, where $ \tilde X $ is something like an "extension" of $ X\colon U\to \mathrm TM $ to all of $ M $. I think that such a $ \tilde X $ could be defined using partitions of unity, but I'm not sure about that so I'm asking here for help.

$\endgroup$
3
  • 2
    $\begingroup$ The "extension" you're looking for exists, because sheaves of sections of smooth vector bundles on smooth manifolds are flabby (they even are soft). Flabby means that the restriction morphisms are surjective. $\endgroup$
    – DamienC
    Commented Jan 11 at 17:23
  • 1
    $\begingroup$ @DamienC: Flabby means the restriction morphism F(V)→F(U) is surjective for any inclusion of opens U⊂V. If F is the sheaf of sections of the trivial vector bundle, it is easy to construct examples where the restriction map is not surjective, e.g., the smooth section x↦1/x over (-∞,0)∪(0,∞) does not extend to the real line. $\endgroup$ Commented Jan 11 at 20:10
  • 2
    $\begingroup$ Dimitri: you are right (and I'm totally wrong). Sheaves of sections are soft but not flabby. $\endgroup$
    – DamienC
    Commented Jan 11 at 23:48

1 Answer 1

2
$\begingroup$

A smooth section $X$ of a vector bundle (e.g., a smooth vector field) over an open subset $U⊂M$ can be presented in the form $X=g Y$, where $g$ is a smooth function on $U$ and $Y$ is a smooth section over $U$ that extends to a smooth section $Z$ over $M$.

Using such a presentation, it is easy to define $ω_U$: $$ω_U(X) = ω_U(g Y) = g ω_U(Y) = g ω(Z).$$

To show the existence of a pair $(g,Y)$, use a smooth partition of unity to pick a smooth metric on the vector bundle and also pick a smooth function $h$ on $M$ that vanishes on $M∖U$ and is strictly positive elsewhere. Set $$g(x)=\exp(‖X(x)‖^2) h(x)^{-1}.$$ Now the section $Y=g^{-1}X$ is smooth and extends by zero to a smooth section $Z$ over $M$.

$\endgroup$
2
  • $\begingroup$ I was expecting something more "canonical", but yeah glad to know things go like this. Why there's the need for the factor $ \exp(\lVert X(x) \rVert^2) $? It is to guarantee that the extension $ Z $ exits? (I tried to prove it, but I'm not very comfortable with these objects). $\endgroup$ Commented Jan 14 at 0:03
  • $\begingroup$ @GeometriaDifferenziale: Dividing by exp(…) makes the resulting section smooth and bounded in norm. Then multiplying by h makes the resulting section extendable by zero to a section over M. If we did not divide by exp(…), the resulting quotient could go to infinity as we approach the boundary of U. $\endgroup$ Commented Jan 14 at 4:45

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .