Is It More Precise to Define Manifolds as Topological Vector Spaces?

[TrickyDicky]

I was referring to statements like this where you externalize your misunderstandings. Everyone else is "not so on target" rather than you having made a mistake. It makes it annoying to answer your questions.

I see in your subsequent responses you acted differently. In the past, you often got belligerent; maybe you're not anymore.

Nobody makes you answer my questions, if you feel so annoyed by that just don't do it, but please don't lie saying that I jump down anybody's throat, there is a stretch between that and suggesting an answer might be slightly off target wrt what I was referring to, however wrong I may be. More so when micromass had already assertively pointed me to where my confusion was.
And your tone is so patronizing, please quit it.

Having said that I have always praised you as an expert in differential geometry in these forums so I encourage you to keep helping people around here.

 

[lavinia]

In the theory of relativity there are local coordinate systems where the observer feels that he is in a Euclidean domain. These are so called free float coordinates. Here the observer can imagine that he can extend his local world beyond the confines of his measuring instruments to a vector space. I think these coordinates are in some sense canonical.

On a general manifold there are no canonical coordinates but on a Riemannian manifold one always has Gaussian polar coordinates and on manifolds with different structures e.g. Riemann surfaces( conformal coordinates) one has other natural coordinates.

 

[micromass]

So what?. As micromass noticed I was actually demanding smooth charts, can you have a smooth chart on the point of the cone?

Yes. You can give a cone a smooth structure. But you can't make it an immersed submanifold of Euclidean space.

 

[TrickyDicky]

Yes. You can give a cone a smooth structure. But you can't make it an immersed submanifold of Euclidean space.

Ok, so I admit then I don't have a clue how "You can have a chart on the point of a cone" contradicts anything I wrote in my post. Unless Ben was thinking only about cones in R^3, but that defeats the definition of manifold as an intrinsically defined object.

 

[TrickyDicky]

In the theory of relativity there are local coordinate systems where the observer feels that he is in a Euclidean domain. These are so called free float coordinates. Here the observer can imagine that he can extend his local world beyond the confines of his measuring instruments to a vector space. I think these coordinates are in some sense canonical.

On a general manifold there are no canonical coordinates but on a Riemannian manifold one always has Gaussian polar coordinates and on manifolds with different structures e.g. Riemann surfaces( conformal coordinates) one has other natural coordinates.

More specifically in general relativity the domain would be Minkowskian rather than Euclidean. That is basically the content of the Equivalence principle.
Of course in the presence of a (pseudo)Riemannian metric you may have those kinds of natural local coordinates :geodesic (Fermi) normal coordinates, once you have these it is easy to derive polar coordinates, but I guess they rely on the Riemannian metric.

 

[Ben Niehoff]

Tricky, I'm sorry I snapped at you. I've gotten annoyed with you in the past, but I think my previous impression of you is wrong.

As for charts at the point of the cone: You can define polar coordinates that are centered at the point, and therefore cover the neighborhood of the point in a single patch. You can always scale these coordinates so that the point of the cone is homeomorphic to R^n...for a 2d cone, imagine simply projecting down into a plane to give you the homeomorphism.

As Micro points out, you can also use this projection into a flat plane to define a smooth structure at the point, but then you don't really have a conical point anymore.

 

[TrickyDicky]

Tricky, I'm sorry I snapped at you. I've gotten annoyed with you in the past, but I think my previous impression of you is wrong.

As for charts at the point of the cone: You can define polar coordinates that are centered at the point, and therefore cover the neighborhood of the point in a single patch. You can always scale these coordinates so that the point of the cone is homeomorphic to R^n...for a 2d cone, imagine simply projecting down into a plane to give you the homeomorphism.

As Micro points out, you can also use this projection into a flat plane to define a smooth structure at the point, but then you don't really have a conical point anymore.

Thanks Ben, no worries, I probably overreacted a bit.

 

[lavinia]

One general principal that this discussion of the cone illustrates is that geometry is a structure that is added onto a topological space and a topological space can be given many geometries.

Another is that a smooth manifold may be embedded non-smoothly in another manifold. The cone is a non- differentiable embedding of a disk in three space.