Está en la página 1de 80

1

Lecture Notes

An Introduction to Riemannian Geometry


(version 1.102 - September 1996)

Sigmundur Gudmundsson
(Lund University)

The latest version of this document can be obtained from: http://www.maths.lth.se/matematiklu/personal/sigma/index.html

Preface

These lecture notes grew out of an M.Sc.-course on di erential geometry which I gave at the University of Leeds 1992. Their main purpose is to introduce the beautiful theory of Riemannian Geometry a still very active area of Mathematics. This is a subject with no lack of interesting examples. They are indeed the key to a good understanding of it and will therefore play a major role throughout this work. Of special interest are the classical Lie groups allowing concrete calculations of many of the abstract notions on the menu. The study of Riemannian Geometry is rather meaningless without some basic knowledge on Gaussian Geometry i.e. the di erential geometry of curves and surfaces in Euclidean 3-space. For this we recommend the excellent textbook: M. P. do Carmo, Di erential Geometry of Curves and Surfaces, Prentice Hall (1976). These lecture notes are written for students with a good understanding of linear algebra, real analysis of several variables, the classical theory of ODEs and some topology. The most important results stated in the text are also proved there. Other smaller ones are left to the reader as exercises, which follow at the end of each chapter. This format is aimed at students willing to put hard work into the course. It is my intention to extent this very incomplete rst draft, which unfortunately still contains typing errors, and include some of the di erential geometry of the Riemannian symmetric spaces. For further reading we recommend the very interesting textbook: M. P. do Carmo, Riemannian Geometry, Birkhauser (1992). Lund University, May 1996 Sigmundur Gudmundsson

Contents
Chapter 1. Chapter 2. Chapter 3. Chapter 4. Chapter 5. Chapter 6. Chapter 7. Chapter 8. Chapter 9. Chapter 10. Introduction Di erentiable Manifolds The Tangent Space The Tangent Bundle Immersions, Embeddings and Submersions Riemannian Manifolds The Levi-Civita Connection Geodesics The Curvature Tensor Curvature and Local Geometry 5 7 15 23 31 35 43 51 63 69

CONTENTS

CHAPTER 1

Introduction
On the 10th of June 1854 Georg Friedrich Bernhard Riemann gave his famous "Habilitationsvortrag" in the Colloquium of the Philosophical Faculty at Gottingen. His talk with the title "Uber die Hypothesen, welche der Geometrie zu Grunde liegen" is often said to be the most important in the history of di erential geometry. Gauss, at the age of 76, was in the audience and is said to have been very impressed. Riemann's revolutionary ideas generalized the geometry of surfaces which had been studied earlier by Gauss, Bolyai and Lobachevsky. Later they lead to an exact de nition of the modern concept of an abstract n-dimensional Riemannian manifold.

1. INTRODUCTION

CHAPTER 2

Di erentiable Manifolds
The main purpose of this chapter is to introduce the concept of a di erentiable manifold, generalizing the idea of a di erentiable surface studied in most introductory courses on Di erential Geometry. Furthermore we study submanifolds and di erentiable maps between manifolds. De nition 2.1. Let M be a topological Hausdor space with a countable basis. M is called a topological manifold if there exists an m 2 N and for every point p 2 M an open neighbourhood Up of p, such that Up is homeomorphic to some open subset Vp of R m . The natural number m is called the dimension of M . To denote that the dimension of M is m we write M m . For an open subset U of R m and r 2 N we denote by C r (U R n ) the r-times continuously di erentiable maps from U to R n . By smooth r ! we mean C 1 = \1 r=1 C and C means real analytic. De nition 2.2. Let M m be a topological manifold, U be an open and connected subset of M and : U ! R m be a continuous map which is a homeomorphism onto its image (U ). Then (U ) is called a chart (or local coordinate) on M . A collection A = f(U )j 2 I g of charts on M is called a C r -atlas if i. M = U , ii. The corresponding transition maps ;1 j (U \U ) : (U \ U ) ! R m are C r for all 2 I. De nition 2.3. Let A be a C r -atlas on M . A chart (U ) on M is said to be compatible with A if A f(U )g is a C r -atlas on M . A ^ is said to be maximal if it contains all charts compatible C r -atlas A ^ on M is also called a C r -structure on M . with it. A maximal atlas A r A C -manifold is a topological manifold M with a C r -structure. A manifold is said to be smooth if it is C 1 and real analytic if it is C!.
7

2. DIFFERENTIABLE MANIFOLDS

Example 2.5. Let R m be the real m-dimensional vector space with the usual topology T induced by the distance function p d(x y) = (x1 ; y1)2 + : : : + (xm ; ym)2 : For the topological space (R m T ) we have the trivial C ! -atlas A = f(R m idRm )j idRm : x 7! xg ^ on R m . inducing the standard C ! -structure A Example 2.6. Let S m denote the unit sphere in R m+1 i.e. 2 S m = fx 2 R m+1 j x2 1 + : : : + xm+1 = 1g equipped with the subset topology TSm induced by T on Rm+1 . Let n be the north pole n = (1 0) 2 R R m and s be the south pole s = (;1 0) on S m , respectively. Put Un = S m ; fng, Us = S m ; fsg and de ne n : Un ! R m , s : Us ! R m by 1 (x : : : x ) m+1 n : (x1 : : : xm+1 ) 7! 1;x 2
1

A.

C r -structure

Remark 2.4. Note that a C r -atlas A on M determines a unique ^ containing A. A ^ consists of all charts compatible with A

: (x1 : : : xm+1 ) 7! 1 (x2 : : : xm+1 ): 1 + x1 ; 1 1 m m Then the transition maps s n n ; s : R ; f0g ! R ; f0g are given by x 7! x=jxj2 so A = f(Un n) (Us s)g is a C ! -atlas on ^) is called the m-dimensional standard S m. The C ! -manifold (S m A sphere. Example 2.7. On the set R m+1 ; f0g we de ne the equivalence relation by x y if and only if there exists 2 R such that x = y: Let : R m+1 ; f0g ! (R m+1 ; f0g)= be the natural projection : x 7! x] onto the quotient space which we denote by R P m and equip with the quotient topology induced by and T on R m+1 . For k 2 f1 : : : m + 1g put Uk = f x] 2 R P m j xk 6= 0g and de ne k : Uk ! R m by x1 : : : xk;1 xk+1 : : : xm+1 ): k : x] 7! ( xk xk xk xk
s

2. DIFFERENTIABLE MANIFOLDS

If x] y] then x = y for some 2 R so xl =xk = yl=yk for all l. This means that k is well de ned for all k. The transition maps ; 1 m k l j l (Ul \Uk ) : l (Ul \ Uk ) ! R are given by 1 : : : xl;1 xl+1 : : : xm+1 ) 7! ( x1 : : : xk;1 xk+1 : : : xm+1 ) (x xl xl xl xl xk xk xk xk ! m so A = f(Uk k )j k = 1 : : : m + 1g is a C -atlas on R P . The mani^) is called the m-dimensional real projective space. fold (R P m A ^ = C f1g, C = C ; f0g, U0 = C and Example 2.8. Put C ^ ; f0g. Then de ne the charts 0 : U0 ! C , 1 : U1 ! C U1 = C by 0 : z 7! z and 1 : w 7! 1=w, respectively. Then the transition 1 ; maps 1 ; C ! C are given by z 7! 1=z so A = 0 0 11 : ! ^ . The C ! -manifold (C ^ A ^) is f(U0 0) (U1 1)g is a C -atlas on C called the Riemann sphere. ^ always mean the In what follows we shall by R m , S m, R P m or C standard manifolds above. ~ of (M m A ^). SubWe now de ne the concept of a submanifold M manifolds are very important objects which will be given considerable attention as we go along. The additional structures that we will intro^) give the corresponding induced structures on M ~ in a duce on (M m A natural manner. We shall often be interested in how they are related. ^) be a C r -manifold and M ~ be a subset De nition 2.9. Let (M m A ~ is said to of M equipped with the subset topology of M . The set M ~ there exists be an n-dimensional submanifold of M if for each p 2 M ^ such that a chart (Up p) 2 A i. p 2 Up ii. p : Up ! R n R m;n satis es ~ ) = p(Up) \ (R n f0g): p(Up \ M ~ The positive natural number (m ; n) is called the codimension of M m in M . ^) be a C r -manifold and M ~ be an Proposition 2.10. Let (M m A ^). For every point p 2 M ~ let n-dimensional submanifold of (M m A ^ be a chart on M such that p 2 Up and the map p : Up ! (Up p) 2 A n m ; n R R satis es ~ ) = p(Up) \ (R n f0g): p(Up \ M Further let : R n R m;n ! R n be the natural projection onto the rst
factor. Then

~ B = f(Up \ M

~ )j p jUp\M

~g p2M

10

2. DIFFERENTIABLE MANIFOLDS

~ . In particular (M ~ B ^) is an n-dimensional C r is a C r -atlas for M manifold. Proof. See Exercise 2.2. We shall now see how the well-known Inverse Function Theorem for nite dimensional real vector spaces can be used to construct di erentiable manifolds as submanifolds of R m . Fact 2.11 (The Inverse Function Theorem). Let U be an open subset of R m and F : U ! R m be a C r -map. If p 2 U and the derivative DFp : R m ! R m of F at p is invertible, then there exist open neighbour^ = F jV : V ! W is hoods V around p and W around F (p) such that F ^ );1 : W ! V is a C r -map. The derivative bijective and the inverse (F ^ ;1)F (p) of F ^ ;1 at F (p) is given by D (F ^ ;1)F (p) = (DFp);1: D (F De nition 2.12. Let U be an open subset of R m and F : U ! R n be a C r -map. A point p 2 U is called a critical point for F if DFp : R m ! R n is not of full rank, and a regular point if it is not critical. A point q 2 F (U ) is called a regular value of F if every point p 2 F ;1(fqg) is a regular point for F and a critical value otherwise. Theorem 2.13 (The Implicit Function Theorem). Let m > n and F : U ! R n be a C r -map from an open subset U of R m . If q 2 F (U ) is a regular value of F then F ;1(fqg) is an (m ; n)-dimensional submanifold of R m . ;1(fqg) and Kp be the kernel Proof. Let p be an element of F of the derivative DFp i.e. the (m ; n)-dimensional subspace of R m given by Kp = fv 2 R m j DFp v = 0g. Let p : R m ! R m;n be a linear map such that pjKp : Kp ! R m;n is bijective and de ne Gp : U ! R n R m;n by Gp : x 7! (F (x) p(x)). The derivative (DGp)p : R m ! R m of Gp is, with respect to the ? Kp, given by decomposition R m = Kp

DFp (0 p) so it is bijective. It now follows from the inverse function theorem that there exist open neighbourhoods Vp around p and Wp around Gp(p) 1 ^ p = GpjVp : Vp ! Wp is bijective. The inverse G ^; such that G p : Wp ! 1 ;1 1 ^; ^; Vp is C r and D(G p )Gp (p) = (DGp )p so that D(G p )y is bijective for ~p = F ;1(fqg) \ Vp then all y 2 Wp. Now put U 1 m;n ) \ W ) ~p = G ^; U p p ((fq g R DGp =

so if : R n R m;n ! R m;n is the natural projection onto the second factor, then the map ~p = Gp : U ~p ! (fqg R m;n ) \ Wp ! R m;n ~p of p. The point q 2 F (U ) is a local chart on the open neighbourhood U is a regular value so the set ~p ~p)j p 2 F ;1(fqg)g B = f(U is a C r -atlas for F ;1(fqg). Example 2.14. Let F : R m+1 ! R be the C ! -map given by m +1 X 2 F : (x1 ::: xm+1 ) 7! 1 2 i=1 xi : The derivative DFx of F at x is given by DFx = x1 ::: xm+1 ] so (DFx) (DFx)t = jxj2 2 R . This means that 1=2 2 R is a regular value of F so the bre S m = fx 2 R m+1 j jxj2 = 1g = F ;1(f1=2g) of F is an m-dimensional submanifold of R m+1 . It is called the mdimensional sphere. Example 2.15. Let F : R m+1 R m+1 ! R 2 be the C ! -map de ned by F : (p v) 7! ((jpj2 ; 1)=2 hp vi). The derivative DF(p v) of F at (p v) is given by p 0 : DF(p v) = v p Hence det DF (DF )t] = jpj2(jpj2 + jvj2) = (1 + jvj2) > 0 on F ;1(f0g). This implies that F ;1(f0g) = f(p v) 2 R m+1 R m+1 j jpj2 = 1 and hp vi = 0g which we denote by TS m is a 2m-dimensional submanifold of R 2m+2 . We shall see later that TS m is what is called the tangent bundle of the m-dimensional sphere. We are now interested in di erentiable maps between two di erentiable manifolds i.e. those which respect the di erentiable structures of the manifolds involved. ^) and (N n B ^) be two C r -manifolds. De nition 2.16. Let (M m A A map : M m ! N n is said to be a C r -map if for all charts (U 1) 2 ^ and (V 2) 2 B ^ the maps A ; 1 ;1(V )) ! R n 2 1 j 1 (U \ ;1 (V )) : 1 (U \

2. DIFFERENTIABLE MANIFOLDS

11

12

2. DIFFERENTIABLE MANIFOLDS

^) ! R is called a C r -function are of class C r . A C r -map f : (M m A on M . ^) ! (N ^ C ^) and 2 : (N ^ C ^) ! Proposition 2.17. Let 1 : (M A ^) be two C r -maps, then the composition 2 1 : (M A ^) ! (N B ^) (N B r is also a C -map. Proof. See exercise 2.7. Example 2.18. It is easily seen that the following maps are C ! i.e. real analytic, see Exercise 2.8. i. 1 : S 2 R 3 ! S 3 R 4 , 1 : (x y z) 7! (x y z 0). ii. 2 : S 3 C 2 ! S 2 C R , 2 : (z1 z2) 7! (2z1 z2 jz1 j2 ; jz2 j2). iii. 3 : R 1 ! S 1 C , 3 : t 7! eit. iv. 4 : S m ! R P m , 4 : x 7! x]. v. 5 : R m+1 ; f0g ! R P m , 5 : x 7! x]. vi. 6 : R m+1 ; f0g ! S m, 6 : x 7! x=jxj. Example 2.19. Let R m m denote the set of all real m m-matrices and Sym(R m ) be the subset of those which are symmetric i.e. Sym(R m ) = fA 2 R m m j A = At g: Then Sym(R m ) can be identi ed with R n where n = m(m + 1)=2. De ne F : R m m ! Sym(R m ) by F : A 7! AAt . The di erential DFA of F at A 2 R m m is given by DFA(X ) = AX t + XAt 2 Sym(R m ). For A 2 O(m) = F ;1(feg) = fA 2 R m m j AAt = eg and Y 2 Sym(Rm ) we see that DFA(Y A) = 2Y so DFA is surjective for every A 2 O(m). Hence e is a regular value of F and following the implicit function theorem O(m) is an m(m ; 1)=2-dimensional C ! -submanifold 2 m m m of R = R . The set O(m) is the well known orthogonal group and the usual matrix multiplication is a group structure on it. It is easily checked that the map : O(m) O(m) ! O(m) with : (x y) 7! x y;1 is C ! . De nition 2.20. A Lie group is a C ! -manifold G with a group structure such that the map : G G ! G with : (x y) 7! x y;1 is C ! . Example 2.21. Let + denote the usual addition in R m . Then the pair (R m +) is an abelian Lie group. Example 2.22. Let denote the usual multiplication of the real numbers R , the complex numbers C or the quaternions H . Then (R ), (C ), and (H ) are Lie groups. The corresponding unit spheres S 0, S 1 and S 3 are compact Lie subgroups in the trivial way.

