LECTURE NOTES ON BASIC DIFFERENTIAL TOPOLOGY
ANTONIO LERARIO
These are notes for the course “Advanced Geometry 2” for the Master Diploma in Mathemat- ics at the University of Trieste and at SISSA. These notes are by no mean complete: excellent references for the subject are the books [1, 3, 4, 5, 6], from which in fact many proofs are taken or adapted. The notes contain some exercises, which the reader is warmly encouraged to solve (sometimes part of a proof is left as an exercise). I apologize in advance for the many mistakes and imprecisions that the reader might find: I will greatly appreciate if he/she could point out any of them.
References [1] R. Bott and L. W. Tu. Di↵erential forms in algebraic topology,volume82ofGraduate Texts in Mathematics. Springer-Verlag, New York-Berlin, 1982. [2] M. Cornalba. Note di geometria di↵ereziale. http://mate.unipv.it/cornalba/dispense/geodiff.pdf. [3] V. Guillemin and A. Pollack. Di↵erential topology.AMSChelseaPublishing,Providence,RI,2010.Reprint of the 1974 original. [4] M. W. Hirsch. Di↵erential topology,volume33ofGraduate Texts in Mathematics.Springer-Verlag,NewYork, 1994. Corrected reprint of the 1976 original. [5] J. M. Lee. Introduction to smooth manifolds,volume218ofGraduate Texts in Mathematics.Springer,New York, second edition, 2013. [6] J. W. Milnor. Topology from the di↵erentiable viewpoint.PrincetonLandmarksinMathematics.Princeton University Press, Princeton, NJ, 1997. Based on notes by David W. Weaver, Revised reprint of the 1965 original.
Date:October16,2018. 1 2 ANTONIO LERARIO
Schedule and plan of the course
Lessons on Wednesdays and Thursdays 11am–1pm in room A134, located in SISSA main building There is no lesson on Thursday, November 1st (national hoilday)
Lesson 01, 03/10 Di↵erentiable manifolds and smooth maps: definition and examples (spheres, projective spaces and Grassmanians) Lesson 02, 04/10 The tangent bundle and the tangent map; smooth maps and their di↵erential Lesson 03, 10/10 Submanifolds and how to produce them. Immersions, embeddings and regular value theorem (notion of transversality). Lesson 04, 11/10 More examples. Lesson 05, 17/10 How abstract is the notion of manifold? Part 1: compact manifolds embed in Rn Lesson 06, 18/10 Sets of measure zero and mini-Sard Lesson 07, 24/10 How abstract is the notion of manifold? Part 2: Weak Whitney embedding theorem Lesson 08, 25/10 Transversality and Sard’s lemma; every closed set is the zero set of a smooth function Lesson 09, 31/10 Proof of Sard’s Lemma Lesson 10, 07/11 Parametric transversality and applications Lesson 11, 08/11 Parametric transversality and applications, continuation Lesson 12, 14/11 The normal bundle of an immersion Lesson 13, 15/11 Approximations of continuous functions and applications (e.g. homotopy groups of spheres) Lesson 14, 21/11 Vector bundles: definition Lesson 15, 22/11 Examples of vector bundles Lesson 16, 28/11 Vector bundles and sections of vector bundles on projective spaces Lesson 17, 28/11 Associated bundles, classification theorem Lesson 18, 05/12 Di↵erential forms Lesson 19, 06/12 The Mayer-Vietoris sequence and first applications Lesson 20, 12/12 Orientation and integration Lesson 21, 13/12 Stokes theorem and Poincar´eLemma Lesson 22, 18/12 Examples: cohomology of spheres and projective spaces note: the time, date and location of this lesson is flexible Lesson 23, 19/12 Kunneth formula and Poincar´edual Lesson 24, 20/12 Final applications (degree, invariant cohomology...) LECTURE NOTES ON BASIC DIFFERENTIAL TOPOLOGY 3
1. Differentiable manifolds
