3

Let $T^m = \mathbb{R}^m / \mathbb{Z}^m$ be the $m$-dimensional torus equipped with quotient topology, so the canonical surjection $\pi: \mathbb{R}^m \to T^m$ is a covering map.

  1. Show that there is an atlas on $T^m$ such that $\pi$ is a smooth immersion.
  2. Show that there are unique smooth vector fields $\xi_1, \dotsc, \xi_m$ on $T^m$ such that for every $p \in \mathbb{R}^m$,
  3. $$d\pi(p)\left( \frac{\partial}{\partial x_i} (p) \right) = \xi_i(\pi(p)).$$
  4. Let $X_1, \dotsc, X_m$ be commuting vector fields on a compact $m$-manifold $M$, and assume that $X_1(p), \dotsc, X_m(p)$ are linearly independent in $T_p(M)$ for every $p\in M$. Show that there is a diffeomorphism $f:M\to T^m=\mathbb{R}^m/\mathbb{Z}^m$ such that for each $i$,

    $$f_*(X_i)= \sum_j a_j \xi_j,$$

    where $a_j$ is a constant function on $T^m$.

My attempt:

Anyway, for #1, given $p\in T^m$ I would like to choose a chart $(U, \varphi)$ which is a ball of radius $1/2$. Then we can construct a diffeomorphism to some subset of the cube $[-1,2]^m$, by just taking the component of $\pi^{-1}(B)$ which contains the preimage of $p$ in $[0,1]^m$. Then $\pi$ is a local diffeomorphism (the transition maps are just translation), so it is a smooth immersion.

For #2, we must show that $d\pi$ agrees on fibers of $\pi$. But about the points in $\pi^{-1}(q)$ there are disjoint neighborhoods (components of the inverse image of a small neighborhood of $q$) on which $\pi$ acts identically. Therefore the function has the same derivative at these points.

For #3, we know that each point $p$ has a neighborhood $U$ in $M$ such that there is a diffeomorphism $f:U\to \mathbb{R}^m$ such that $f_*(X_i) = \frac{\partial}{\partial x^i}$. Now my idea is to show that the compactness of $M$ allows us to glue the little $f$'s together on all of $M$. Is this the right path?

Please let me know if anything is unclear, or if I should explain certain points more.

Eric Auld
  • 28,127
  • Eric, as it stands, 2. doesn't make sense. Your error in thinking 3. is wrong is that you're forgetting $n$ commuting vector fields. That is an exceedingly rare condition :) – Ted Shifrin Nov 02 '13 at 02:48
  • @TedShifrin Sorry, Ted, in number 2 I had made a transcription error. I'm now writing my attempt for #3, perhaps you can look at it when you get a chance – Eric Auld Nov 02 '13 at 02:54
  • @TedShifrin Edited. – Eric Auld Nov 02 '13 at 03:06
  • You'd better quote the big theorem that gives those coordinates in 3. You might want to consider the flows, but at the moment I don't see how to avoid the dense line on the torus. ... Who's teaching your course? – Ted Shifrin Nov 02 '13 at 03:16
  • @TedShifrin Dr. Bruce Kleiner – Eric Auld Nov 02 '13 at 03:27
  • Ha! Haven't seen him in years, but say hi. Same goes for Sylvain Cappell, if you're learning alg top. I'm now suspicious about this problem. Who says the integral curves of the commuting vector fields have to all be closed curves? Hmm ... I'll ponder more in the morning. – Ted Shifrin Nov 02 '13 at 03:30
  • @TedShifrin Thanks for your help. Luckily I have a few days. I love Dr. Cappell...I'll pass that along. – Eric Auld Nov 02 '13 at 03:38

0 Answers0