2. DIFFERENTIABLE MANIFOLDS

13

Example 2.23. We now give a few examples of matrix Lie groups. In all cases the operation is the usual matrix multiplication. i. GL(Rm ) = fA 2 R m m j det A 6= 0g is called the real general linear group, ii. SL(R m ) = fA 2 R m m j det A = 1g is called the real special linear group, iii. O(m) = fA 2 R m m jAt A = eg is called the orthogonal group, iv. SO(m) = fA 2 O(m)j det A = 1g is called the special orthogonal group, v. GL(C m ) = fA 2 C m m j det A 6= 0g is called the complex general linear group, vi. SL(C m ) = fA 2 C m m j det A = 1g is called the complex special linear group, vii. U(m) = fA 2 C m m jAt A = eg is called the unitary group, and viii. SU(m) = fA 2 U(m)j det A = 1g is called the special unitary group. ^) and (N B ^) are said De nition 2.24. Two C r -manifolds (M A ^) ! to be di eomorphic if there exists a bijective C r -map : (M A ^), such that its inverse ;1 : (N B ^) ! (M A ^) also is C r . The (N B ^) and (N B ^). map is called a di eomorphism between (M A ^ and B ^ be two C r -structures on the same De nition 2.25. Let A ^ and B ^ are said to be di erent if the identity topological manifold M . A ^ ^ map idM : (M A) ! (M B) is not a di eomorphism. ^), (N n B ^) be C r -manifolds of the Deep Result 2.26. Let (M m A same dimension i.e. m = n. If M and N are homeomorphic as topo^) and (N B ^) are di eomorphic. logical spaces and m 3 then (M A The following remarkable result was proved by J.Milnor in his famous paper: Di erentiable structures on spheres, Amer. J. Math. 81 (1959), 962-972. Deep Result 2.27. The seven dimensional sphere S 7 has exactly 28 di erent di erentiable structures.

14

2. DIFFERENTIABLE MANIFOLDS

Find a C 1-structure on M . Exercise 2.2. Find a proof for Proposition 2.10. Exercise 2.3. Let S 1 be the 1-dimensional sphere given by fz 2 C j jz j2 = 1g. Use the maps 1 : C ; f;1g ! C with 1 : z 7! (z ; 1)=(z + 1) and 2 : C ; f1g ! C with 2 : z 7! (1 + z)=(1 ; z) to show that S 1 is a 1-dimensional submanifold of C = R 2 . Exercise 2.4. Show that the m-dimensional torus T m = fz 2 C m j jz1j = ::: = jzmj = 1g is an analytic submanifold of C m = R 2m . Exercise 2.5. Let F : R m+1 R m+1 ! R 2 be the analytic map given by F : (x y) 7! 1 (jxj2 + jyj2 jxj2 ; jyj2) 2 and R be the set of critical values of F . Determine whether the sets F ;1(f(1 0)g) and R are submanifolds of their ambient spaces. Exercise 2.6. Let (G ) and (H ) be two Lie groups. Prove that the product manifold G H has a Lie group structure. Exercise 2.7. Find a proof of Proposition 2.17. ^) and (N B ^) be two smooth manifolds Exercise 2.8. Let (M A ~, N ~ be submanifolds of M and N , respectively. Prove that if and M ~ ) contained in N ~ , then the : M ! N is a smooth map with (M ~ !N ~ is smooth. restriction ~ = jM ~ :M ^ and B ^ on (R T ) by the Exercise 2.9. De ne two C ! -structures A following atlases A = f(R id R)j idR : x 7! xg and B = f(R )j : x 7! x3 g. i. Is the chart (R ) compatible with A? ^) and (R B ^) di eomorphic? ii. Are (R A ^ f o 1g) as deExercise 2.10. Prove that (S 2 f s ng) and (C ned above are di eomorphic. Exercise 2.11. Prove the following di eomorphies S 1 = SO(2) S 3 = SU(2) SO(n) O(1) = O(n) SU(n) U(1) = U(n):

^1) and (M2 A ^2) be two smooth manifolds Exercise 2.1. Let (M1 A and M = (M1 M2 T ) be the product space with the product topology.

Exercises

CHAPTER 3

The Tangent Space


In this chapter we develope the idea of a tangent space from the theory of surfaces in R 3 to the more general situation of a di erentiable manifold M m . We see a tangent vector X at a point p as a rst order linear di erential operator on the set of locally de ned functions on the manifold or rather that of functions germs at p. We then prove that the tangent space TpM i.e. the set of all tangent vectors at p is a vector space isomorphic to R m . From now on we shall assume, when not stating otherwise, that our manifolds and maps are smooth i.e. in the C 1-category. Let M m be a manifold. For a point p 2 M let " ^(p) be the set of all smooth functions de ned on an open neighbourhood of p i.e. " ^(p) = ff : Uf ! R j Uf is an open subset of M containing pg: On " ^(p) we de ne the equivalence relation by: f g if and only if there exists an open neighbourhood V Uf \ Ug such that f jV = gjV . By "(p) we denote the set of equivalence classes "(p) = " ^(p)= . The elements f ] of "(p) are called the function germs at p. By the following operations + and on "(p) we make it into an R -algebra. i. f ] + g] = f + g], ii. f ] = f ], iii. f ] g] = f g] for all f g 2 "(p) and 2 R. De nition 3.1. A tangent vector Xp at p 2 M is a map Xp : "(p) ! R such that i. Xp( f + g) = Xp(f ) + Xp(g), ii. Xp(f g) = g(p) Xp(f ) + f (p) Xp(g) for all 2 R and f g 2 "(p). By TpM we denote the set of all tangent vectors Xp at p 2 M . TpM is called the tangent space of M at p. The following operations + and make the tangent space TpM into a real vector space. i. (Xp + Yp)(f ) = Xp(f ) + Yp(f ),
15

16

3. THE TANGENT SPACE

ii. ( Xp)(f ) = Xp(f ) for all Xp Yp 2 TpM , f 2 "(p) and 2 R . For M = R m we denote by "m the set of function germs at 0 2 R m i.e. "m = "(0). For v 2 R m and f 2 "m the directional derivative of f at 0 in the direction of v is given by f (tv) ; f (0) : @v f = lim t!0 t P v @f (0) It is well known that for v = (v1 : : : vm) we have @v f = m i=1 i @xi and that i. @v ( f + g) = @v f + @v g, ii. @v (f g) = g(0) @v f + f (0) @v g, iii. @( v+ w) f = @v f + @w f for all 2 R , v w 2 R m and f g 2 "m Corollary 3.2. If v 2 R m then the directional derivative @v is an element of the tangent space T0 R m . Lemma 3.3. If f 2 "m then there exist functions k 2 "m such that m X @f (0): f (x) = f (0) + xk k (x) and k (0) = @x
Proof.

R Put k (x) = 01 @f=@xk (tx1 : : : txm )dt and the statement immediately follows. Theorem 3.4. The map : R m ! T0 R m given by v 7! @v is a vector space isomorphism. Proof. That the map is linear follows directly from @( v+ w) f = @v f + @w f for all 2 R, m v w 2 R m and f 2 "m. Let v w 2 R such that v 6= w. Choose an element u 2 R m such that hu vi 6= hu wi and de ne f : R m ! R by f (x) = hu xi. Then @v f = hu vi 6= hu wi = @w f so @v 6= @w . This proves that the map is injective.

It follows from the Fundamental Theorem of Calculus that Z 1 @f (tx1 : : : txm )dt f (x) ; f (0) = 0 @t Z 1 @f m X = xk @x (tx1 : : : txm )dt: k 0 k=1

k=1

3. THE TANGENT SPACE


Rm

17

Let be an arbitrary element of T0R m . For k = 1 : : : m let x ^k : ! R be the map (x1 : : : xm) 7! xk and put vk = (^ xk ). For the constant function 1 : (x1 : : : xm) 7! 1 we have (1) = (1 1) = 1 (1) + 1 (1) = 2 (1), so (1) = 0. By the linearity of it follows that (c) = 0 for all constants c 2 R. Let f 2 "m and following Lemma 3.3 write

f (x) = f (0) +
where that
k

m X k=1

x ^k (x)

k (x)

2 "m with k (0) = @f=@xk (0). Then by applying we obtain


(f (0) +
m X k=1 m

(f ) = = = =

x ^k

k) k (0)

(f (0)) +
k=1 @v f

X
k=1

(^ xk ) +

m X k=1

x ^k (0) ( k )

m X vk @f (0)

@xk

where v = (v1 : : : vm) 2 R m . This implies that that the map is surjective.

= @v and proves

Corollary 3.5. Let fek jk = 1 : : : mg be a basis for R m . Then the set f@ek jk = 1 : : : mg is a basis for the tangent space T0 R m . De nition 3.6. Let : M ! N be a map between two manifolds. For a point p 2 M we de ne the map d p : TpM ! T (p)N by
(d p)(Xp)(f ) = Xp(f ) for all Xp 2 TpM and f 2 "( (p)). The map d p is called the di erential of at p 2 M . ~ and : M ~ ! N be maps Proposition 3.7. Let : M ! M between manifolds, then ~ is linear, i. the map d p : TpM ! T (p) M ii. if idM is the identity map, then d(idM )p = idTpM , iii. d( )p = d (p) d p for all p 2 M . The last equation is called the Chain Rule.

18

3. THE TANGENT SPACE

Proof. The only point which is not trivial is the chain rule. If Xp 2 TpM and f 2 "( (p)), then (d (p) d p(Xp))(f ) = (d p(Xp))(f ) = Xp(f ) = d( )p(Xp)(f ): This proves the statement. Corollary 3.8. Let : M ! N be a di eomorphism with inverse = ;1 : N ! M . Then the di erential d p : TpM ! T (p) N at p is bijective and (d p);1 = d (p) . Proof. The statement follows directly from d p d (p) = d( ) (p) = d(idN ) (p) = idT (p) N d (p) d p = d( )p = d(idM )p = idTpM :

As a direct consequence of Corollaries 3.5 and 3.8 we obtain the following result which generalizes the case when M 2 is a surface in R 3 . Corollary 3.9. Let M m be an m-dimensional manifold and p 2 M . Then the tangent space TpM at p is an m-dimensional real vector space. Proof. Let (U y ) be a chart on M with y (p) = 0. Then the linear map dyp : TpM ! T0 R m = R m is a vector space isomorphism. Next we show that a local chart around a point p 2 M gives a canonical basis for the tangent space TpM . Proposition 3.10. Let M m be a manifold and (U y) be a local coordinate on M . Further let fek j k = 1 : : : mg be the canonical basis @ j 2 T M by for R m . For p 2 M we de ne @y p k p @ j : f 7! @f (p) = @ (f y;1)(y(p)): ek @yk p @yk @ j j k = 1 2 : : : mg is a basis for the tangent space T M for Then f @y p k p all p 2 U . Proof. We are assuming as usual that the manifold M is smooth so the inverse y;1 of y is smooth with di erential (dy;1)y(p) : Ty(p) R m ! TpM satisfying (dy;1)y(p) (@ek )(f ) = @ek (f y;1)(y(p)) @ j )(f ) = ( @y p k

3. THE TANGENT SPACE

19

for all f 2 "(p). This proves the statement. We shall now give an alternative description of the tangent space ^ (p) be the set of all locally de ned C 1-curves passing TpM . Let C through the point p 2 M i.e. ^ (p) = f : (; ) ! M j is C 1 and (0) = pg: C ^ (p) we have the di erential d 0 : T0 R ! TpM . For an element 2 C Let e1 = @e1 2 R = T0 R be the positive unit tangent of T0R . On the ^ (p) we de ne the equivalence relation by set C 1 2 if and only if (d 1)0 (e1 ) = (d 2)0 (e1 ): This means that the two parametrized curves 1 and 2 are identi ed if they have the same tangent at the point p. By C (p) we denote the ^ (p)= . set of equivalence classes i.e. C (p) = C It is an easy exercise to show that the map : C (p) ! TpM with : ] 7! (d )0(e1) is bijective. This implies that TpM can be identi ed with C (p) being the set of all possible tangents to curves going through the point p. Hence a vector v 2 TpM can be thought of as a rst order di erential operator acting on the functions de ned locally around the point p 2 M as follows: Let f : U ! R be a function de ned on an open subset U of M containing p. Furthermore let : I ! U be a curve with (0) = p and _ (0) = v. Then the action of v on f is given by d (f (t))j : v(f ) = dt t=0 Note that this is independent of the choice of the curve as long as (0) = p and _ (0) = v. This second interpretation of TpM shows that if m < n and M m is a submanifold of R n , then TpM is simply the tangent space of M at p in the classical sense i.e. the space of all tangents to curves at the point p 2 M . We can now use this alternative description to determine the tangent space TeO(m) of the orthogonal group O(m) at the neutral element e. Example 3.11. Let e denote the neutral element of the orthogonal group O(m) i.e. the identity matrix. Let A : (; ) ! O(m) be a curve in O(m) such that A(0) = e. Then A(s) A(s)t = e for all s 2 (; ). Di erentiation yields fA0(s) A(s)t + A(s) A0(s)t gjs=0 = 0 or equivalently A0 (0)+ A0 (0)t = 0. This means that each tangent vector of O(m) at e is a skew-symmetric matrix.

