1. Manifolds
We will introduce the machinery necessary for defining a smooth manifold.
1.1. Charts
1. Definition. Let {X\subset M} be some set. An {n}-dimensional Chart consists of
- an open subset {U\subset\mathbb{R}^{n}}
- a map {\varphi\colon U\to X}
2. Remark. We call {\varphi\colon U\to X} a Parametrization of {X}, and {\varphi^{-1}\colon X\to U} a Local System of Coordinates.
3. Remark. Since {\varphi} is an isomorphism, the literature mixes up using {U\to X} and {X\to U}. Milner uses {\varphi\colon U\to X}, but John Lee uses the opposite convention.
4. Definition. Let {(U,\varphi)}, {(V,\psi)} be two charts. We say they are Compatible if
- the set {(\varphi^{-1}\circ\psi)(V)\subset U} is an open set;
- the set {(\psi^{-1}\circ\varphi)(U)\subset V} is an open set;
- the map {\psi^{-1}\circ\varphi\colon\varphi^{-1}(\psi(V))\to\psi^{-1}(\varphi(U))} is smooth; and
- the map {\varphi^{-1}\circ\psi\colon\psi^{-1}(\varphi(U))\to\varphi^{-1}(\psi(V))} is smooth.
In particular, the charts are compatible if {\varphi(U)\cap\psi(V)=\emptyset} is disjoint.
5. Remark. We refer to the maps {\psi^{-1}\circ\varphi} as Transition Functions. The condition of smooth is {C^{\infty}(\mathbb{R}^{n})}, but different manifolds have different conditions (we could have {C^{k}} charts, or {C^{0}} charts, or analytic {C^{\omega}} charts, or...).
In the older literature (e.g., Kobayashi and Nomizu's Foundations of Differential Geometry), the collection of transition functions form a gadget called a Pseudogroup.
6. Remark. We abuse notation, and could be more explicit by writing \psi^{-1}\circ\varphi\colon\varphi^{-1}(\varphi(U)\cap\psi(V))\to\psi^{-1}(\varphi(U)\cap\psi(V)) \tag{1}
▸ Exercise 1. Prove chart compatibility is an equivalence relation.
1.2. Atlases
- Covers {M}: \displaystyle{\bigcup_{\alpha\in A}\varphi_{\alpha}(U_{\alpha})=M}
- Pairwise compatible: for any {\alpha}, {\beta\in A} the charts {(U_{\alpha},\varphi_{\alpha})} and {(U_{\beta},\varphi_{\beta})} are compatible.
8. Definition. Two {n}-dimensional atlases on {M}, {\mathcal{A}} and {\mathcal{B}}, are called Equivalent if their union {\mathcal{A}\cup\mathcal{B}} is also an atlas. That is to say, if any chart of {\mathcal{A}} is compatible with any chart of {\mathcal{B}}.
9. Remark. Remember: charts are compatible, but atlases are equivalent.
10. Lemma. Let {\mathcal{B}} be an atlas, let {(U,\varphi)} and {(V,\psi)} be two charts not contained in {\mathcal{B}}. If {(U,\varphi)} is compatible with every chart of {\mathcal{B}}, and if {(V,\psi)} is compatible with every chart of {\mathcal{B}}, then {(U,\varphi)} is compatible with {(V,\psi)}.
11. Theorem. Equivalence of atlases is an equivalence relation.
Proof: Let {\mathcal{A}}, {\mathcal{B}}, {\mathcal{C}} be arbitrary atlases on {M}.
- Reflexivity: {\mathcal{A}} is equivalent to itself, since by definition any pair of charts in {\mathcal{A}} are compatible.
- Symmetry: let {\mathcal{A}} and {\mathcal{B}} be equivalent atlases, then {\mathcal{B}} and {\mathcal{A}} are equivalent atlases.
- Transitivity: this is the nontrivial part. Let {\mathcal{A}} and {\mathcal{B}} be equivalent atlases, and {\mathcal{B}} be equivalent to {\mathcal{C}}. Then transitivity follows by considering arbitrary charts {(U,\varphi)\in\mathcal{A}} and {(V,\psi)\in\mathcal{C}}, then applying Lemma 10.
Thus "equivalence of atlases" forms an equivalence relation. ∎
12. Proposition. The collection of atlases on a given set {M} is a set, not a proper class.
Proof: The class of atlases is a subcollection of \mathcal{X}=\mathcal{P}\left(\bigcup_{U\in\mathcal{P}(\mathbb{R}^{n})}\mathop{\rm Hom}\nolimits(U,M)\right) \tag{2} where {\mathop{\rm Hom}\nolimits(U,\mathbb{R}^{n})} is the collection of (appropriately smooth, or continuous, or holomorphic, or...) functions from {U} to {M}. By ZF axioms, {\mathcal{X}} is a set. ∎
1.3. Manifolds
13. Definition. Let {M} be a set, let {\mathcal{A}} be an {n}-dimensional atlas on {M}. We call a subset {B\subset M} Open (with respect to {\mathcal{A}}) if for any chart {(U,\varphi)\in\mathcal{A}} the preimage {\varphi^{-1}(B)} is open (in {U}, and thus open in {\mathbb{R}^{n}}). In particular, the images {\varphi(U)} are open.
14. Theorem. If two atlases {\mathcal{A}_{1}} and {\mathcal{A}_{2}} on {M} are equivalent, then a subset {B\subset M} is open with respect to {\mathcal{A}_{1}} if and only it is open with respect {\mathcal{A}_{2}}.
15. Remark. This theorem shows an equivalence class of atlases on {M} makes {M} a topological space. We may therefore meaningfully speak about topological properties of {M} (like compactness, connectedness, and so forth).
16. Corollary. Let {\mathcal{A}} be an {n}-dimensional atlas for {M}. Then the collection of open sets with respect to {\mathcal{A}} form a topology on {M}.
17. Definition. Let {M} be a fixed set. A {n}-Dimensional Differential Structure (or {n}-Dimensional Smooth Structure) on {M} consists of an equivalence class {\mathfrak{D}} of {n}-dimensional atlases on {M} such that
- Second-Countable: {\mathfrak{D}} contains an at most countable atlas;
- Hausdorff: for any distinct {p,q\in M}, there exists disjoint open neighborhoods {U,V\subset M} such that {p\in U} and {q\in V}.
18. Remark (Smooth Structure using a Maximal Atlas). Equivalence classes are awkward to work with, and so it is more popular to consider maximal atlases. An atlas {\mathcal{A}} is maximal if it contains all charts compatible with every chart in {\mathcal{A}}. Given an equivalence class {\mathfrak{A}} of atlases, we may obtain a maximal atlas by considering \mathcal{A}_{\text{max}} = \bigcup_{\mathcal{A}\in\mathfrak{A}}\mathcal{A}. \tag{3} This may be used instead of an equivalence class of atlases in defining a differential structure, provided the second-countable axiom is reworded as: {\mathcal{A}_{\text{max}}} contains an at most countable subatlas.
19. Remark (Convenient Fiction). No one actually constructs either a maximal atlas or a differential structure. We typically construct a smooth atlas on {M}, then announce we are working with the differential structure containing our atlas. Thus maximal atlases and, to some degree, differential structures are a convenient fiction.
20. Definition. A (Smooth) {n}-Dimensional Manifold consists of a set {M} equipped with an {n}-dimensional differential structure.
21. Puzzle. Is this definition correct? By this, I mean: is an "{n}-Dimensional Differential Structure" actually structure (in the sense of "stuff, structure, and properties")?
No comments:
Post a Comment