I am trying to get started with differential geometry, and am having a difficult time wrapping my head around the concept of a manifold. One thing that would make it easier to understand would be if authors would give more examples of things that aren't manifolds. Anyway, here's one potential example that I came up with that I think will help me along quite a bit if I can understand it.

So, let's say we're in $\Bbb R^3$ and we have the unit sphere $x^2+y^2+z^2=1$. Even I can tell that this is a manifold. But now let's take the disc $z=0, x^2+y^2<2$, cut a hole out of it at $x^2+y^2<1$ and "attach" it to our sphere. Now, is the resulting object a manifold? Why or why not? What happens if we take our disc again and cut out the hole $x^2+y^2 \leq 1$?

I can't see how this object violates any explicit part of the definition of a manifold, but it just doesn't seem right.

Thanks for helping me get some sleep again.

  • 1
    $\begingroup$ So you're looking at the sphere with like a ring around it thats touching it? This won't be locally homeomorphic to $\mathbb R^2$. $\endgroup$ – Eric O. Korman Aug 14 '14 at 0:16
  • $\begingroup$ Yep, that's the idea. $\endgroup$ – bob.sacamento Aug 14 '14 at 0:16

This is a great question. You don't say whether you're thinking of topological manifolds or of smooth manifolds, but it doesn't really matter.

Let's reduce the dimension by $1$, and consider the unit circle $x^2+y^2=1$, together with a whisker, $y=0$, $x\ge 1$. Does the point $(1,0)$ have a neighborhood that is homeomorphic (diffeomorphic) to an interval $I=\{|t|<\epsilon\}\subset\Bbb R$? Suppose you had such a homemorphism, and suppose it sends $(1,0)$ to $0\in I$. Can you see an elementary reason that this is impossible?

The point-set topology won't quite work in higher dimensions, and you need to use a bit more. But in the world of smooth manifolds, it's easy to see things won't work, using linear algebra.

  • 1
    $\begingroup$ Nice answer, @Ted Shifrin. Could you please elaborate on the use of "linear algebra" in higher dimensions? Do you mean that the tangent space wouldn't have the right structure in a space such as the OP's example, or something like that? $\endgroup$ – Phillip Andreae Aug 14 '14 at 2:35
  • 1
    $\begingroup$ @PhillipAndreae, yup. The tangent space must be a (linear) subspace, and yet it has (by continuity of the derivative) too many dimensions' worth of vectors in it. $\endgroup$ – Ted Shifrin Aug 14 '14 at 3:18
  • $\begingroup$ @Ted Shifrin Thanks for the answer, and the positive feedback. Now, to your question, my messed up thinking is that there wouldn't be a homeomorphism between $I$ and our "circle+whisker" because the point where the two objects connect would have to map to two different coordinates on $I$. But then I think: we're supposed to be able to cover the object with overlapping "coordinate patches", and that seems to me -- incorrectly, no doubt -- to offer a way out. $\endgroup$ – bob.sacamento Aug 14 '14 at 17:04
  • $\begingroup$ No, if we had a parametrization there'd just be a single point (which I assumed was $0$) mapping to the junction point. Of course, realistically, this cannot happen, as the elementary proof shows. (Pull out $0$ from the interval and you have two pieces. Pull out the junction point from a neighborhood in the "circle+whisker" and you'll have three pieces.) $\endgroup$ – Ted Shifrin Aug 14 '14 at 17:58

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.