20

3. THE TANGENT SPACE

On the other hand, for an arbitrary real skew-symmetric matrix Z de ne : R ! R m m by : s 7! Exp(s Z ), where Exp is the usual exponential map for matrices de ned in Exercise 3.1. Then (s) (s)t = Exp(s Z ) Exp(s Z )t = Exp(s Z ) Exp(s Z t) = Exp(s(Z + Z t )) = Exp(0) = e: This shows that is a curve on the orthogonal group, (0) = e and 0(0) = Z so Z is an element of Te O(m). Hence TeO(m) = fX 2 R m m j X + X t = 0g: The dimension of TeO(m) is therefore m(m ; 1)=2. This con rms our calculations of the dimension of O(m) in chapter 2 since we know that dim(O(m)) = dim(TeO(m)).

3. THE TANGENT SPACE

Exercises

21

de ned by

Exercise 3.1. Let the exponential map Exp : C m


Exp : A 7!

1 Ak X
k=0

! Cm

be

Prove that for all A B 2 C m m i. det Exp(A)] = etrace(A) , ii. if A B = B A then Exp(A + B ) = Exp(A) Exp(B ), iii. Exp(At ) = Exp(A)t . Exercise 3.2. Use the results from Exercise 3.1 to determine the tangent space TeG at the neutral element e and the dimension of the following Lie groups: GL(R m ), SL(R m ), O(m), SO(m), GL(C m ), SL(C m ), U(m), SU(m). Exercise 3.3. Let p be an arbitrary point on the unit sphere S 2n+1 of C n+1 = R 2n+2 . Determine the tangent space TpS 2n+1 and show that it contains an n-dimensional complex subspace of C n+1 .

k! :

22

3. THE TANGENT SPACE

CHAPTER 4

The Tangent Bundle


^) In this chapter we construct for each di erentiable manifold (M A its tangent bundle TM . Intuitively this is the object we get by glueing at each point p 2 M the corresponding tangent space TpM . This way ^ on we obtain a 2m-dimensional topological manifold. The structure A c c) M induces a di erentiable structure A on TM which makes (TM A into a di erentiable manifold. We then study vector elds which can be thought of as being maps from M into TM . These are fundamental tools for the geometric study of manifolds. De nition 4.1. Let E and M be topological manifolds and : E ! M be a continuous surjective map. The triple (E M ) is called an n-dimensional topological vector bundle over M if i. for each p 2 M the bre Ep = ;1(p) is an n-dimensional vector space, ii. for each p 2 M there exists a bundle chart ( ;1(U ) ) consisting of the pre-image ;1(U ) of an open neighbourhood U of p and a homeomorphism : ;1(U ) ! U R n such that for all q 2 U the map q = jEq : Eq ! fqg R n is a vector space isomorphism. De nition 4.2. Let (E M ) be an n-dimensional topological vector bundle over M . It is said to be trivial if there exists a global bundle chart : E ! M R n . Example 4.3. Let M be a topological manifold and : M R n ! M be the natural projection : (x v) 7! x. Then (M R n M ) is a trivial n-dimensional vector bundle. De nition 4.4. Let (E M ) be a topological vector bundle. A continuous map : M ! E is called a section of the bundle if (p) = p for each p 2 M . De nition 4.5. Let (E M ) be an n-dimensional topological vector bundle over M . A collection B = f( ;1(U ) )j 2 I g
23

24

4. THE TANGENT BUNDLE

of bundle charts is called a bundle atlas for (E M ) if M = U . For each pair ( ) there exist a function A : U \ U ! GL(Rn ) such that the corresponding continuous map ;1 j(U \U ) Rn : (U \ U ) R n ! (U \ U ) R n is given by (p v) 7! (p (A (p))(v)): The elements of fA j 2 I g are called the transition maps of the bundle atlas B. De nition 4.6. Let E and M be smooth manifolds and : E ! M be a smooth map such that (E M ) is an n-dimensional topological vector bundle. A bundle atlas B for (E M ) is said to be smooth if the corresponding transition maps are smooth. A smooth vector bundle is a topological vector bundle together with a maximal smooth bundle atlas. A smooth section of (E M ) is called a vector eld. By C 1(E ) we denote the set of all vector elds of (E M ). From now on we assume that all vector bundles are smooth. De nition 4.7. Let M be a manifold and (E M ) be a vector bundle over M . By the following operations we make C 1(E ) into a C 1(M R )-module. In particular, C 1(E ) is a vector space over the real numbers. i. (v + w)p = vp + wp, ii. (f v)p = f (p) vp for all v w 2 C 1(E ) and f 2 C 1(M R ). De nition 4.8. Let M be a manifold and (E M ) be an ndimensional vector bundle over M . A set F = fv1 : : : vng of vector elds v1 : : : vn : U M ! E is called a local frame for E on U if for each p 2 U the set f(v1)p : : : (vn)pg is a basis for the vector space Ep . ^) we denote by TM the tanExample 4.9. For a manifold (M A gent bundle of M given by TM = f(p v)j p 2 M v 2 TpM g and de ne the projection : TM ! M by : (p v) 7! p. ^ we de ne x : ;1(U ) ! R m R m by For a chart x : U ! R m in A m X @ j ) 7! (x(p) (v : : : v )): x : (p vk @x p 1 m k k=1

4. THE TANGENT BUNDLE

25

Then the collection ^ and W x(U ) R m openg f(x );1 (W ) TM j (U x) 2 A is a basis for a topology TTM on TM and ( ;1 (U ) x ) is a chart on the 2m-dimensional topological manifold (TM TTM ). If (U x) and (V y) ^ such that p 2 U \ V , then the transition map are two charts in A (y ) (x );1 : x ( ;1(U \ V )) ! R m R m is given by m m X @y1 (x;1 (a))b : : : X @ym (x;1(a))b ): (a b) 7! (y x;1 (a) k k k=1 @xk k=1 @xk We are assuming that y x;1 is smooth so it follows that (y ) (x );1 is also smooth. This means that ^ A = f( ;1(U ) x )j (U x) 2 Ag c ) has the structure of a smooth is a C 1-atlas on TM so (TM A manifold. It is trivial that the projection : TM ! M is smooth and surjective. For each point p 2 M the bre ;1 (p) of is the tangent space TpM of M at p so it is an m-dimensional vector space. For a chart ^ we de ne x : ;1 (U ) ! U R m by x : U ! R m in A m X @ j ) 7! (p (v : : : v )): x : (p vk @x p 1 m k k=1 The restriction xp = xjTpM : TpM ! fpg R m to TpM is given by m X @ j 7! (v : : : v ) xp : vk @x p 1 m k k=1 hence a vector space isomorphism. This implies that the map x : ;1(U ) ! U R m is a bundle chart. It is not di cult to see that ^ B = f( ;1(U ) x)j (U x) 2 Ag is a bundle atlas making (TM M ) into an m-dimensional topological vector bundle. It immediately follows from above that (TM M ) ^ de ned by B is a smooth together with the maximal bundle atlas B vector bundle. Example 4.10. Let S 3 be the unit sphere in C 2 on which we de ne the operation by (x y) ( ) = (x ; y y + x ):

26

4. THE TANGENT BUNDLE

It is easily checked that (S 3 ) is a Lie group with neutral element e = (1 0). Put v1 = (i 0), v2 = (0 1) and v3 = (0 i) and for k = 1 2 3 de ne the curves k : R ! S 3 by k : t 7! cos t (1 0) + sin t vk : Then k (0) = e and _ k (0) = vk for each k so v1 , v2 and v3 are elements of the tangent space TeS 3. The tangent vectors v1 , v2 and v3 are linearily independent so they span Te S 3. The left translations Lp : S 3 ! S 3 with Lp : h 7! p h induce vector elds X1 X2 X3 2 C 1(TS 3) by d (L ( (t)))j : (Xk )p = (dLp)e(vk ) = dt p k t=0 It is left as an exercise for the reader to show that at a point p = (x y) 2 S 3 the values of Xk at p is given by (X1)p = (x y) (i 0) = (ix iy) (X2)p = (x y) (0 1) = (;y x) (X3)p = (x y) (0 i) = (;iy ix): Lemma 4.11. Let M m be a smooth manifold and X : M ! TM be a continuous section on M . Then the following conditions are equivalent i. the section X is smooth, ii. if (U x) is a chart on M then the functions a1 : : : am : U ! R given by m X @ = Xj ak @x U k k=1 are smooth, iii. if f : V ! R de ned on an open subset V of M is smooth, then the function X (f ) : V ! R with X (f )(p) = Xp (f ) is smooth. Proof. i: ) ii: The functions ak = m+k x X jU : U ! TM ! x(U ) R m ! R are restrictions of compositions of smooth maps so therefore smooth. ii: ) iii: Let (U x) be a chart on M such that U is contained in P @f V . By assumption the map X (f jU ) = m i=1 ai @xi is smooth. This is true for each such chart (U x) so the function X (f ) is smooth. iii: ) i: Note that the smoothness of the section X is equivalent to x X jU : U ! R 2m being smooth for all charts (U x) on M . On the other hand, this is equivalent to xk = k x X jU : U ! R being smooth for all k = 1 2 : : : 2m and all charts (U x) on M . It is trivial that the coordinates xk = xk for k = 1 : : : m are smooth. But xm+k = ak = X (xk ) for k = 1 : : : m hence also smooth by assumption.

4. THE TANGENT BUNDLE

27

De nition 4.12. Let M be a smooth manifold. For two vector elds X Y 2 C 1(TM ) we de ne the Lie bracket X Y ]p of X and Y at p 2 M by X Y ]p(f ) = Xp(Y (f )) ; Yp(X (f )) 2 R where f 2 C 1(M R ). Lemma 4.13. Let M be a smooth manifold, X Y 2 C 1(TM ) be vector elds on M , f g 2 C 1 (M R ) and 2 R . Then i. X Y ]p( f + g) = X Y ]p(f ) + X Y ]p(g) ii. X Y ]p(f g) = f (p) X Y ]p(g) + g(p) X Y ]p(f ):
Proof.

= = = = =

X Y ]p( f + g) = Xp(Y ( f + g)) ; Yp(X ( f + g)) = Xp(Y (f )) + Xp(Y (g)) ; Yp(X (f )) ; Yp(X (g)) = X Y ]p(f ) + X Y ]p(g): X Y ]p(f g) Xp(Y (f g)) ; Yp(X (f g)) Xp(f Y (g) + g Y (f )) ; Yp(f X (g) + g X (f )) Xp(f )Yp(g) + f (p)Xp(Y (g)) + Xp(g)Yp(f ) + g(p)Xp(Y (f )) ;Yp(f )Xp(g) ; f (p)Yp(X (g)) ; Yp(g)Xp(f ) ; g(p)Yp(X (f )) f (p)fXp(Y (g)) ; Yp(X (g))g + g(p)fXp(Y (f )) ; Yp(X (f ))g f (p) X Y ]p(g) + g(p) X Y ]p(f ):

Then i. X Y ]p is an element of TpM for all p 2 M , ii. the section X Y ] : p 7! X Y ]p is smooth. Proof. The rst statement is a direct consequence of Lemma 4.13. It implies that X Y ] : M ! TM is a section of TM . If f : M ! R is a smooth function, then X Y ](f ) = X (Y (f )) ; Y (X (f )) is smooth. It then follows from Lemma 4.11 that the section X Y ] is smooth. Theorem 4.15. Let M be a smooth manifold. The vector space 1 C (TM ) of smooth vector elds on M equipped with the Lie bracket ] : C 1(TM ) C 1(TM ) ! C 1(TM ) is a Lie algebra over the real numbers i.e. if X Y Z 2 C 1 (TM ) and 2 R then i. X + Y Z ] = X Z ] + Y Z ],

Proposition 4.14. Let M be a manifold and X Y 2 C 1(TM ).

