5
$\begingroup$

I will link the following lecture notes, because it makes no sense to keep pasting from them.

When reading them, there are two things I do not understand. The author introduces smooth manifolds by defining charts (which in a first stage are just bijections with open images, Definition 2.2.1) and then shows how these charts define a topology on the manifolds (Proposition 2.2.5). He then shows that W.r.t. this topology the charts (Proposition 2.2.6).

But much later, he actually needs that charts are diffeomorphisms, or at least smooth - e.g., he defines on pp. 34 tangent maps $T_p f$ only for smooth maps $f$ and on the next page considers $T_p \psi $ for a chart $\psi$, so $\psi$ should be smooth as well?

So is a chart smooth, or a diffeomorphism? If not, the lecture notes contain errors, should the requirements of the definitions be relaxed?

$\endgroup$

3 Answers 3

4
$\begingroup$

In the definition, it is only required that overlap maps be smooth, as said in the answer of @AndresMejia.

But, here also is an answer to your question about whether a chart $\psi$ itself is smooth. Once smooth manifolds have been defined, and once smooth maps between smooth manifolds have been defined, it is then very easy exercise to prove that for each chart $\psi : U \to \mathbb R^n$ (where $U$ is some open subset of the manifold), the map $\psi$ is indeed smooth.

$\endgroup$
1
$\begingroup$

continuous charts indeed define a topology on a manifold,

but if you want a smooth manifold, then transition maps $\psi_{u,v}=\psi_u\psi_v^{-1}:\mathbb R^n \to \mathbb R^n$ are required to be diffeomorphisms.

$\endgroup$
2
  • $\begingroup$ so are you saying that the way he defined charts in 2.2.1 doesn't give a smooth manifold? $\endgroup$
    – user38525
    Aug 15, 2018 at 18:42
  • $\begingroup$ No, it does. He requires that transition maps are diffeomorphisms. I edited my answer to make this clearer. $\endgroup$ Aug 15, 2018 at 18:45
1
$\begingroup$

Ok, so a priori charts are only required to be homeomorphisms, with the transition maps $u \circ v^{-1}$ smooth from $u(U \cap V) \to v(U \cap V)$. The idea behind a smooth atlas however is that it induces a smooth structure, which makes the charts smooth by fiat. This isn't hard to show: our definition of a smooth map $f:M \to \mathbb{R}^n$ is that $f \circ v^{-1}$ be smooth for all charts $v: V \to \mathbb{R}^n$ for which the composition is defined. But when we replace $f$ by some chart $u: U \to \mathbb{R}^n$, this is exactly the smoothness criterion for transition maps we mentioned above! So a chart, in addition to being a homeomorphism, is also a diffeomorphism

$\endgroup$

You must log in to answer this question.

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