4

Let $(M,\mathcal{A})$ be a manifold with smooth structure $\mathcal{A}$. For any point $x\in M$, we define a tangent at x by the triplet $(c,x,h)$, where $c=(U,\phi)$ is a chart at $x$, $h\in R^n$ ($n$ is the dimension of the manifold). For two charts $c,c'$, define the following equivalence relation: $(c_1,x,h_1)\sim (c_2,x,h_2)$ if $D(\psi\circ\phi^{-1})(\phi(x))h_1=h_2$, where $c_1=(U,\phi), c_2=(V,\psi)$. Now, the tangent vectors at $x$ is defined by $T_xM=\{[c,x,h]: x\in M, h\in R^n\}$ and $[c,x,h]$ is the equivalence class. Finally, the tangent space is defined as $TM=\cup_{x\in M} T_xM$. My question is the following:

Can we show that $TM$ is diffeomorphic to $M\times R^n$ using this definition of tangent space? In particular, is it true when $M=\mathbb{S}^{n-1}$?

Ribhu
  • 762
  • 1
    $TM$ is rather the tangent bundle of $M$, and $T_xM$ is a tangent space. Tangent bundles are generally not trivial, i.e. $TM$ is generally not diffeomorphic to $M\times\Bbb R^n$. See https://math.stackexchange.com/questions/893519/tangent-bundle-of-s2-not-diffeomorphic-to-s2-times-mathbbr2. – edm Oct 03 '18 at 06:57
  • Thanks! Actually I have to show that $TS^{n-1}$ is diffeomorphic to ${(x,y): \ x\in S^{n-1}, \ y\perp x}$. Is it easy to show using the definition of tangent bundle given here? – Ribhu Oct 03 '18 at 15:18

1 Answers1

2

In general this is not true. One can show that a vector bundle $(E, \pi,M)$(https://en.wikipedia.org/wiki/Vector_bundle) is trivial (i.e. $E \simeq M \times \mathbb{R}^k$) iff exists a global frame $\lbrace \sigma_j : M \rightarrow E \rbrace_{j=1,\ldots , k} $ (i.e. a family of sections of $\pi$) which never vanish on $M$ and form a pointwise base of $E_p\simeq \pi^{-1}(p)$ for all $p\in M$. The tangent bundle is an example of vector bundle. You can use this criterion for checking whether a vector bundle is parallelizable (that is $TM \simeq M \times \mathbb{R}^m$).

As for the spheres it has been shown that only $S^1$, $S^3$ and $S^7$ are parallelizable. For $S^1 \hookrightarrow \mathbb{R}^2$ you just take the vector field $x_2 \partial_1- x_1 \partial_2$. For $S^3$ you note that it is a Lie group (quaternions with norm $1$), then Lie groups are always parallelizable with $TG\simeq G\times \mathfrak{g}$ with $\mathfrak{g} \simeq T_eG$ the Lie algebra. The fact that $S^7$ is parallelizable by a result of Milnor.

Baol
  • 642
  • 1
    The fact that even-dimensional spheres $S^{2m}$, $m > 0$, are not parallelizable follows from the Hairy Ball Theorem, which says those spheres don't admit even a single nonvanishing vector field, let alone a global frame of them. – Travis Willse Oct 03 '18 at 13:45