28

4. THE TANGENT BUNDLE

ii. X Y ] = ; Y X ], iii. X Y Z ]] + Z X Y ]] + Y Z X ]] = 0. Proof. See exercise 4.4. De nition 4.16. Let M be a smooth manifold. Two vector elds X Y 2 C 1(TM ) are said to commute if X Y ] = 0. Lemma 4.17. Let : M ! N be a smooth bijective map between two manifolds. If X Y 2 C 1 (TM ) are vector elds on M , then i. d (X ) 2 C 1(TN ), ii. the map d : C 1(TM ) ! C 1(TN ) is a Lie algebra homomorphism i.e. d (X ) d (Y )] = d ( X Y ]). 1 Proof. That d (X ) 2 C (TN ) follows directly from the fact that d (X )(f )( (p)) = X (f )(p): Let f : N ! R be a smooth function, then d (X ) d (Y )](f ) = d (X )(d (Y )(f )) ; d (Y )(d (X )(f )) = X (d (Y )(f ) ) ; Y (d (X )(f ) ) = X (Y (f )) ; Y (X (f )) = X Y ](f ) = d ( X Y ])(f ): This completes the proof. De nition 4.18. Let G be a Lie group with neutral element e. For x 2 G the left translation by x is the map Lx : G ! G de ned by Lx : y 7! xy. A vector eld X 2 C 1(TG) is said to be left-invariant if for all x y 2 G Xxy = (dLx)y (Xy ): We denote the set of all left-invariant vector elds on G by g. Note that if X 2 g, then Xx = (dLx)e(Xe). This implies that the value Xx of X at x 2 G is completely determined by the value Xe of X at e. Proposition 4.19. Let G be a Lie group and g be the set of all left-invariant vector elds on G. Then i. g is a Lie subalgebra of C 1(TG) i.e. if X Y 2 g then X Y ] 2 g, ii. the vector spaces g and TeG are isomorphic.

4. THE TANGENT BUNDLE


Proof.

29

If x y 2 G and f a smooth function on G, then ((dLx)y X Y ]y )(f ) = dLx(X ) dLx(Y )]y (f ) = dLx(X )y (dLx(Y )(f )) ; dLx(Y )y (dLx(X )(f )) = dLx(X )y (Y (f Lx )) ; dLx(Y )y (X (f Lx)) = X (Y (f Lx))(y) ; Y (X (f Lx))(y) = X Y ]xy (f ) for all X Y 2 g. This proves that X Y ] 2 g and thereby that the vector space g is a Lie subalgebra of C 1(TG). To see that g is isomorphic to the tangent space at the neutral element TeG note that the map : TeG ! g given by : Z 7! (Z : x 7! (dLx)e(Z )) is a vector space isomorphism. De nition 4.20. Let G be a Lie group and de ne a Lie bracket ] on TeG by X Y ] = X Y ]e. Then (TeG ]) = (g ]) is called the Lie algebra of G. Proposition 4.21. Let GL(R m ) be the general linear group of the real m-dimensional vector space R m . Then Te GL(R m ) = R m m and the Lie bracket ] : TeGL(Rm ) Te GL(R m ) ! TeGL(R m ) de ned above is given by A B ] = AB ; BA. Proof. See exercise 4.6. The Lie algebras of the matrix groups introduced in Example 2.23 are denoted by gl(R m ), sl(Rm ), o(m), so(m), gl(C m ), sl(R m ), u(m) and su(m). Theorem 4.22. Let G be a Lie group. Then the tangent bundle TG is trivial. Proof. Let fX1 : : : Xm g be a basis for Te G. Then the map : TG ! G R m given by : (p
m X k=1

vk (Xk )p) 7! (p (v1 : : : vm))

is a global bundle chart so the tangent bundle TG is trivial.

30

4. THE TANGENT BUNDLE

Exercises

^) be a smooth manifold and (U x), (V y) Exercise 4.1. Let (M m A ^ be two charts in A such that U \ V 6= . Let f = y x;1 : x(U \ V ) ! related by
@ j i = 1 : : : mg and f @ j j = 1 : : : mg for TM on U \ V are f @x @yj i
Rm

be the corresponding transition map. Show that the local frames

m @ (fj x) @ : @ =X @xi j=1 @xi @yj Exercise 4.2. Let O(m) be the orthogonal group. i. Find a basis for the tangent space TeO(m), ii. construct a non-vanishing vector eld Z 2 C 1(T O(m)), iii. determine all smooth vector elds on O(2). The Hairy Ball Theorem. If m 2 N + then there does not exist a non-vanishing vector eld X 2 C 0 (TS 2m). Exercise 4.3. Let m 2 N + . Use the Hairy Ball Theorem to prove that the tangent bundle TS 2m of S 2m is not trivial. Construct a nonvanishing vector eld X 2 C 1(TS 2m+1 ). Exercise 4.4. Find a proof for Theorem 4.15. Exercise 4.5. Let f@=@xk j k = 1 : : : mg be the standard global frame for T R m . Let X Y 2 C 1(T Rm ) be two vector elds given by m m X X @ @ X= and Y = k k k=1 @xk k=1 @xk where k k 2 C 1(R m R ). Calculate the Lie brackets @=@xk @=@xl ] and X Y ]. Exercise 4.6. Find a proof for Proposition 4.21.

CHAPTER 5

Immersions, Embeddings and Submersions


called i. an immersion if for each p 2 M the di erential d p : TpM ! T (p) N is injective, ii. an embedding if it is an immersion and a homeomorphism onto its image (M ), iii. a submersion if for each p 2 M the di erential d p is surjective. Example 5.2. Let S 1 be the unit circle in C . For each positive natural number k de ne k : S 1 ! C and k : S 1 ! S 1 by k k k : z 7! z : For a point p 2 S 1 let : R ! S 1 be the curve with : t 7! eit p. Then (0) = p and _ (0) = ip. For the di erentials of k and k we have d ( (t))j = d (eiktpk )j = ikpk : (d k )p( _ (0)) = (d k )p( _ (0)) = dt k t=0 t=0 dt The di erentials (d k )p : TpS 1 = R 1 ! Tpk S 1 = R are all bijective, so the maps k are both immersions and submersions. The only one that is an embedding is 1 . The di erentials (d k )p : TpS 1 ! Tpk C = R 2 are all injective and not surjective. This means that the maps k are all immersions, but none of them is a submersion. The only one that is an embedding is 1. Example 5.3. For p 2 S m let p : R m+1 ! R m+1 be the re ection about the line ftp 2 R m+1 j t 2 R g spanned by p. Then de ne the map : S m ! R (m+1) (m+1) by : p 7! ( p : q 7! 2hq pip ; q): Then the matrix describing the linear map p is given by (2ppt ; I ) 2 R (m+1) (m+1) , since 2hq pip ; q = 2php q i ; q = (2ppt ; I )q Proposition 5.4. The map is an immersion and the image (S m) is di eomorphic to the m-dimensional real projective space R P m and lies in Sym(R m+1 ) \ O(m + 1).
31

De nition 5.1. A map : M m ! N n between two manifolds is

32
Proof.

5. IMMERSIONS, EMBEDDINGS AND SUBMERSIONS

Let p be an arbitrary point on S m and : I ! S m be two curves meeting at p, that is (0) = p = (0), with a = _ (0) and b = _ (0). For 2 f g we have : t 7! (q 7! 2hq (t)i (t) ; q) so d ( (t))j = (q 7! 2hq _ (0)i (0) + 2hq (0)i _ (0)): (d )p( _ (0)) = dt t=0 This means that d p(a) = (q 7! 2hq aip + 2hq pia) d p(b) = (q 7! 2hq bip + 2hq pib): If a 6= b then d p(a) 6= d p(b) so the di erential d p is injective. This proves that is an immersion. If two points p q 2 S m are linearily independent, then p 6= q since their images are di erent. On the other hand, if p = q then p = q . This means that the image (S m ) is di eomorphic to the quotient space S m = where is the equivalence relation de ned by x y if and only if x = y. This proves that (S m) = R P m . It is obvious that (S m) lies on Sym(R m+1 ) \ O(m + 1). Corollary 5.5. The map ^ : R P m ! Sym(R m+1 ) \ O(m +1) given by ^ : p] 7! (2ppt ; I ) is an embedding. The following result was proved by H. Whitney in his very famous paper, Di erentiable Manifolds, Ann. of Math. 37 (1936), 645-680. Deep Result 5.6. For 1 r 1 let M m be an m-dimensional C r -manifold. Then there exists a C r -embedding : M ! R 2m+1 into the (2m + 1)-dimensional real vector space R2m+1 . De nition 5.7. Let : M m ! N n be a map between manifolds. A point p 2 M is called a critical point if the di erential d p : TpM ! T (p) N is not of full rank, and a regular point if it is not critical. A point q 2 (M ) is called a regular value if every point on the pre-image ;1(fqg) is regular.

Theorem 5.8 (The Implicit Function Theorem). Let : M m ! be a map between two manifolds. If q 2 (M ) is a regular value, then the pre-image ;1(fq g) is an (m ; n)-dimensional submanifold of M m . The tangent space Tp ;1(fqg) of ;1(fqg) at p is the kernel of the di erential d p i.e. Tp ;1(fqg) = Ker d p.
Nn

5. IMMERSIONS, EMBEDDINGS AND SUBMERSIONS

33

Proof. Let (Vq q ) be a chart on N with q 2 Vq and q (q ) = 0. For a point p 2 ;1(fqg) we choose a chart (Up p) on M such that p 2 Up, p(p) = 0 and (Up) Vq . The di erential of the map ;1 n ^= q p j p (Up ) : p (Up ) ! R at the point 0 is given by ;1)0 : T0 R m ! T0 R n : d ^0 = (d q )q d p (d p The pairs (Up p) and (Vq q ) are charts so the di erentials (d q )q and ;1)0 are bijective. This means that the di erential d ^0 is surjective (d p since d p is. It then follows from the implicit function theorem 2.13 that p( ;1(fqg)\Up) is an (m;n)-dimensional submanifold of p (Up). Hence ;1(fqg) \ Up is an (m ; n)-dimensional submanifold of Up. This is true for each point p 2 ;1(fqg) so we have proved that ;1(fqg) is an (m ; n)-dimensional submanifold of M m . Let : (; ) ! ;1(fqg) be a curve, such that (0) = p. Then (d )p( _ (0)) = d ( (t))jt=0 = dq jt=0 = 0: dt dt This implies that Tp ;1 (fqg) is contained in, and has the same dimension as the kernel of d p, so Tp ;1(fqg) = Ker d p. De nition 5.9. Let M m be a smooth manifold and U be an open subset of R m . An immersion : U ! M is called a local parametrization of M . Example 5.10. Let (U ) be a chart on M m . Then the inverse ;1 : (U ) ! U of is a local parametrization of U M . Example 5.11. Let S 3 and S 2 be the unit spheres in C 2 and C R = R 3 , repectively. Let : S 3 ! S 2 be the map given by : 2 2 (x y) 7! (2xy jxj ; jyj ). Then one easily shows that and d p : TpS 3 ! T (p) S 2 are surjective for each p 2 S 3. This implies that each point q 2 S 2 is a regular value and the bres of are 1-dimensional submanifolds of S 3. They are the great circles given by ;1(f(2xy jxj2 ; jyj2)g) = fei (x y)j 2 Rg: This means that S 3 is the union of disjoint great circles ;1(fqg): S3 = q2S 2 It is also what is called a S 1-bundle over S 2.

34

5. IMMERSIONS, EMBEDDINGS AND SUBMERSIONS

Exercises

C N 0 are k k immersions, submersions or embeddings. Exercise 5.2. Let S 2 and S 3 be the unit spheres of R 3 and C 2 , respectively. The Hopf-map : S 3 ! S 2 is de ned by : (x y) 7! (2xy jxj2 ; jyj2). Prove that is a submersion.

Exercise 5.1. For each k 2 N 0 de ne by k k : z 7! zk . For which k 2

k:C

! C and

k:C

CHAPTER 6

Riemannian Manifolds
Let M be a smooth manifold and as before we denote by C 1(TM ) 1(TM ) = C 1(M R ) be the ring the set of vector elds on M . Let C0 of all smooth functions de ned on M . For k 2 N + let k 1(TM ) = O C 1(TM ) Ck

1(TM ) is a C01(TM ) be the k-fold tensor product of C 1(TM ). Then Ck module in the trivial way. A tensor eld B on M of type (r s) is 1(TM ) i.e. an r-linear map B : Cr1(TM ) ! Cs1(TM ) over the ring C0 B (X1 : : : Xl;1 f Xl + g Y Xl+1 : : : Xr ) = f B (X1 : : : Xr ) + g B (X1 : : : Xl;1 Y Xl+1 : : : Xr ) 1(TM ) and l = 1 : : : r. for all X1 : : : Xr Y 2 C 1(TM ), f g 2 C0 Proposition 6.1. Let B : Cr1(TM ) ! Cs1(TM ) be a tensor eld of type (r s) and p 2 M . Let X1 : : : Xr and Y1 : : : Yr be smooth vector elds on M such that (Xk )p = (Yk )p for each k = 1 : : : r. Then B (X1 : : : Xr )(p) = B (Y1 : : : Yr )(p): Proof. It is su cient to prove the statement for r = 1 since the rest follows by induction. Put X = X1 and Y = Y1. Let (U x) be 1(TM ) such that a local coordinate on M . Choose a function f 2 C0 f (p) = 1 and support(f ) is contained in U . Then de ne v1 : : : vm 2 C 1(TM ) by @ j if q 2 U q) @x k q (vk )q = f (0 if q 2 =U 1 Then there exist functions k k 2 C0 (TM ) such that
f X=
Now
m X k=1 k vk

l=1

and f Y =

m X k=1

k vk :

B (X )(p) = f (p)B (X )(p) = B (f X )(p) =


35

m X k=1

k (p)B (vk )(p)

36

6. RIEMANNIAN MANIFOLDS

and similarily B (Y )(p) = m k=1 k (p)B (vk )(p). Now Xp = Yp so ( p ) = ( p ) for all k . This implies that B (X )(p) = B (Y )(p). k k By Bp we denote the restriction Bp = B j r l=1 Tp M of B to the r -fold tensor product of TpM given by Bp : ((X1)p : : : (Xr )p) 7! B (X1 : : : Xr )(p): The tensor eld B is said to be smooth if for all X1 : : : Xr 2 C 1(TM ) the map B (X1 : : : Xr ) : M ! Cs1(TM ) with B (X1 : : : Xr ) : p 7! Bp((X1)p : : : (Xr )p) is smooth. De nition 6.2. Let M be a smooth manifold. A Riemannian metric on M is a smooth tensor eld g : C21(TM ) ! C01(TM ) such that for each p 2 M the restriction gp = gjTpM TpM : TpM TpM ! R with gp : (Xp Yp) 7! g(X Y )(p) is an inner product on TpM . The pair (M g) is called a Riemannian manifold. The study of Riemannian manifolds is called Riemannian Geometry. Geometric properties of (M g) which only depend on the metric g are called intrinsic (or metric) properties. De nition 6.3. Let (M g) be a Riemannian manifold and : I ! M be a curve in M . Then the length L( ) of is de ned by Zp g( 0(t) 0(t))dt: L( ) =

Example 6.4. By the m-dimensional Euclidean space we mean the Riemannian manifold E m = (R m h iRm ) where

hu viRm =

m X k=1

uk vk :

Let : R + !

