3
$\begingroup$

For any $C^{\infty}$ manifold $M$, the tangent bundle $TM$ of $M$ is also a $C^{\infty}$ manifold.

Hence we can think about the differential $df:TM\rightarrow TN$ of maps $f:M\rightarrow N$ between smooth manifolds as a smooth map between two smooth manifolds $TM$ and $TN$.

When I come up with this, I'm totally convinced that for any (smooth) embedding $\iota:N\rightarrow M$ between smooth manifolds, the differential $d\iota:TN \rightarrow TM$ also becomes an embedding (It must be in common sense; embedding is subobject!), but I cannot prove this in detail.

In particular, I'm stuck on thinking about the differential of the differential $d(d\iota)$. It's quite hard to imagine for me.

Is this true? In that case, I want to see a detailed proof for this one.

$\endgroup$
  • 2
    $\begingroup$ Can you prove this when $\iota$ is the embedding of $\mathbb{R}^n$ into $\mathbb{R}^m$, with $n\leq m$? That's the best place to start, I think. $\endgroup$ – froggie Dec 4 '12 at 16:12
  • 1
    $\begingroup$ to write down $d(d\iota)$ explicitely at some tangent vector $X\in TM$, you only need to write down $d\iota$ in local oordinates and differentiate it, try it for yourself, it's not hard. $\endgroup$ – Olivier Bégassat Dec 4 '12 at 16:18
  • 1
    $\begingroup$ I think it is possible to get around explicit calculations. It is clear that $d\iota$ is bijective, has full rank along the zero section and is linear restricted to fibers. Maybe with the help of local trivializations it should become clear $\endgroup$ – wspin Dec 4 '12 at 16:45
  • $\begingroup$ Thanks everyone, in particular froggie. I can use the result on $\mathbb{R}^n$ with the normal chart for embedded manifold. $\endgroup$ – cjackal Dec 4 '12 at 17:21
  • $\begingroup$ The result is also true when $i:N\to M$ is merely an immersion, see my answer in math.stackexchange.com/questions/3233772/… $\endgroup$ – Zero May 26 at 0:17
0
$\begingroup$

Take normal charts $\left\{ (U_\alpha , \psi_\alpha)\right\}$ for $N$, that is, charts of $M$ s.t $\psi_\alpha \circ \iota \circ(\psi_\alpha \mid_{U_\alpha \cap N})^{-1}(a^1,\cdots , a^n)=(a^1,\cdots,a^n,0,\cdots,0)$.

($\psi_\alpha \mid_{U_\alpha\cap N}:U_\alpha\cap N \rightarrow \mathbb{R}^n$ : restriction onto first $n$ coordinates)

Then $(\hat{U_\alpha},\hat{\psi_\alpha})$ is a chart for $TM$, where $\hat{U_\alpha}=\pi^{-1} (U_\alpha), \hat{\psi_\alpha}(p,X)=(\psi_\alpha (p),d\psi_\alpha^1(X),\cdots,d\psi_\alpha^m(X))$.

($\pi:TM\rightarrow M$:natural projection)

Same for $TN$ : $(\widehat{U_\alpha\cap N}, \widehat{\psi_\alpha \mid_{U_\alpha \cap N}})$ is a chart for $TN$.

Identify $\mathbb{R}^n \cong \mathbb{R}^n \times \left\{0,0,\cdots,0\right\}\subset \mathbb{R}^m$, hence tangent vector $X$ at $p\in T_p N$ looks locally like $\sum_{i=1}^n a^i \frac{\partial}{\partial x^i}$ where $x^i = (\psi_\alpha\mid_{U_\alpha \cap N})^i=\psi_\alpha^i$

Since $d\iota (p,X)=(p,d\iota_p X)$, $\hat{\psi_\alpha}\circ d\iota (p,X)=(\psi_\alpha(p), d\psi_\alpha^1 (d\iota_p X),\cdots, s\psi_\alpha^m (d\iota_p X))=(\psi_\alpha(p),a^1,\cdots,a^n,0,\cdots ,0).$

$\therefore \hat{\psi_\alpha}\circ d\iota \circ \widehat{\psi_\alpha \mid_{U_\alpha\cap N}}^{-1}(p^1,\cdots, p^n,a^1,\cdots , a^n)=(p^1,\cdots,p^n,0,\cdots,0,a^1,\cdots,a^n,0,\cdots,0)$ : embedding.

$\endgroup$

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.