Manifolds: Topological Versus Differential - Geometry of Manifolds | MATH 7350, Study notes of Mathematics

Material Type: Notes; Professor: Torok; Class: Geometry of Manifolds; Subject: (Mathematics); University: University of Houston; Term: Spring 2009;

Typology: Study notes

Pre 2010

Uploaded on 08/19/2009

koofers-user-fc8
koofers-user-fc8 🇺🇸

10 documents

1 / 2

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Manifolds: Topological versus Differentiable
Here are some facts about the differential (smooth) structures that a topological manifold
of dimension ncan support, and related results.
Unless stated otherwise, all manifolds are without boundary. Counting (e.g., uniqueness)
is meant in the appropriate sense (up to smooth diffeomorphisms for smooth manifolds).
See also the remarks on pages 14 and 37 of the textbook.
(a) n3
Each n-dimensional topological manifold has a unique smooth structure
(J. Munkres, E. Moise).
(b) n4
For each nthere is a connected compact topological manifold that does not admit
a smooth structure.
Each n-dimensional compact connected topological manifold admits at most
countably many different smooth structures.
S7admits exactly 28 different smooth structures (J.W. Milnor and M.A. Ker-
vaire).
Classification of smooth (even compact) manifolds of dimension n4 is very
hard.
Does S4or CP2admit more than one smooth structures?
(c) The Poincar´e conjecture (1904): a compact simply-connected1smooth manifold is
homeomorphic to the sphere of that dimension.
n5 proved by S. Smale (1961).
n= 4 proved by M.H. Freedman (1982).
n= 3 proved by G. Perelman (2002–2003), using the Ricci flow method of
R.S. Hamilton.
This makes the “geometrization conjecture” of W. Thurston close to being
proved2: there are eight standard Riemannian models in dimension 3 (compare
to the dimension-2 case below, where there are three models).
Last updated: Jan 27, 2009; only limited claim of accuracy is made.
1A (path connected) topological space Xis simply-connected if its fundamental group, π1(X), is trivial:
any closed path in Xcan be continuously “shrunk” to a point. The 1-sphere is not simply-connected but all
higher dimensional spheres are.
2Might be already proven, using Perelman’s work.
1
pf2

Partial preview of the text

Download Manifolds: Topological Versus Differential - Geometry of Manifolds | MATH 7350 and more Study notes Mathematics in PDF only on Docsity!

Manifolds: Topological versus Differentiable

Here are some facts about the differential (smooth) structures that a topological manifold of dimension n can support, and related results. Unless stated otherwise, all manifolds are without boundary. Counting (e.g., uniqueness) is meant in the appropriate sense (up to smooth diffeomorphisms for smooth manifolds). See also the remarks on pages 14 and 37 of the textbook.

(a) n ≤ 3

  • Each n-dimensional topological manifold has a unique smooth structure (J. Munkres, E. Moise).

(b) n ≥ 4

  • For each n there is a connected compact topological manifold that does not admit a smooth structure.
  • Each n-dimensional compact connected topological manifold admits at most countably many different smooth structures.
  • S^7 admits exactly 28 different smooth structures (J.W. Milnor and M.A. Ker- vaire).
  • Classification of smooth (even compact) manifolds of dimension n ≥ 4 is very hard.
  • Does S^4 or CP^2 admit more than one smooth structures?

(c) The Poincar´e conjecture (1904): a compact simply-connected^1 smooth manifold is homeomorphic to the sphere of that dimension.

  • n ≥ 5 proved by S. Smale (1961).
  • n = 4 proved by M.H. Freedman (1982).
  • n = 3 proved by G. Perelman (2002–2003), using the Ricci flow method of R.S. Hamilton. This makes the “geometrization conjecture” of W. Thurston close to being proved^2 : there are eight standard Riemannian models in dimension 3 (compare to the dimension-2 case below, where there are three models). ∗Last updated: Jan 27, 2009; only limited claim of accuracy is made. (^1) A (path connected) topological space X is simply-connected if its fundamental group, π 1 (X), is trivial:

any closed path in X can be continuously “shrunk” to a point. The 1-sphere is not simply-connected but all higher dimensional spheres are. (^2) Might be already proven, using Perelman’s work.

(d) n = 1

  • Any 1-dimensional connected smooth manifold is diffeomorphic to either R or S^1 with the canonical structure.

(e) n = 2

  • Any 2-dimensional connected compact oriented smooth manifold is diffeomorphic to the sphere S^2 with zero or more handles attached.
  • Any 2-dimensional compact manifold M admits a Riemannian metric with makes it locally isometric to one of the following (because they are quotients of the manifolds listed below, under a free discrete group action): (1) S^2 with the canonical metric (i.e., the one inherited from S^2 ⊂ R^3 ); this has constant +1 curvature. (2) R^2 with the Euclidean metric; this has constant 0 curvature. (3) The hyperbolic plane, H^2 , with the Poincar´e metric; this has constant − 1 curvature. The model above that applies to M is determined by the sign of its Euler char- acteristic, χ(M ) :=

i≥ 0

(−1)i^ dim Hi(M, R).

[Recall that by Gauss-Bonnet

M KdA^ = 2πχ(M^ ), so the Euler characteristic determines the Gauss curvature if the latter is constant.]

  • For example, the only compact 2-dimensional smooth manifolds that fall in the first case above are S^2 and RP^2 ; in the second case above are the 2-torus, S^1 × S^1 , and the Klein bottle. The surfaces of genus at least 2 (in the orientable case, these are the sphere with at least two handles) admit a metric with constant curvature −1.
  • The only smooth non-compact manifolds that admit a complete metric^3 of con- stant zero curvature are R^2 , R × S^1 and the M¨obius strip (without boundary); there are no such manifolds for curvature +1. The hyperbolic plane is such an example for curvature −1.

(f) Rn

  • For n 6 = 4, Rn^ admits a unique smooth structure.
  • R^4 admits uncountably many non-diffeomorphic smooth structures (S.K. Donald- son and M.H. Freedman, 1984)

(^3) A Riemannian metric is complete if each geodesic can be extended to infinite time. For example, Rn

and any compact manifold are complete, but Rn^ \ { 0 } is not complete with the canonical metric because some geodesics (straight lines in this case) run in the missing origin in finite time.