Example 6.5. By the punctured round sphere we mean the Riemannian manifold 4 m = (R m m (1 + jxj2 )2 h iR ):
be the curve with : t 7! (t 0 : : : 0). Then Z 1 ph 0 0 i Z 1 dt L( ) = 2 dt =2 = 2 arctan(t)]1 0 = 2 2 0 1+j j 0 1+t
m
Rm

6. RIEMANNIAN MANIFOLDS

37

4 m (0) H m = (B1 2 h iRm ) (1 ; jxj2 Rm ) m (0) is the m-dimensional open unit ball where B1 m (0) = fx 2 R m j jxj m < 1g: B1 R m Let : (0 1) ! H be a curve given by : t 7! (t 0 : : : 0). Then Z 1 ph 0 0 i Z 1 dt 1 + t )]1 = 1 L( ) = 2 1 ; j j2 dt = 2 1 ; t2 = log( 1 ;t 0 0 0 ~ ~m De nition 6.7. Let (M m g) be a Riemannian manifold and M be an m ~ -dimensional submanifold of M . Then the smooth tensor eld 1(T M ~ ) ! C0 ~ ) with h : C21(T M h(X Y ) : p 7! gp(Xp Yp): ~ called the induced metric on M ~ in is a Riemannian metric on M (M g). Example 6.8. The Euclidean metric h iRm on R m induces Riemannian metrics on the following submanifolds. i. the (m ; 1)-dimensional sphere S m;1 R m , ii. the tangent bundle TS n R m where m = 2n + 2, iii. the n-dimensional torus T n R 2n , iv. the n-dimensional real projective space R P n Sym(R n+1 ) R m where m = (n + 2)(n + 1)=2. Example 6.9. On C n n we have the Euclidean metric given by hA B i = Reftrace(At B )g: This induces metrics on submanifolds of C n n such as R n n and all the matrix Lie groups GL(C n ), SL(C n ), U(n), SU(n), GL(R n ), SL(Rn ), O(n) and SO(n). We now need the following fact which should be known to any graduate student from a course on topology.

ian manifold

Example 6.6. By the hyperbolic space we mean the Riemann-

Fact 6.10. Every locally compact Hausdor space with countable basis is paracompact. ^) be a topological manifold. Let the colCorollary 6.11. Let (M A lection (U ) 2I be an open covering of M such that for each 2 I the pair (U ) is a chart on M . Then there exists i. a locally nite open re nement (W ) 2J such that for all 2 J , ^, and W is an open neighbourhood for a chart (W ) 2 A

38

6. RIEMANNIAN MANIFOLDS

ii. a partition of unity (f ) 2J such that support(f ) W . ^) be a smooth manifold. Then there Theorem 6.12. Let (M m A exists a Riemannian metric g on M . Proof. For each p 2 M let (Up p) be a chart such that p 2 Up . Then (Up)p2M is an open covering as in Corollary 6.11. Let (W ) 2J be a locally nite open re nement, (W x ) be charts on M and (f ) 2J be a partition of unity such that support(f ) is contained in W . Let h1 iRm be the Euclidean metric on R m . Then for 2 J de ne g : 1 C2 (TM ) ! C0 (TM ) by

p) hek el iRm if p 2 W g ( @ @ )(p) = f ( 0 if p 2 =W @xk @xl P 1(TM ) ! C01(TM ) given by g = 2J g is a Riemannian Then g : C2 metric on M . De nition 6.13. Let (M g) and (N h) be Riemannian manifolds. A map : (M g) ! (N h) is said to be conformal if there exists a function : M ! R such that e (p) gp(Xp Yp) = h (p) (d p(Xp) d p(Yp)) for all X Y 2 C 1(TM ) and p 2 M . The function e is called the conformal factor of . A conformal map with 0 is said to be isometric. An isometric di eomeorphism is called an isometry. Example 6.14. Equip the orthogonal group O(m) R m m with the induced metric given by hA B i := trace(At B ). For x 2 O(m) the left translation Lx : O(m) ! O(m) by x is given by Lx : y 7! xy. The tangent space Ty O(m) of O(m) at y is Ty O(m) = fy Z j Z + Z t = 0g and the di erential (dLx)y : Ty O(m) ! Txy O(m) is given by (dLx)y : yZ 7! xyZ . We then have h(dLx)y (yZ ) (dLx)y (yW )ixy = trace((xyZ )txyW ) = trace(Z tytxt xyW ) = trace(yZ )t(yW ): = hyZ yW iy This shows that for each x 2 O(m) the left translation Lx : O(m) ! O(m) is an isometry. De nition 6.15. Let G be a Lie group. A Riemannian metric h i on G is said to be left-invariant if for each x 2 G the left-translation Lx : G ! G is an isometry.

6. RIEMANNIAN MANIFOLDS

39

Proposition 6.16. Let G be a Lie group and h ie be an inner product on the tangent space Te G at the neutral element e. Then for each x 2 G the bilinear map h ix : TxG TxG ! R with h(dLx)e(Ze) (dLx)e(We)ix = hZe Weie is an inner product on the tangent space Tx G. The smooth tensor eld h i : C21(TG) ! C01(TG) given by h i : (Z W ) 7! (hZ W i : x 7! hZx Wxix) is a left-invariant Riemannian metric on G. Proof. See Exercise 6.3. Example 6.17. Let (S m h iRm+1 ) be the standard sphere. Let the linear space of symmetric R (m+1) (m+1) matrices Sym(R m+1 ) be equipped with the metric g given by t g(A B ) = 1 8 trace(A B ): As in Example 5.3 de ne a map : S m ! Sym(R m+1 ) by : p 7! ( p : q 7! 2hq pip ; q): Let : R ! S mbe two curves such that (0) = p = (0) and put a = 0(0), b = 0(0). Then for 2 f g we have d p( 0 (0)) = (q 7! 2hq 0(0)ip + 2hq pi 0(0)): Let B be an orthonormal basis for R m+1 , then 1 trace(d (a)t d (b)) g(d p(a) d p(b)) = 8 p p X = 1 2 q2B hhq aip + hq pia hq bip + hq pibi 1 Xfhp piha qihq bi + ha bihp qihp qig = 2 q2B = 1 2 fha bi + ha big = ha bi This proves that the immersion is isometric. The following result was proved by J. Nash in his famous paper: The embedding problem for Riemannian manifolds, Ann. of Math. 63 (1956), 20-63. It implies that every Riemannian manifold can be realized as a submanifold of a Euclidean space.

40

6. RIEMANNIAN MANIFOLDS

clidean space. We will now see that parametrizations can be very useful tools for the study of the intrinsic geometry of a Riemannian manifold (M gM ). Let p 2 M , ^ : U ! M be a parametrization of M with q 2 U and ^(q) = p. The di erential d ^q : Tq R m ! TpM is bijective so there exist neighbourhoods Uq of q and Up of p such that the restriction = ^jUq : Uq ! Up is a di eomorphism. On Uq we have the frame fe1 : : : emg for TUq so fd ^(e1) : : : d ^(em)g is a local frame for TM over Up. We then de ne the pull-back metric g = ^ gM on Uq by gkl = g(ek el ) = gM (d ^(ek ) d ^(el )): Then ^ : Uq ! Up is an isometry so the intrinsic geometry of (Uq g) and that of (Up gM ) are exactly the same. Example 6.19. Let G be a matrix Lie group and e be the neutral element of G. Let fX1 : : : Xmg be a basis for the Lie algebra TeG. For x 2 G de ne x : R m ! G by
x

Deep Result 6.18. For 3 r 1 let (M g) be a Riemannian C r -manifold. Then there exists an isometric C r -embedding into a Eu-

: (t1 : : : tm) 7! Lx (

m Y

Then (d x)0(ek ) = Xk (x) for all k. This means that the di erential (d x)0 : T0 Rm ! TxG is an isomorphism so there exist open neighbourhoods U0 of 0 and Ux of x such that the restriction of to U0 is bijective onto its image Ux . ~ be De nition 6.20. Let (M g) be a Riemannian manifold and M ~ a submanifold of M . For a point p 2 M we de ne the normal space ~ of M ~ at p by NpM ~ = fv 2 TpM j gp(v w) = 0 for all w 2 TpM ~ g: NpM For all p we have the orthogonal decomposition ~ NpM: ~ TpM = TpM ~ in M is de ned by The normal bundle of M ~ = f(p v)j p 2 M ~ v 2 NpM ~ g: NM ~ ~m Theorem 6.21. Let (M m g) be a Riemannian manifold and M ~ M ~ ) be a smooth submanifold of M . Then the normal bundle (N M ~. is a smooth (m ; m ~ )-dimensional vector bundle over M

k=1

exp(tk Xk )):

6. RIEMANNIAN MANIFOLDS
Proof.

41

See Exercise 6.5. Example 6.22. The orthogonal group O(m) is a subset of the linear space R m m equipped with the Riemannian metric hA B iRm m = trace(At B ): We have already seen that the tangent space TeO(m) of O(m) at the neutral element e is Te O(m) = fZ 2 R m m jZ + Z t = 0g and the tangent bundle T O(m) of O(m) is given by T O(m) = f(x xZ )j x 2 O(m) Z 2 TeO(m)g: The space R m m has a linear decomposition R m m = Sym(R m ) Te O(m) and every element X 2 R m m can be decomposed X = X N + X T in its symmetric and skew-symmetric parts given by 1 (X + X t) and X T = 1 (X ; X t): XN = 2 2 m If Z 2 TeO(m) and W 2 Sym(R ) then hZ W iRm m = trace(Z tW ) = trace(W t Z ) = trace(ZW t) = trace(;Z t W ) = ;hZ W iRm m This implies that the normal bundle N O(m) of O(m) in Rm m is given by N O(m) = f(x xW )j x 2 O(m) W 2 Sym(R m )g:

42

6. RIEMANNIAN MANIFOLDS

Exercises
+1

1 (x : : : x ): : (x0 : : : xm ) 7! 1 ; m x0 1 Prove that m is an isometry for each m. Exercise 6.2. Let B12 (0) be the open unit disk in the complex plane equipped with the hyperbolic metric g( ) = 4=(1 ; jzj2 )2h iR2 . Prove that the map 1 h i 2) 2 (0) ! (fz 2 C j Im(z ) > 0g : B1 Im(z)2 R given by : z 7! (z + i)=(iz + 1) is an isometry. Exercise 6.3. Find a proof for Proposition 6.16. Exercise 6.4. On the real general linear group GL(R m ) we de ne metrics g h by gx(xZ xW ) = trace((xZ )t xW ) and hx(xZ xW ) = trace(Z t W ): They induce metrics g ^^ h on the real special linear group SL(R m ). ^ are left-invariant? i. Which of the metrics g h g ^h ii. Find the normal space NeSL(R m ) of SL(Rm ) in GL(R m ) w.r.t g iii. Find the normal bundle N SL(R m ) of SL(R m ) in GL(R m ) w.r.t h. Exercise 6.5. Find a proof for Theorem 6.21.
m

be given by

Exercise 6.1. For m 2 N + let the stereographic projection 4 h i m) m m m : (S ; f(1 0 : : : 0)g h iRm ) ! (R (1 + jxj2)2 R

CHAPTER 7

The Levi-Civita Connection


De nition 7.1. Let (E M ) be a smooth vector bundle over M . ^ : C 1(TM ) C 1(E ) ! C 1(E ) A connection on (E M ) is a map r
such that ^ ( v + w) = r ^ v+ r ^ w, i. r X X X ^X(f v) = X (f ) v + f r ^Xv, ii. r ^(f X + g Y )v = f r ^Xv + g r ^Y v. iii. r for all 2 R , X Y 2 C 1(TM ), v w 2 C 1(E ) and f g 2 C01(TM ), ^Xv) : M ! E with (r ^Xv) : p 7! (r ^Xv)p is smooth. A and the map (r 1 section v 2 C (E ) is said to be parallel with respect to the connection ^ if r ^ v = 0 for all X 2 C 1(TM ). r X ^ be a connecDe nition 7.2. Let M be a smooth manifold and r tion on the tangent bundle (TM M ). Then we de ne the torsion ^ by T : C21(TM ) ! C11(TM ) of r ^ Y ;r ^ X ; X Y] T (X Y ) = r X Y 1 where ] is the Lie bracket on C (TM ). ^ De nition 7.3. Let M be a smooth manifold. A connection r on the tangent bundle (TM M ) is said to be torsion-free if the corresponding torsion T vanishes i.e. T (X Y ) = 0 for all X Y 2 ^ is said to be C 1(TM ). If g is a Riemannian metric on M , then r metric (or compatible with g) if ^ Y Z ) + g(Y r ^ Z) X (g(Y Z )) = g(r X X for all X Y Z 2 C 1(TM ). Lemma 7.4. Let M be a smooth manifold and ] be the Lie bracket on the tangent bundle TM . Then i. X f Y ] = X (f ) Y + f X Y ], ii. f X Y ] = ;Y (f ) X + f X Y ] 1(TM ), for all X Y 2 C 1(TM ) and f 2 C0
43

44
Proof.

1(TM ), then If h 2 C0

7. THE LEVI-CIVITA CONNECTION

X f Y ](h) = X (f Y (h)) ; f Y (X (h)) = X (f ) Y (h) + f X (Y (h)) ; f Y (X (h)) = (X (f ) Y + f X Y ])(h)


This proves the rst statement and the second follows from the skewsymmetry of the Lie bracket.

