11
$\begingroup$

Some mathematical structures are visualized very well. I imagine how a shapeless bunch of points (a set; the only property of which is quantity) is collected in one or another soft form (topological space), depending on what topology is set on it. I imagine how this soft (floating) shape becomes hard when I set the metric. The first examples of deformation retractions give a vivid geometric intuition of the homotopy type as a "(homotopy) framework". Thus, the concepts of homotopy type, topology, and metrics have a clear physical content. In contrast to these three examples, for the concept of a smooth structure on a manifold, everything is not so clear due to the fact that manifolds up to dimension 3 have a unique smooth structure

Do you have any intuition for a smooth structure on a manifold? Do you see any meaning/physical content in it? Do visually homeomorphic non-diffeomorphic spaces differ for you?

After a smooth manifold is provided with additional structures (Riemannian, symplectic, etc.), the meaning of these objects becomes clear. But a smooth manifold devoid of additional structures is still very mysterious for me.

P.S. I'm not sure if this question is for a forum (on the other hand, the system showed me many questions with similar titles). If so, then feel free to close it.

$\endgroup$
15
  • 2
    $\begingroup$ Is it that different from "visualizing" a function (of one real variable, say) which is continuous but not differentiable? $\endgroup$ Mar 22, 2022 at 22:23
  • 1
    $\begingroup$ Probably yes, because the smoothness of a function is a property, and "the smoothness of a manifold" is an additional structure. At least continuous non-smooth functions themselves are easy to visualize, for example $|x|$. $\endgroup$ Mar 22, 2022 at 22:26
  • 3
    $\begingroup$ I think the question is not about the distinction between manifolds that have or don't have smooth structures, or between maps of manifolds that are or aren't smooth, but given a manifold that does have at least one smooth structure, what do the different ones mean. A related question is "what's the simplest example of two distinct smooth structures on a given topological manifold ?" $\endgroup$ Mar 22, 2022 at 22:30
  • 2
    $\begingroup$ @MaximeRamzi: maybe the two questions are not so separate, since whenever you have two smooth manifolds that are homeomorphic but not diffeomorphic, any homeomorphism gives a continuous, non-differentiable function. $\endgroup$ Mar 22, 2022 at 22:41
  • 5
    $\begingroup$ As for the "simplest example", there are at least two candidates: 1. $3\mathbb{CP}^2\#20\overline{\mathbb{CP}}\vphantom{C}^2$ and $K3\#\overline{\mathbb{CP}}\vphantom{C}^2$; 2. $S^7$ and $\{v^2+w^2+x^2+y^3+z^7 = 0\} \cap S^9 \subset \mathbb{C}^5$. $\endgroup$ Mar 22, 2022 at 22:45

1 Answer 1

2
$\begingroup$

$\newcommand{\R}{\mathbb{R}}$The purpose of a manifold structure is to be able to differentiate functions. And initially we know how to differentiate functions only if the domain is an open set in $\R$. Moreover, once you know how to differentiate functions, this opens the door to doing all the other stuff you do with manifolds.

Here's how I think of it: Initially, think of a manifold as just a set $M$ and nothing else. Think of a coordinate map as a bijection $\phi: O \rightarrow \R^n$, where $O\subset M$ and $\phi(O)$ is an open subset of $\R^n$. An atlas is a collection of coordinate maps such that the domains of the coordinate maps cover $M$. No assumptions on topology or differentiability yet.

A topological manifold is $M$ with an atlas $\mathcal{A}$ such that for any two coordinate maps $\phi_1: O_1 \rightarrow \R^n$ and $\phi_2: O_2 \rightarrow \R^n$ such that $O_1\cap O_2\ne \emptyset$, then the map $\phi_2\circ\phi_1^{-1}: \phi_1(O_1\cap O_2) \rightarrow \phi_2(O_1\cap O_2)$ is a homeomorphism. Notice that such an atlas immediately defines a topology on $M$ where the domains of the coordinate maps in $\mathcal{A}$ form a base of open sets. Assume that this topology is Hausdorff, and you have a topological manifold.

A smooth manifold is defined in exactly the same way, except you assume that the change of coordinate maps, $\phi_2\circ\phi_1^{-1}: \phi_1(O_1\cap O_2) \rightarrow \phi_2(O_1\cap O_2)$, are smooth.

Clearly, a smooth manifold is a topological manifold. If you start with a topological manifold, extend its atlas to a maximal atlas, then you can ask whether there is a subatlas (i.e., a subset of coordinate maps) that satisfies the definition of an atlas of a smooth manifold. If so, you say that the topological manifold is smoothable.

There is no reason why there couldn't be two different smooth subatlases of a topological manifold. And, if there are two such subatlases, there is no reason why they should be compatible with each other. In other words, if you have a coordinate map $\phi_1: O_1 \rightarrow \R^n$ in the first smooth atlas and a coordinate map $\phi_2: O_2 \rightarrow \R^n$ in the second smooth atlas, it does not necessarily follow that $\phi_2\circ\phi_1^{-1}: \phi_1(O_1\cap O_2) \rightarrow \phi_2(O_1\cap O_2)$ is smooth. It is homeomorphic, since both lie in the topological atlas.

Now, to make things even more complicated, it is possible that there is a global homeomorphism $\Phi: (M,\mathcal{A}_1) \rightarrow (M,\mathcal{A}_2)$ that is a smooth diffeomorphism of the two apparently different smooth manifolds. So, even though the two subatlases are incompatible, they actually define two smooth structures on $M$ that are diffeomorphic.

Finally, we say that a topological manifold has more than one smooth structure on it if there are two smooth subatlases such that no map $\Phi$ as described in the previous paragraph exists.

$\endgroup$
7
  • $\begingroup$ Oops. I forgot return to functions. You can now say that a function $f: M \rightarrow \R$, where $M$ is a smooth manifold, is smooth if $f\circ\Phi^{-1}: \phi(O) \rightarrow \R$ is smooth for any coordinate map $\phi$. The assumption on the change of coordinate maps is the necessary and sufficient condition that you don't run into any logical inconsistencies in this definition. $\endgroup$
    – Deane Yang
    Mar 22, 2022 at 23:26
  • $\begingroup$ I understand this (and love the language of sheaves), but it does not answer the question of what physical idea the notion of smooth structure captures. Imagine two homeomorphic non-isometric metric spaces. You see how they differ. Imagine two homeomorphic non-diffmomorphic manifolds. Do you see how they differ? $\endgroup$ Mar 22, 2022 at 23:37
  • 1
    $\begingroup$ Alas, I didn't really answer the question you asked. Unfortunately, I don't see any physical intuition for two homeomorphic but non-diffeomorphic manifolds. Notice that this is a global concept. In general, there will exist diffeomorphisms between dense open subsets of the two manifolds that can be extended uniquely as homeomorphisms that are not diffeomorphisms. $\endgroup$
    – Deane Yang
    Mar 23, 2022 at 0:23
  • $\begingroup$ MathJax note: As unpleasant as it is, you have to write $\newcommand\R{\mathbb R}$The purpose …, not $\newcommand\R{\mathbb R}$ The purpose …, or your text will have a spurious space. I have edited accordingly. $\endgroup$
    – LSpice
    Mar 23, 2022 at 19:10
  • 1
    $\begingroup$ @LSpice, thanks! $\endgroup$
    – Deane Yang
    Mar 23, 2022 at 20:29

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.