In the Lee‘s book there is a proposition stating:

If $M$ is a smooth $n$-manifold with or without boundary, and $M$ can be covered by a single smooth chart, then $TM$ is diffeomorphic to $M\times \mathbb{R}^n$.

This covers simple examples when $M$ is an open subset of $\mathbb R^n$. However, closed ball and closed annulus in $\mathbb R^n$ are not contained in the above statement.

My question is do we have (for any $n$)

  • $TB_r \cong B_r\times \mathbb R^n$
  • $TC(r_1,r_2) \cong C(r_1,r_2)\times \mathbb R^n$,

where $B_r:=\{x\in\mathbb R^n :|x|\leq r\}$ is closed ball, and $C(r_1,r_2):=\{x\in\mathbb R^n : r_1\leq|x|\leq r_2\}$ closed annulus.

By the stereographic projection, the problem with closed ball should be equivalent to one with closed hemisphere.


Look at the hypotheses: the closed ball IS a smooth $n$-manifold with boundary, covered by a single chart. The answer in each case you asked about is "yes". In fact, the map you get is not just a diffeomorphism, but a bundle map. Once you see this,

The annular case follows from the ball case, since if you have a map $$ F: TB(r_2) \to B_r^2 \times R^n $$ such that for $v \in TB(r_2)$, we have $$ \pi(v) = \pi_1 (F(v)) $$ where $\pi$ is the projection to the base space on the left, and $\pi_1$ is projection to the first factor on the right (i.e., it's a bundle map), you can simply restrict the domain of $F$ to $TC(r_1, r_2) \subset TB(r_2))$ to get a trivialization of $TC$.

  • 1
    $\begingroup$ One thing is still not clear to me. I know that the closed ball is a smooth $n$-manifold, but I still missing why it can be covered by a single chart. $\endgroup$ Mar 15 '15 at 12:19
  • $\begingroup$ I think of definition that every point needs to have a neighbourhood diffeomorphic to a (relatively) open subset in $\mathbb H^n:=\mathbb R^n\cap\{x_n\geq 0\}$, so I know how to do it using $n$ charts. $\endgroup$ Mar 15 '15 at 12:27
  • $\begingroup$ The single chart is the inclusion map: you send $x \in B$ to $x \in \mathbb R^n$. Look at the north pole $(0, \ldots 0, 0,1)$. The artic (points $(x_1, x_2, \ldots)$ with $x_n > 1/2$, say), is diffeomorphic to a relatively open subset of $H^n$ via the map $x \mapsto x + u(x)$, where $u(x_1, x_2, \ldots) = -1 + \sqrt{1 - (x^2_1 + x^2_3 + x^2_{n-1})}$. Under this map, the surface of the artic maps to the $x_n = 0$ plane, while the rest of the arctic maps to a relatively open subset of $\mathbb R^n \cap \{ x_1 \le 0\}$. You can remap that to $H^n$ by negation. :) $\endgroup$ Mar 15 '15 at 12:47

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.