Theorem 7.5. Let (M g) be a Riemannian manifold and let the 1(TM ) ! C 1(TM ) be given by map r : C2
1 g(r XY Z ) = 2 fX (g(Y Z )) + Y (g(Z X )) ; Z (g(X Y )) +g(Z X Y ]) + g(Y Z X ]) ; g(X Y Z ])g:
Then r is a connection on the tangent bundle (TM M ).

It follows from De nition 3.1, Theorem 4.15 and the fact that g is a tensor eld that
Proof.

g(r X( Y1 +
and for all = = = =

Y2) Z ) =

g(r XY1 Z ) +

g(r XY2 Z )

2 R and X Y1 Y2 Z 2 C 1(TM ). Furthermore we have


g(r XfY Z ) 1 fX (f g(Y Z )) + f Y (g(Z X )) ; Z (f g(X Y )) 2 +g(Z X f Y ]) + f g(Y Z X ]) ; g(X f Y Z ])g 1 fX (f ) g(Y Z ) + f X (g(Y Z )) + f Y (g(Z X )) 2 ;Z (f ) g(X Y ) ; f Z (g(X Y )) + g(Z X (f ) Y + f X Y ]) +f g(Y Z X ]) ; g(X ;Z (f ) Y + f Y Z ])g X (f ) g(Y Z ) + f g(r XY Z ) g(X (f ) Y + f r XY Z )

g(r Y1 + Y2X Z ) = g(r Y1X Z ) + g(r Y2X Z )

7. THE LEVI-CIVITA CONNECTION

45

g(r f XY Z ) = 1 2 ff X (g(Y Z )) + Y (f g(Z X )) ; Z (f g(X Y )) +g(Z f X Y ]) + g(Y Z f X ]) ; f g(X Y Z ])g = 1 ff X (g(Y Z )) + Y (f ) g(Z X ) + f Y (g(Z X )) 2 ;Z (f ) g(X Y ) ; f Z (g(X Y )) +g(Z ;Y (f ) X ) + g(Z f X Y ]) + g(Y Z (f ) X ) f g(Y Z X ]) ; f g(X Y Z ])g = f g(r XY Z ): This proves that r is a connection on the tangent bundle (TM M ).

and

Levi-Civita connection is a unique torsion-free and metric connection on (TM M ). Proof. The di erence g (r Y Z ) ; g (r X Z ) equals twice the X Y skew-symmetric (w.r.t the pair (X Y )) part of the right hand side in Theorem 7.5. This is the same as = 1 fg(Z X Y ]) ; g(Z Y X ])g = g(Z X Y ]): 2 This proves that the Levi-Civita connection is torsion-free. The sum g(r XY Z )+ g(r XZ Y ) equals twice the symmetric (w.r.t the pair (Y Z )) part on the right hand side of Theorem 7.5. This is exactly 1 fX (g(Y Z )) + X (g(Z Y ))g = X (g(Y Z )): =2 This shows that the Levi-Civita connection is compatible with the Riemannian metric g on M . ^ is a torsion-free and metric connection. Let us now assume that r Then it is easily seen that the following equations hold ^XY Z ) = X (g(Y Z )) ; g(Y r ^XZ ) g(r

De nition 7.6. The connection r : C 1(TM ) C 1 (TM ) ! C11(TM ) de ned in Theorem 7.5 is called the Levi-Civita connection. The next result is called The Fundamental Theorem of Riemannian Geometry Theorem 7.7. Let (M g) be a Riemannian manifold. Then the

46

7. THE LEVI-CIVITA CONNECTION

^XY Z ) = g( X Y ] Z ) + g(r ^Y X Z ) g(r ^Y Z ) = g( X Y ] Z ) + Y (g(X Z )) ; g(X r ^ZX Y ) + g(X r ^ZY ) 0 = ;Z (g(X Y )) + g(r ^XZ + Z X ] Y ) + g(X r ^Y Z ; Y Z ]): = ;Z (g(X Y )) + g(r We add these equations and obtain ^XY Z ) = fX (g(Y Z ) + g( X Y ] Z ) + Y (g(X Z )) 2 g(r ;Z (g(X Y )) + g(( Z X ] Y ) ; g(X Y Z ])g = 2 g(r XY Z ): ^ = r and thereby proves the uniqueness of r. This implies that r ^ on (TM M ) can be thought of as a rule for di erA connection r entiating a vector eld Y 2 C 1(TM ) in the direction of another X 2 ^XY . The last theorem shows that given a Riemannian C 1(TM ) by r metric g on M there is only one way of doing this in a metric and torsion-free manner. The Levi-Civita connection r is by de nition determined by the metric g so it is an intrinsic object. De nition 7.8. Let G be a Lie group. For a left-invariant vector eld X 2 g we de ne a map ad(X ) : g ! g by ad(X ) : Z 7! X Z ]: Proposition 7.9. Let (G h i) be a Lie group equipped with a leftinvariant metric such that for all X 2 g, ad(X ) is skew-symmetric with
respect to h i i.e. for all X Y Z 2 g. Then the Levi-Civita connection of (G h i) is given 1 by r XY = 2 X Y ] for all left-invariant X Y 2 g. Proof. If X Y Z 2 g then it follows from the fact that h i is leftinvariant that the function hY Z i : G ! R is constant so X (hY Z i) = 0. It then follows from the de nition of the Levi-Civita connection and the fact that ad is skew-symmetric that 1 fhY Z X ]i ; hX Y Z ]ig hrXY ; 1 X Y ] Z i = 2 2 = 1 fhY ad(Z )X i + had(Z )Y X ig 2 = 0

had(X )Y Z i = ;hY ad(X )Z i

7. THE LEVI-CIVITA CONNECTION

47

Example 7.10. Let (M gM r) be a Riemannian manifold with Levi-Civita connection. Further let (U x) be a local coordinate on M @ 2 C 1(TU ). Then fX : : : X g is a local frame of and de ne Xi = @x 1 m i TM on U . For (U x) we de ne the Christo el symbols ;k ij : U ! R of r with respect to (U x) by
m X

On the subset x(U ) of

de ne the metric g by @ @ ): gij = g(ei ej ) = gM ( @x i @xj Following Lemma 4.17 and Exercise 4.5 we obtain for the di erential dx of x dx( Xi Xj ]) = dx(Xi ) dx(Xj )] = ei ej ] = 0 so Xi Xj ] = 0 since dx is bijective. From the de nition of the LeviCivita connection we get
m X k=1

k=1 m R we

;k ij Xk = r XiXj :

;k ij gkl

= h

m X k=1

;k ij Xk Xl i

If gkl = (g;1)kl then ;k ij

= hr XiXj Xli 1 fX hX X i + X hX X i ; X hX X ig = 2 i j l j l i l i j @gjl + @gli ; @gij g: f = 1 2 @x @x @x


i j l m X 1 kl f @gjl + @gli ; @gij g: = g 2 l=1 @xi @xj @xl

~ be a submanDe nition 7.11. Let M be a smooth manifold, M 1 ~ ~ ~ ifold and X 2 C (T M ) be a vector eld on M . Let U be an open ~= ~ on U is a subset of M such that U \ M 6 . A local extension of X 1 ~p = Xp for all p 2 M ~ . If U = M vector eld X 2 C (TU ) such that X then X is called a global extension. ~ 2 C 1(T M ~ ) has a global extension Fact 7.12. Every vector eld X 1 X 2 C (TM ). ~ be a Remark 7.13. Let (M g) be a Riemannian manifold and M 1 submanifold equipped with the induced metric g ~. Let Z 2 C (TM ) ~ ~ be a vector eld on M and Z = Z jM ~ : M ! TM be the restriction

48

7. THE LEVI-CIVITA CONNECTION

~ . Note that Z ~ is not necessarily an element of C 1(T M ~ ). of Z to M ~ the tangent vector Z ~p 2 TpM can be decomposed For each p 2 M ~p = (Z ~p)T + (Z ~p)N in a unique way such that (Z ~p)T 2 TpM ~ and Z N T N ~p) 2 NpM ~ . For the vector eld we write Z ~=Z ~ +Z ~ . (Z ~ Y ~ 2 C 1(T M ~ ) be vector elds on M ~ and X Y 2 C 1(TM ) Let X ~ then (r be their extensions onto M . If p 2 M XY )p only depends on the ~p and the value of Y along some curve : (; ) ! M value Xp = X ~p. For this see Remark 8.3. such that (0) = p and _ (0) = Xp = X Hence we can choose the curve such that ((; )) is contained in ~ . Then Y ~ (t) = Y (t) for t 2 (; ). This implies that (r M XY )p only ~p and the value of Y ~ along , but not on the way X ~ and depends on X ~ are extended. This implies that the following maps are well-de ned. Y De nition 7.14. For the above situation we de ne 1(T M 1(T M 1 (T M 1(N M ~ : C2 ~ ) ! C1 ~ ) and B : C2 ~ ) ! C1 ~) r with T N ~X ~ = (r ~ Y ~ ) = (r r ~Y XY ) and B (X XY ) : It is easily proved that B is symmetric and hence tensorial in both its arguments, see Exercise 7.4. B is called the second fundamental ~ in (M g). form of M The Levi-Civita connection on (M g) induces the Levi-Civita con~ and a metric connection on its normal nection on any submanifold M ~: bundle N M ~ be Theorem 7.15. Let (M g) be a Riemannian manifold and M 1(T M ~ : C2 ~) ! a submanifold of M with the induced metric g ~. Then r 1 ~ ) is the Levi-Civita connection of the submanifold (M ~ g C1 (T M ~). Proof. See Exercise 7.5. ~ Proposition 7.16. Let (M g) be a Riemannian manifold and M 1 be a submanifold with the induced metric g ~. Let X Z 2 C (TM ) be 1 ~ ~ ~ 2 C 1(N M ~ ). Then the vector elds extending X 2 C (T M ) and Z ~ ) C 1 (N M ~ ) ! C 1 (N M ~ ) given by map r : C 1(T M N ~ = (r rX ~Z XZ ) ~ . Furthermore is a well-de ned connection on the normal bundle N M ~ (hZ ~W ~ i) = hr ~W ~ i + hZ ~r ~i X ~Z ~W X X ~ 2 C 1(T M ~ ) and Z ~W ~ 2 C 1(N M ~ ). for all X Proof. See Exercise 7.6.

7. THE LEVI-CIVITA CONNECTION

Exercises

49

eld of type (2 1). with the metric

Exercise 7.1. Prove that the torsion T in De nition 7.2 is a tensor

Exercise 7.2. Let SO(m) be the special orthogonal group equipped

Prove that h i is left-invariant and that for left-invariant vector elds 1 X Y 2 so(m) we have r XY = 2 X Y ]. Let A B C be elements of the Lie algebra so(3) with 0 0 ;1 0 1 0 0 0 ;1 1 00 0 0 1 Ae = @ 1 0 0 A Be = @ 0 0 0 A Ce = @ 0 0 ;1 A : 0 0 0 1 0 0 0 1 0 Prove that fA B C g is an orthonormal basis for so(3) and calculate (r AB )e, (r BC )e and (r CA)e. Exercise 7.3. Let SL(R 2 ) be the real special linear group equipped with the metric hX Y ip = trace((p;1X )t (p;1Y )): Find a formula for the Levi-Civita connection r 2 sl(R 2 ). XY for X Y 2 Calculate (r AB )e (r BC )e and (r CA)e where A B C 2 sl(R ) are given by ;1 Be = 0 1 Ce = 1 0 : Ae = 0 1 0 1 0 0 ;1

hX Y i = 1 trace(X t Y ): 2

Exercise 7.4. Let (M g) be a Riemannian manifold with Levi~ g Civita connection r and (M ~) be a submanifold with the induced ~ in M is metric. Prove that the second fundamental form B of M symmetric and tensorial in both its arguments. Exercise 7.5. Find a proof for Theorem 7.15. Exercise 7.6. Find a proof for Proposition 7.16.

