13
$\begingroup$

Generally, local invertibility does not imply invertibility. However, for differentiable functions from $\mathbb{R}$ to $\mathbb{R}$ then surjectivity and local invertibility do imply invertibility.

As well as being the most obvious, it's also the only (non-contrived) case that I can think of. Are there any more?

Specifically, I'm looking for examples of spaces $X$ which are at least topological spaces (but may be more structured) and subsets of endomorphisms on $X$ for which local invertibility implies invertibility.

$\endgroup$
1
  • 4
    $\begingroup$ I'm having a hard time seeing why continuous (surjective) functions from R to R don't work. Is there a standard counterexample that forces you to restrict to differentiable ones? It seems like a continuous function is locally invertible iff locally monotone iff globally monotone iff globally invertible. And the set-theoretic inverse of an invertible continuous map R-->R is continuous. $\endgroup$ Dec 31, 2009 at 2:25

6 Answers 6

7
$\begingroup$

The "Jacobian Conjecture" is an example of such a statement. It reads :

"If a polynomial $P:\mathbb{C}^N\rightarrow \mathbb{C}^N$ has invertible differential everywhere, then it is globally invertible."

It is open, and considered as difficult.

$\endgroup$
5
$\begingroup$

For a locally invertible holomorphic function on the unitary disk $f:\Delta\rightarrow \mathbb{C}$ we have a result of Becker (1972): if $(1-|z|^2)\left|\frac{zf''(z)}{f'(z)}\right|\leq 1$, then $f$ is invertible. We have also two criteria of Nehari involving the Schwarzian derivative $S(f)$ of $f$: if $|S(f)(z)|\leq\frac{2}{(1-|z|^2)^2}$ or $|S(f)(z)|\leq \frac{\pi^2}{2}$, then $f$ is invertible. A good reference is the book "geometric function theory in one and higher dimensions" by Graham-Kohr.

$\endgroup$
5
$\begingroup$

Following Deane Yang, the answer is a definite yes: the map in question is a global diffeo, provided that (a) it is `locally invertible': i.e. its derivative is everywhere invertible, and (b) the domain and range are compact, simply connected, without boundary.

Proof: the map must be a covering map (``stack of records theorem'' -- see for example Guillemin and Pollack). But a covering map between simply connected spaces is an isomorphism -- here a diffeo.

To make this `non-contrived' take domain and range to be the two-sphere.

$\endgroup$
1
  • $\begingroup$ This follows immediately from my comment to Deane's answer, since properness of the map f follows trivially from compactness of the domain X. The assumptions of compactness of the range Y, simple connectedness of X, boundarylessness of both are completely superfluous. $\endgroup$ Dec 31, 2009 at 9:26
4
$\begingroup$

Another class of topological spaces of which this property holds are trees, and more generally R-trees. http://en.wikipedia.org/wiki/Real_tree

$\endgroup$
3
$\begingroup$

There are a number of different ways to show that a local diffeomorphism $f: X \rightarrow Y$ is a global diffeomorphism:

For example, this follows if $X$ and $f(X)$ are both connected and simply connected. (this appears to be incorrect based on the comments below)

Or if $X$ and $Y$ are both simply connected compact manifolds without boundary.

I think but am not sure that the statement also follows if $X$ and $Y$ are both simply connected manifolds without boundary and the map $f$ is proper.

$\endgroup$
3
  • 1
    $\begingroup$ If f:X--->Y is a local homeomorphism between hausdorff topological spaces, then f is a finite covering if and only f is proper. Hence if moreover Y is simply connected and X connected, a proper local homeomorphism will be a homeomorphism. (No assumption of simple connectedness is needed for X) $\endgroup$ Dec 30, 2009 at 19:27
  • $\begingroup$ The first claim doesn't seem quite right; I think you need (at least) surjectivity of f as well. For instance there are local diffeomorphisms from the interval (0, 2) to itself, with image (0, 1). $\endgroup$
    – macbeth
    Dec 30, 2009 at 19:54
  • 2
    $\begingroup$ In fact, I don't think the first claim is right at all. For instance, there's a surjective, not-injective local diffeomorphism from $\mathbb{R}^2$ to $S^2$: you can wind an infinitely long rubber ribbon around a ball so as to cover it completely. $\endgroup$
    – macbeth
    Dec 30, 2009 at 20:28
2
$\begingroup$

A local isometry between complete connected Riemannian manifolds must be a covering map. So a local isometry between complete connected Riemannian manifolds, with simply-connected range, should be a global isometry?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.