home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!snorkelwacker.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!tycchow
- From: tycchow@athena.mit.edu (Timothy Y Chow)
- Subject: Why Atlases? (was: Re: What is a manifold?)
- Message-ID: <1992Nov5.205854.1485@athena.mit.edu>
- Sender: news@athena.mit.edu (News system)
- Nntp-Posting-Host: alfredo.mit.edu
- Organization: None. This saves me from writing a disclaimer.
- Date: Thu, 5 Nov 1992 20:58:54 GMT
- Lines: 42
-
- Vaughan Pratt asks why the complicated notion of atlas is essential,
- since it seems to involve just as much arbtirariness as a retract
- definition.
-
- In my previous articles I glossed over the distinctions between
- topological and smooth manifolds. In the case of topological manifolds
- I think the embedding-independent definition is well-motivated. (Feel
- free to disagree, of course.) There isn't any of the arbitrariness of
- the choice of open charts that there is in the case of smooth manifolds:
- we just require that at EVERY point there is SOME neighborhood that's
- homeomorphic to R^n.
-
- In the case of differentiable manifolds, however, there is a certain
- arbitrariness in choosing an atlas. The arbitrariness can be removed by
- requiring the set of charts to be maximal, but this doesn't really answer
- the question of why this definition is superior to the retract definition
- that Vaughan Pratt suggests.
-
- I think the point is that a lot of basic theorems about smooth manifolds
- require you to work with coordinate neighborhoods of points. In order
- to apply the inverse function theorem or other techniques from calculus,
- you need to identify a small portion of the manifold with an open set in
- Euclidean space. Since you need the existence of such things all the
- time, you might as well build it into the definition. Even with the
- retract definition, at some point you're going to have to prove that
- you can take coordinate neighborhoods.
-
- To some extent it is just a matter of taste whether you begin with the
- existence of global tubular neighborhoods and prove the existence of
- local coordinate neighborhoods or proceed the other way around. In
- favor of the standard approach I think it can be said that it is more
- in keeping with the approach in modern mathematics of defining objects
- as intrinsically as possible, i.e., without having to create a habitat
- for them to live in before allowing them to exist, but focusing on the
- object itself as much as possible. IMHO this is a more natural way to
- define mathematical objects.
-
- Incidentally, another way to avoid the apparent arbitrariness of the
- choice of charts for the atlas is to define a smooth manifold by
- specifying, instead of these little overlapping charts, the sheaf of
- differentiable functions on the manifold. This is more in the spirit
- of scheme theory, but I think we're starting to digress here.
-