50

7. THE LEVI-CIVITA CONNECTION

CHAPTER 8

Geodesics
all smooth vector elds along . It is easily seen that the operations and + given by i. ( X )(t) = X (t), ii. (X + Y )(t) = X (t) + Y (t), make (C 1(TM ) + ) into a vector space. Proposition 8.2. Let (M g) be a smooth Riemannian manifold and : I ! M be a smooth curve. Then there exists a unique operator D 1 1 2 R and f 2 C 1(I R ), dt : C (TM ) ! C (TM ) such that for all D X ) + ( D Y ), D( X + Y ) = ( dt i. dt dt D (f Y ) = df Y + f ( D Y ), and ii. dt dt dt iii. if J0 is an open subset of I such that t0 2 J0 and X 2 C 1(TM ) is a vector eld with X (t) = Y (t) for all t 2 J0 then ( D Y )(t0 ) = (r_X ) (t0 ) : dt Proof. Let us rst prove the uniqueness, so for the moment we D exists. For a point t 2 I choose a assume that such an operator dt 0 chart (U x) on M such that (t0) 2 U . Then a vector eld Y along can be written in the form m X @ j Y (t) = k (t) @xk (t) k=1 for some functions k 2 C 1(I R). The second condition implies that m m X X D D @ @ j : (1) Y ( t ) = )( t ) + _ k (t)( k (t) dt dt @xk @xk (t) k=1 k=1 Let x (t) = (x1 (t) : : : xm (t)) then m X _ (t) = (x _ k )(t) @ j (t) @xk k=1
51

De nition 8.1. Let (TM M ) be the tangent bundle of the smooth manifold M . A vector eld X along a curve : I ! M is a curve X : I ! TM such that X = . We denote by C 1(TM ) the set of

52

8. GEODESICS

D imply that and the third condition for dt m X @ ) = @ ) : (2) ( D @ ) (t) = (r_ @x (x _ k )(t)(r @ @x (t) j (t) dt @xj @xk j

Together equations (1) and (2) give m m X X D (3) Y (t) = f _ k (t) + ;k (t)(x _ i )(t) j (t)g @ j (t) : ij dt @xk i j =1 k=1 D is uniquely determined. This shows that the operator dt It is easily seen that if we use equation (3) for de ning an operator D then it satis es the necessary conditions of Proposition 8.2. This dt D. proves the existence of the operator dt Remark 8.3. It follows from the fact that the Levi-Civita connection is tensorial in its rst argument i.e. rf XY = f rXY and Proposition 8.2 that the value (r XY )p of r XY at p only depends on the value of Xp of X at p and the values of Y along some curve satisfying (0) = p and _ (0) = Xp. The equality D Y )(t ) = (r X ) ( dt 0 _ (t0 ) DY . in Proposition 8.2 allows us to use the notation r _ Y for dt De nition 8.4. Let (M g) be a Riemannian manifold and : I ! M be a C 2-curve. A vector eld X along is said to be parallel along if r_ X = 0: The curve : I ! M is called a geodesic if the vector eld _ is parallel along i.e. r_ _ = 0: Proposition 8.5. Let (M g) be a Riemannian manifold, : I ! M be a smooth curve and X Y be parallel vector elds along . Then the function g(X Y ) : I ! R given by t 7! g (t) (X (t) Y (t) ) is constant. In particular if is a geodesic then g( _ _ ) is constant along . Proof. Using the fact that the Levi-Civita connection is metric we obtain d (g(X Y )) = g(r X Y ) + g(X r Y ) = 0: _ _ dt This proves that the function g(X Y ) is constant along .

k=1

8. GEODESICS

53

Corollary 8.6. Let (M g) be a Riemannian manifold and : I ! M be a smooth curve. If X1 : : : Xm are parallel vector elds along such that for some p 2 (I ) the set f(X1 )p : : : (Xm )pg is a (orthonormal) basis for the tangent space Tp M , then the set f(X1 )q : : : (Xm )q g is a (orthonormal) basis for Tq M for all q 2 (I ). Theorem 8.7. Let (M g) be a Riemannian manifold and I = a b] be an interval on the real line R . Further let : I ! M be a smooth curve, t0 2 I and X0 2 T (t0 ) M . Then there exists a unique parallel vector eld Y along such that X0 = Y (t0 ) . Proof. Without loss of generality we may assume that the image of lies in a chart (U x). On the interval I the tangent eld _ is represented in our local coordinate by
_ (t) =
m X i=1 i (t)

@ j @xi (t)

with some functions i 2 C 1(I R). Similarily let Y be a vector eld along represented by

Y (t) =
Then
m X

m X k=1

k (t)

@ j : @xk (t)

r_ Y (t) =
=

X
k=1

j =1 m

@ j + (t)(r @ ) g f _ j (t) @x (t) j _ @xj (t)


j m X i j =1 k j (t) i (t);ij

f _ k (t ) +
m X i j =1

(t)g @ j (t) : @xk

This implies that r_ Y _ k (t) +

0 if and only if
k j (t) i (t);ij

(t) = 0

for all k = 1 : : : m. Since I is compact if follows from classical results on ODEs that to each initial value (t0) = (v1 : : : vm) 2 R m with
m X @ j X0 = vi @x (t ) k k=1
0

54

8. GEODESICS

there exists a unique solution = ( 1 : : : m ) to the above system. This gives us the unique parallel vector eld Y m X @ j Y (t) = k (t) @xk (t) k=1 along I . Theorem 8.8. Let (M g) be a Riemannian manifold. If p 2 M and v 2 Tp M then there exists an open interval I = (; ) and a unique geodesic : I ! M such that (0) = p and _ (0) = v . Proof. Let (U x) be a local chart on M such that p 2 U . For a C 2 -curve : J ! U we put i = xi : J ! R . The curve x : J ! R m is C 2 so we have (dx) (t) ( _ (t)) = This implies that _ (t) = By di erentation we obtain
m X i=1 m X i=1

_ i(t)ei :

@ j : _ i(t) @x (t)
i

m X @ @ ) g = f j (t) @x j (t) + _ j (t) _ i(t)(r@x @ @x (t) j i j i=1 j =1 m m X X @ j : = f k (t) + _ j (t) _ i(t);k (t)g @x (t) ij k i j =1 k=1 Hence the curve is a geodesic if and only if

r_ _ =

m X j =1 m

@ j r_ _ j (t) @x j (t)

k (t) +

m X

for all k = 1 : : : m. It follows from classical results on ODEs that for initial values q0 = x(p) and w0 = (dx)p(v) there exists an open interval (; ) and an unique solution ( 1 : : : m) satisfying the initial conditions ( 1(0) : : : m(0)) = q0 and ( _ 1(0) : : : _ m(0)) = w0:

i j =1

_ j (t) _ i(t);k ij

(t) = 0

8. GEODESICS

55

De nition 8.9. A Riemannian manifold (M g) is called complete if for each point (p v) 2 TM there exists a geodesic : R ! M de ned on the whole of R such that (0) = p and _ (0) = v. Example 8.10. Let (M g) = E m be the Euclidean space. For the trivial chart idRm : R m ! R m the metric is given by gij = ij , so ;k ij = 0 for all i j k = 1 : : : m. This means that : I ! R m is a geodesic if and only if (t) = 0 or equivalently (t) = t a + b for some a b 2 R m . This proves that the geodesics are the straight lines. ~ Proposition 8.11. Let (M g) be a Riemannian manifold and M be a submanifold equipped with the induced metric g ~. A curve : I ! ~ is a geodesic in M ~ if and only if (r _ )T = 0. M _ ~_ _ = Proof. The statement follows directly from the fact that r (r_ _ )T . Example 8.12. Let E m+1 be the (m + 1)-dimensional Euclidean space and S m E m+1 be the unit sphere with the induced metric. At a point p 2 S m the normal space NpS m is simply the line spanned by p. If : I ! S m is a curve on the sphere, then r_ _ = T = ; N = ; h i : This shows that is a geodesic if and only if (4) =h i : For a point (p v) 2 TS m de ne the curve = (p v) : R ! S m by if v = 0 : t 7! cos(jvjt) p + p sin(jvjt) v=jvj if v 6= 0. Then one easily checks that (0) = p, _ (0) = v and that satis es the geodesic equation (4). This implies that i. every geodesic on S m is a great circle, ii. the standard sphere is complete. Example 8.13. Let Sym(R m+1 ) be equipped with the metric t hA B i = 1 8 trace(A B ): Then we know that the map : S m ! Sym(R m+1 ) with : p 7! (2ppt ; I ) is an isometric immersion such that (S m) = R P m i.e. the m-dimensional real projective space. The map is locally an isometry, so the geodesics on R P m are exactly the images of geodesics on S m. This shows that the real projective space is complete.

56

8. GEODESICS

De nition 8.14. Let (M g) be a Riemannian manifold and : I ! M be a C r -curve on M . A variation of is a C r -map : (; ) I ! M such that for all s 2 I , 0 (s) = (0 s) = (s). If the interval is compact i.e. of the form I = a b], then the variation is called proper if for all t 2 (; ), t (a) = (a) and t (b) = (b). De nition 8.15. Let (M g) be a Riemannian manifold and : I ! M be a C 2-curve on M . For every compact interval a b] I we de ne the energy functional E a b] by Zb 1 E a b]( ) = 2 g( _ (t) _ (t))dt: a A C 2-curve : I ! M is called a critical point for the energy functional if every proper variation of j a b] satis es d (E ( ))j = 0: dt a b] t t=0 Theorem 8.16. A C 2-curve is a critical point for the energy
functional if and only if it is a geodesic. 2 Proof. For a C -map : (; ) I ! M , : (t s) 7! (t s) we de ne the vector elds X = d (@=@s) and Y = d (@=@t) along . The following shows that the vector elds X and Y commute: rXY ;rY X = X Y ] = d (@=@s) d (@=@t)] = d ( @=@s @=@t]) = 0 since @=@s @=@t] = 0. We now assume that is a proper variation of j a b]. Then

d (E ( )) = 1 d (Z b g(X X )ds) dt a b] t 2 dt a Zbd 1 = 2 dt (g(X X ))ds a


= = =

Zb Zb
a a

g(r Y X X )ds g(r XY X )ds

= g(Y X )]b a;

Zb d ( (g(Y X )) ; g(Y r XX ))ds a ds Zb


a

g (Y r XX )ds:

8. GEODESICS

57

The variation is proper, so Y (a) = Y (b) = 0. Furthermore X (0 s) = @ =@s(0 s) = _ (s), so d (E ( ))j = ; Z b g(Y (0 s) (r _ )(s))ds: _ dt a b] t t=0 a The last integral vanishes for every proper variation of if and only if r_ _ = 0. Remark 8.17. A geodesic : I ! (M g) is a special case of what is called a harmonic map : (M g) ! (N h) between Riemannian manifolds. Other examples are conformal immersions : (M 2 g) ! (N h) which parametrize the so called minimal surfaces in (N h). For references on harmonic maps see i. J. Eells, L. Lemaire, A report on harmonic maps, Bull. London Math. Soc. 10, (1978), 1-68. ii. J. Eells, L. Lemaire, Selected topics in harmonic maps, CBMS Regional Conf. Ser. in Math. 50, AMS (1983). iii. J. Eells, L. Lemaire, Another report on harmonic maps, Bull. London Math. Soc. 20, (1988), 385-524. iv. J. Jost, Harmonic maps - analytic theory and geometric signi cance, in Lecture Notes in Math. 1357, Springer (1988) iv. F.E. Burstall, L. Lemaire, J. Rawnsley, The Harmonic Maps Bibliography, v. S. Gudmundsson, The Harmonic Morphisms Bibliography,
http://www.maths.lth.se/matematiklu/personal/sigma/ harmonic/bibliography.html http://www.bath.ac.uk/Departments/Maths/home.html

Let (M m g) be an m-dimensional Riemannian manifold, p 2 M and m;1 = fv 2 T M j g (v v ) = 1g Sp p p be the unit sphere in the tangent space TpM at p. Then every point w 2 TpM ; f0g can be written as w = rw vw , where rw = jwj and m;1 . For v 2 S m;1 let v : (; v v ) ! M be vw = w=jwj 2 Sp p the maximal geodesic such that v v 2 R + f1g, v (0) = p and _ v (0) = v. De ne m;1 p = inf f v v j v 2 Sp g: m;1 is compact, so > 0. Put The unit sphere Sp p 2 Bm p (0) = fv 2 Tp M j gp (v v ) < p g:

58

8. GEODESICS

p if w = 0 ( r vw w ) if w 6= 0. m;1 the line segment v : (; p p) ! Tp M Note that for v 2 Sp with v : t 7! t v is mapped onto the geodesic v i.e. locally we have v = expp v . One can prove that the map expp is smooth and it follows from its de nition that the di erential d(expp)p : TpM ! TpM is the identity map for the tangent space TpM . It then follows from the inverse mapping theorem that there exists an rp 2 R + such that if Up = Brm p (0) and Vp = expp (Up ) then expp jUp : Up ! Vp is a di eomorphism parametrizing the open subset Vp of M . Theorem 8.19. Let (M g) be a Riemannian manifold. Then the geodesics are locally the shortest paths between their endpoints, or more precisly: If p 2 M and : 0 ] ! M is a geodesic with (0) = p, then there exists an with 0 < such that for each q 2 ( 0 ]), is the shortest path from p to q. m Proof. Let p 2 M , U = Br (0) TpM and V = expp(U ) be such that the restriction = expp jU : U ! V of the exponential map at p is a di eomorphism. On V we have the metric g which we pull back via to obtain g ~ = g on U . This makes : (U g ~) ! (V g) into an isometry. It then follows from the construction of the exponential map, that the geodesics in (U g ~) through the point 0 = ;1(p) are exactly the lines v : t ! t v where v 2 TpM . Now let q 2 Brm (0) ; f0g and q : 0 1] ! Brm (0) be the curve q : t 7! t q. Further let : 0 1] ! Brm (0) be any other curve such that (0) = 0 and (1) = q. Along we de ne two vector elds ^ and _ rad by ^ : t 7! ( (t) (t)) and g ~ ( _ (t) ^ (t)) _ rad : t 7! ( (t) (t) ^ (t)): g ~ (t) (^(t) ^ (t)) Then it is easily checked that ~ (t) ( _ (t) ^ (t))j j _ rad (t)j = jg j^j and ~ ( _ ^) : d j ^ (t)j = d qg ~ (t) (^ (t) ^ (t)) = g dt dt j^j Combining these two equations we obtain d j ^ (t)j: j _ rad (t)j dt
expp : w 7!