Definition 1 (Ck manifold). A manifold of dimension m and class Ck is a paracompact1 Hausdor↵space M such that: (1) for every point x M there exists a neighborhood U of x and a continuous function 2 : U Rm which is a homeomorphism onto an open subset of Rm (the pair (U, )is called! a chart); (2) for every pairs of charts (U , ) and (U , ) such that U U = the map 1 1 2 2 1 \ 2 6 ; 1 m 2 1 U U : 1(U1 U2) R | 1\ 2 \ ! is a Ck map (for k = 0 we obtain topological manifolds, for k 1 di↵erentiable manifolds, for k = smooth manifolds and for k = ! analytic manifolds). 1 k A collection of charts (Uj, j) j J as above such that j J Uj = M is called a C atlas. { } 2 2 k Remark 2 (Atlases and di↵erential structures). Two CS atlases = (U↵, ↵) ↵ A and = A { k} 2 B k (V ,' ) B for M, are said to be equivalent if their union is still a C atlas. A C - { } 2 A[B di↵erential structure on M is the choice of an equivalence class of Ck atlases. If we take the union of all atlases belonging to a Ck-di↵erential structure, we obtain a maximal Ck atlas. This atlas contains every chart that is compatible with the chosen di↵erentiable structure. (There is a natural one-to-one correspondence between di↵erentiable structures and maximal di↵erentiable atlases.) From now on we will assume that the atlas we work with is maximal, so that we will have all possible charts available. A simple way to enrich a given atlas is as follows. Given a chart : U Rm around a point x M (as in point (1) of Definition 1), and given a neighborhood V !U of x we can easily 2 m ⇢ construct a chart ' : V R by simply taking ' = V . Note that in this way we can construct ! | 1 a chart (V,') around any point with V contractible: it is enough to take V = (BRm ( (x),✏)) for ✏>0 small enough. Example 3. If ' : M Rn be a homeomorphism, then M is an analytic manifold. In fact one ! 1 m m can cover M with the single chart (M,'), and ' ' =id m : R R is analytic. R ! Example 4 (Product manifolds). If M and N are smooth manifolds with respective atlases (U↵, ↵) ↵ A and (V ,' ) B,thenM N is naturally a smooth manifold with the atlas { } 2 { } 2 ⇥ (U↵ V , ↵ ' ) (↵, ) A B. { ⇥ ⇥ } 2 ⇥ n 2 2 n+1 Example 5 (Spheres). The unit sphere S = x0 + + xn =1 R can be endowed { ··· }⇢ n with the structure of a smooth manifold as follows. Consider the point e0 =(1, 0,...,0) S n n 2 and the two open sets of the the sphere defined by U1 = S e0 and U2 = S e0 .We \{ } n \{ n } produce explicit and nice homeomorphisms (i.e. charts) 1 : U1 R and 2 : U2 R , called stereographic projections (see Figure 1), as follows: ! ! 1 1 (x ,...,x )= (x ,...,x ) and (x ,...,x )= (x ,...,x ). 1 0 n 1 x 1 n 2 0 n 1+x 1 n 0 0
1Recall that a paracompact space is a topological space X for which every open cover has a locally finite refinement. More precisely: given an open cover = U↵ ↵ A fo X,thereexistsanotheropencover = U { } 2 V V B such that (i) for every B there exists ↵( ) A such that V U↵( ) (i.e. refines ); (ii) for { } 2 2 2 ⇢ V U every x X there exists a neighborhood Vx of x which intersects only finitely many elements of (i.e. is 2 V V locally finite). It is worth noticing that if a Hausdor↵space is locally Euclidean (i.e. if it satisfies condition (1) of Definition 1)andconnected,thenthisspaceisparacompactifandonlyifitissecondcountable(i.e.itstopology has a countable basis). 4 ANTONIO LERARIO
e0
Sn
p1
Rn
1(p2) 1(p1)
p2
n n Figure 1. The stereographic projection 1 : S e0 R . \{ }!
1 n n It is easy to verify that : R S is given by: 1 ! 1 1 2 (y ,...,y )= y 1, 2y ,...,2y . 1 1 n 1+ y 2 k k 1 n k k 1 n n This in particular allows to write the explicit expression for 2 1 U U : R 0 R 0 : | 1\ 2 \{ }! \{ } 1 y (y)= , 2 1 y 2 k k n which is a smooth map. Hence (U1, 1), (U2, 2) is a smooth atlas for S and turns it into a smooth manifold. This is called{ the standard di↵erential} structure on Sn. Example 6 (Real projective spaces). The real projective space RPn can be endowed the struc- n n+1 ture of smooth manifold as follows. Recall that RP =(R 0 )/ ,wherep1 p2 if and \{ } ⇠ ⇠ only if there exists = 0 such that p1 = p2. We denote by [x0,...,xn] the equivalence class n+16 of (x0,...,xn) R 0 (the xj are called homogeneous coordinates). For every j =0,...,n 2 \{ } n consider the open set Uj RP defined by: ⇢ U = [x ,...,x ] such that x =0 , j { 0 n j 6 } n together with the homeomorphism j : Uj R given by: ! x x x ([x ,...,x ]) = 0 ,..., j ,..., n j 0 n x x x ✓ j j j ◆ (here the “hat” symbol denotes that this element has beenc removed from the list). The inverse 1 n n : R RP is given by: j ! 1 j (y0,...,yj,...,yn)=[y0,...,1,...yn], where the “1” is in position j. As a consequence, for every i = j we have: b 6 1 y0 yi 1 yn (y ,...,y ,...,y )= ,..., ,..., ,... , i j 0 j n y y y y ✓ i i i j ◆ which is a di↵eomorphism of Rn 0 to itself. c \{ b} Exercise 7. Prove that RP1 and S1 are homeomorphic. Example 8 (real Grassmannians). The real Grassmannian G(k, n) consists of the set of all k-dimensional vector subspaces of Rn, endowed with the quotient topology of the map: n k q : M R ⇥ such that rk(M)=k G(k, n),q(M) = span columns of M . { 2 }! { } LECTURE NOTES ON BASIC DIFFERENTIAL TOPOLOGY 5
In other words, G(k, n) (as a topological space) can be considered as the quotient of the set of n k n k real matrices of rank k (viewed as a subset of R ⇥ ) under the equivalence relation: ⇥ M1 M2 there exists L GL(k, R) such that M1 = M2L. ⇠ () 2 n 1 Observe that G(1,n)=RP and that the above definition mimics the equivalence relation v1 v2 if and only if there exists GL(1, R)=R 0 such that v1 = v2. ⇠ 2 \{ } We want to endow G(k, n) with the structure of a smooth manifold. For every multi-index J =(j ,...,j ) n we denote by M the k k submatrix of M obtained by selecting the 1 k 2 k |J ⇥ rows j ,...,j (in this way M c denotes the (n k) k submatrix of M obtained by selecting 1 k J the complementary rows). For| every such multi-index ⇥J we define the open set: U = [M] G(k, n) such that det(M ) =0 . J { 2 |J 6 } (Note that this set is well defined.) Mimicking again the definition for projective spaces, we (n k) k define the manifold charts J : UJ R ⇥ by: ! 1 ([M]) = (MM ) c . J J |J The expression of the inverse of a matrix in terms of its determinant and its cofactor sshows that n 1 for every pair of indices J ,J the map is smooth. In this way (U , ) n 1 2 k J2 J1 J J J 2 { } 2 k is a smooth atlas for the k(n k)-dimensional manifold G(k, n). { } Exercise 9. Fill in all the details in the previous definition of the smooth structure on the Grassmannian.
Example 10 (The Complex projective line). Recall that the complex line CP1 is defined as the 2 quotient space (C 0 )/ where (z0,z1) (z0,z1) for every C 0 . As we did for real \{ } ⇠ ⇠ 2 \{ } 1 projective spaces, we denote by [z0,z1] the homogeneous coordinates of a point on CP . Consider 2 the two open sets U0 = z0 =0 and U1 = z1 =0 together with the charts j : Uj C R for j =0, 1 which are given{ 6 by:} { 6 } ! ' z0 z1 1([z0,z1]) = and 0([z0,z1]) = . z1 z0 1 1 We have that 0 (z)= , which is a holomorphic map C 0 C 0 and consequently, 1 z \{ }! \{ } using the identification C R2, a smooth map R2 0 R2 0 . If we wanted, we could also ' \{ }! \{ } work with j as a real map, as follows (however, as the reader will see, using the field structure 2 4 of C R simplifies a lot the computations). Given (x0,y0,x1,y1) R , let us denote by ' 2 2 2 (x0 + iy0,x1 + iy1)=(z0,y0) C . We can write 1 : U0 C R as 2 ! ' x0 + iy0 x0x1 + y0y1 y0x1 x0y1 1([x0 + iy0,x1 + iy1]) = = 2 2 + i 2 2 x1 + iy1 x1 + y1 · x1 + y1 2 which means that the real map 1 : U1 R is given by: ! x x + y y y x x y ([x + iy ,x + iy ]) = 0 1 0 1 , 0 1 0 1 , 1 0 0 1 1 x2 + y2 x2 + y2 ✓ 1 1 1 1 ◆ 1 2 1 with inverse : R CP given by: 1 ! 1 1 (x, y)=[1,x+ iy]. 1 1 In particular is given by (x, y) (x, y), which is indeed a smooth map 0 1 U0 U1 x2+y2 | \ 7! R2 0 R2 0 . \{ }! \{ } The complex projective line CP1 is homeomorphic to S2 and in fact, as smooth manifolds, they are indistinguishable (see Exercise 27). 6 ANTONIO LERARIO
x = z + ay x = z + by
✏ H z a b
✏
Az
Figure 2. An open set of the type U3 (grey region) for the topology of the Pr¨ufer surface.
Exercise 11. Generalize the previous example and prove that CPn can be endowed with the structure of a smooth 2n-dimensional manifold. Example 12 (A non-paracompact “manifold”: the Pr¨ufer surface). Let H = (x, y),y >0 be { } the positive half-space and for every z R consider the set 2 3 Az = (x, y, z) R with y 0 { 2 } (each Az is a closed half space). The Pr¨ufer surface is the set
P = H A [ z z ! [2R endowed with the topology generated by open sets of the three following types: (1) open sets U H for the standard euclidean topology; 1 ⇢ (2) open sets U2 Az y<0 (so U2 is a euclidean open set in (x, y, z),y <0 H); (3) open sets of the⇢ form\{U = }A (a, b, ✏) H (a, b, ✏) where (see Figure{ 2): }' 3 z \ z 3 Az(a, b, ✏)= (x, y, z) R ,a