De nition 8.18. For the above situation we de ne the exponential map expp : B m p (0) ! M at p by

8. GEODESICS

59

This implies that

L( ) =

Z1 Z0 Z0
1 1

j _ jdt j _ radjdt

q.

d j ^ (t)jdt 0 dt = j ^ (1)j ; j ^ (0)j = jq j = L( q ): This proves that in fact q is the shortest path connecting p and
~ be De nition 8.20. Let (M g) be a Riemannian manifold and M

a submanifold with the induced metric g ~. Then 1 trace(B ) 2 C 1(N M ~) H= m ~ ~ in M . The submanis called the mean curvature vector eld of M ~ is said to be ifold M i. minimal in M if H 0, and ii. totally geodesic if B 0. ~ Proposition 8.21. Let (M g) be a Riemannian manifold and M be a submanifold equipped with the induced metric g ~. Then the following conditions are equivalent: ~ is totally geodesic in M i. M ~ is a curve, then the following conditions are equivii. if : I ! M alent ~ is a geodesic in M ~, a. : I ! M b. : I ! M is a geodesic in M . Proof. The result immediately follows from the following decomposition formula ~_ _ + B ( _ _ ): r_ _ = (r_ _ )T + (r_ _ )N = r ~ Proposition 8.22. Let (M g) be a Riemannian manifold and M ~ let (p v) : I ! M be a submanifold. For an arbitrary (p v) 2 T M

~ is totally be the geodesic in M with (0) = p and _ (0) = v . Then M ~ for all (p v) 2 T M ~. geodesic in (M g ) if and only if (p v) (I ) M

60
Proof.

8. GEODESICS

See Exercise 8.3. ~ Proposition 8.23. Let (M g) be a Riemannian manifold and M be a submanifold which is the xpoint set of an isometry : M ! M . ~ is totally geodesic in M . Then M ~ , v 2 TpM ~ and : I ! M be the geodesic Proof. Let p 2 M with (0) = p and _ (0) = v. The map : M ! M is an isometry so : I ! M is a geodesic. The uniqueness result of Theorem 8.8, ( (0)) = (0) and d ( _ (0)) = _ (0) then imply that ( ) = . Hence ~ , so following the image of the geodesic : I ! M is contained in M ~ Proposition 8.22 M is totally geodesic in M . Corollary 8.24. If m ~ < m then the m ~ -dimensional sphere ~ = f(x 0) 2 R m ~ +1 R m;m ~ j jxj2 = 1g Sm is totally geodesic in ~ +1 R m;m ~ j jxj2 + jy j2 = 1g: S m = f(x y) 2 R m m ~ of S m is the xpoint set of the isomProof. The submanifold S m m etry : S ! S with (x y) 7! (x ;y).

Exercises
Find all geodesics in H 2.

8. GEODESICS

61

Exercise 8.1. Let H 2 = (R

R + y12

h iR ) be the hyperbolic plane.


2

Exercise 8.2. Let the orthogonal group O(n) be equipped with the left-invariant metric g(A B ) = trace(At B ). Prove that a C 2-curve : (; ) ! O(n) is a geodesic if and only if t = t . Exercise 8.3. Find a proof for Proposition 8.22. Exercise 8.4. Determine for which 2 (0 =2) the topological 2-torus T 2 = f(cos ei sin ei ) 2 S 3j 2 Rg is a minimal submanifold of the 3-dimensional sphere S 3 = f(z1 z2) 2 C 2 j jz1 j2 + jz2j2 = 1g: ~ = f(x 0) 2 R m ~ R m;m ~ j jxj < 1g is Exercise 8.5. Show that H m m a totally geodesic submanifold of H . Exercise 8.6. Determine the totally geodesic submanifolds of the m-dimensional real projective space R P m . Exercise 8.7. Let the orthogonal group O(n) be equipped with the left-invariant metric g(A B ) = trace(At B ) and let K O(n) be a Lie subgroup. Prove that K is totally geodesic in O(n).

62

8. GEODESICS

CHAPTER 9

The Curvature Tensor


De nition 9.1. Let (M g) be a Riemannian manifold with Levi1(TM ) and Civita connection r. For tensor elds A : Cr1(TM ) ! C0 1 1 B : Cr (TM ) ! C1 (TM ) we de ne their covariant derivatives 1 1 1 rA : Cr1 +1 (TM ) ! C0 (TM ) and rB : Cr+1 (TM ) ! C1 (TM ) by rA : (X X1 : : : Xr ) 7! (rXA)(X1 : : : Xr ) =
X (A(X1 : : : Xr )) ;
r X i=1

A(X1 : : : Xi;1 r XXi Xi+1 : : : Xr )

rB : (X X1 : : : Xr ) 7! (rXB )(X1 : : : Xr ) = rX(B (X1 : : : Xr )) ;


r X i=1

B (X1 : : : Xi;1 r XXi Xi+1 : : : Xr ):

A tensor eld E of type (r 0) or (r 1) is said to be parallel if rE 0. An example of a parallel tensor eld of type (2 0) is the Riemannian metric g of (M g). For this see Exercise 9.1. Let (M g) be a Riemannian manifold. A vector eld Z 2 C 1(TM ) 1(TM ) ! C11(TM ) given by ^ : C1 de nes a smooth tensor eld Z ^ : X 7! r Z XZ: For two vector elds X Y 2 C 1(TM ) we de ne the second covariant derivative r2X Y : C11(TM ) ! C11(TM ) by ^)(Y ): r2X Y : Z 7! (rXZ It then follows from the de nition above that ^(Y )) ; Z ^ (r r2X Y Z = rX(Z XY ) = r Xr YZ ; r rXY Z: De nition 9.2. Let (M g) be a Riemannian manifold with Levi1(TM ) ! C11(TM ) be twice the Civita connection r. Let R : C3 skew-symmetric part of the second covariant derivative r2 i.e. R(X Y )Z = r2X Y Z ; r2Y XZ = r Xr YZ ; r Yr XZ ; rX Y ]Z:
63

64

9. THE CURVATURE TENSOR

Then R is a smooth tensor eld of type (3 1) which we call the curvature tensor of the Riemannian manifold (M g). Proof. See Exercise 9.2. Note that the curvature tensor R only depends on the intrinsic object r and hence it is intrinsic itself. Proposition 9.3. Let (M g) be a smooth Riemannian manifold. For vector elds X Y Z W on M we then have i. R(X Y )Z = ;R(Y X )Z , ii. g(R(X Y )Z W ) = ;g(R(X Y )W Z ), iii. g(R(X Y )Z W ) + g(R(Z X )Y W ) + g(R(Y Z )X W ) = 0, iv. g(R(X Y )Z W ) = g(R(Z W )X Y ), v. 6 R(X Y )Z = R(X Y + Z )(Y + Z ) ; R(X Y ; Z )(Y ; Z ) + R(X + Z Y )(X + Z ) ; R(X ; Z Y )(X ; Z ). Proof. See Exercise 9.3. For a point p 2 M let G2 (TpM ) denote the set of all 2-dimensional subspaces of TpM i.e. G2(TpM ) = fV TpM j V is a 2-dimensional subspace of TpM g: Lemma 9.4. Let X Y Z W 2 TpM such that the two 2-dimensional subspaces spanRfX Y g spanR fZ W g 2 G2 (TpM ) are equal. Then g(R(X Y )Y X ) = g(R(Z W )W Z ) : jX j2jY j2 ; g(X Y )2 jZ j2jW j2 ; g(Z W )2 Proof. See Exercise 9.4. De nition 9.5. For a point p 2 M the function Kp : G2 (TpM ) ! R with R(X Y )Y X ) Kp : spanRfX Y g 7! jXgj( 2jY j2 ; g (X Y )2 is called the sectional curvature at p. Furthermore we de ne the functions : M ! R by : p 7! V 2G min K (V ) and : p 7! V 2G max K (V ): (T M ) p (T M ) p
2

The Riemannian manifold (M g) is said to be i. of (strictly) positive curvature if inf p2M (p) 0 (> 0), ii. of (strictly) negative curvature if supp2M (p) 0 (< 0), iii. of constant curvature if = is constant, iv. at if 0.

9. THE CURVATURE TENSOR

65

Proposition 9.6. Let (M g) be a Riemannian manifold and let (U x) be a local coordinate on M . For i j k l = 1 : : : m put Xi = @=@xi and Rijkl = g(R(Xi Xj )Xk Xl ). Then m m s X @ ;s jk @ ;ik X r sl r s Rijkl = g @x ; @x + f;jk ;s ir ; ;ik ;jr g i j s=1 r=1 Proof. Using the fact that Xi Xj ] = 0 we obtain R(Xi Xj )Xk = r rXjrXiXk Xir XjXk ; P P m s m s =r Xi( s=1 ;jk Xs) ; r Xj( s=1 ;ik Xs) m @ ;s m m s X X X jk s ;r X ; @ ;ik X ; r = X + ; ;s s s jk is r ik ;jsXr @x @x i j s=1 r=1 r=1 m m X @;sjk @;sik X r s r s ; @x + f;jk;ir ; ;ik ;jr g Xs: = @x i j r=1 s=1 Example 9.7. Let (M g) be the Euclidean space. Then the set f@=@x1 : : : @=@xm g is a global frame for T R m . We have gij = ij , so
;k ij 0. This implies that R 0 so E m is at. Example 9.8. The standard sphere S m has constant sectional curvature +1 (see Exercises 9.7 and 9.8) and the hyperbolic space H m has constant sectional curvature ;1 (see Exercise 9.9). Our next goal is Corollary 9.12 where we obtain a formula for the curvature tensor of the manifolds of constant sectional curvature . This turns out to be very useful in the study of Jacobi elds later on. Lemma 9.9. Let (M g) be a Riemannian manifold and (p Y ) 2 ~ : TpM ! TpM with Y ~ : X 7! R(X Y )Y is a TM . Then the map Y symmetric endomorphism of the tangent space Tp M . Proof. For Z 2 Tp M we have ~ (X ) Z ) = g(R(X Y )Y Z ) = g(R(Y Z )X Y ) g(Y ~ (Z )): = g(R(Z Y )Y X ) = g(X Y

Lemma 9.9 implies the existence of eigenvectors X1 : : : Xm for the ~ which form an orthonormal basis for the symmetric endomorphism Y tangent space TpM such that the corresponding eigenvalues satisfy 1 (p) m (p):

66

9. THE CURVATURE TENSOR

De nition 9.10. Let (M g) be a Riemannian manifold. Then de1(TM ) ! C11(TM ) of type (3 1) ne the smooth tensor eld R1 : C3 by R1(X Y )Z = g(Y Z )X ; g(X Z )Y: Proposition 9.11. Let (M g) be a smooth Riemannian manifold and X Y Z be vector elds on M . Then 1 ( ; )jX jjY j2 i. jR(X Y )Y ; + R1 (X Y )Y j 2 2 2 ii. jR(X Y )Z ; + 2 R1 (X Y )Z j 3 ( ; )jX jjY jjZ j Proof. Without loss of generality we can assume that jX j = jY j = jZ j = 1. If X = X ? + X T with X ? ? Y and X T is a multiple of Y then R(X Y )Z = R(X ? Y )Z and jX ?j jX j so we can also assume that X ? Y . Then R1 (X Y )Y = hY Y iX ; hX Y iY = X . The rst statement follows from the fact that the symmetric endomorphism of TpM with + Xg X 7! fR(X Y )Y ; 2 ; has eigenvalues in the interval ; 2 2 ]. It is easily checked that the operator R1 satis es the conditions of + R1 as well. This implies that Proposition 9.3 and hence D = R ; 2 6 D(X Y )Z = D(X Y + Z )(Y + Z ) ; D(X Y ; Z )(Y ; Z ) + D(X + Z Y )(X + Z ) ; D(X ; Z Y )(X ; Z ): The second statement then follows from 1 ( ; )fjX j(jY + Z j2 + jY ; Z j2) 6jD(X Y )Z j 2 +jY j(jX + Z j2 + jX ; Z j2)g 2 + jZ j2 ) + 2jY j(jX j2 + jZ j2)g = 1 ( ; ) f 2 j X j ( j Y j 2 = 4( ; ):
curvature . Then the curvature tensor R is given by R(X Y )Z = (hY Z iX ; hX Z iY ): Proof. This follows directly from Proposition 9.11 by using = = . Proposition 9.13. Let (G h i) be a Lie group equipped with a leftinvariant metric such that for all X 2 g the endomorphism ad(X ) :

Corollary 9.12. Let (M g) be a Riemannian manifold of constant

9. THE CURVATURE TENSOR

67

g ! g is skew-symmetric with respect to h i. Then for any leftinvariant vector elds X Y Z 2 g the curvature tensor R is given by R(X Y )Z = ; 1 4 X Y ] Z ]: Proof. See Exercise 9.6. Theorem 9.14 (The Gauss Equation). Let (M g) be a Riemann~ be a submanifold equipped with the induced metric ian manifold and M ~ Y ~ Z ~W ~ 2 g ~. Let X Y Z W 2 C 1(TM ) be vector elds extending X ~ ). Then C 1(T M ~ (X ~ Y ~ )Z ~W ~ i = hR(X Y )Z W i + hB (Y ~ Z ~) B (X ~ W ~ )i hR ~ Z ~ ) B (Y ~ W ~ )i: ;hB (X ~ , the Proof. Using the de nitions of the curvature tensors R, R ~ and the second fundamental form of M ~ in Levi-Civita connection r M we obtain ~ (X ~ Y ~ )Z ~W ~i hR ~X ~Y ~;r ~Y ~X ~;r ~X ~W ~i = hr ~r ~Z ~r ~Z ~ Y ~ ]Z T T = h(r X(r Y Z ; B (Y Z ))) ; (r Y (r XZ ; B (X Z ))) W i ;h(rX Y ]Z ; B ( X Y ] Z ))T W i = hr Xr YZ ; r Yr XZ ; rX Y ]Z W i ;hrX(B (Y Z )) ; rY (B (X Z )) W i = hR(X Y )Z W i + hB (Y Z ) B (X W )i ; hB (X Z ) B (Y W )i:

As a direct consequence we get the following. ~ be Corollary 9.15. Let (M g) be a Riemannian manifold and M 1 a totally geodesic submanifold. Let X Y Z W 2 C (TM ) be vector ~ Y ~ Z ~W ~ 2 C 1 (T M ~ ). Then elds extending X ~ (X ~ Y ~ )Z ~W ~ i = hR(X Y )Z W i: hR

68

9. THE CURVATURE TENSOR

Exercises

the tensor eld g of type (2 0) is parallel with respect to the Levi-Civita connection. Exercise 9.2. Let (M g) be a Riemannian manifold. Prove that R is a smooth tensor eld of type (3 1). Exercise 9.3. Find a proof for Proposition 9.3. Exercise 9.4. Find a proof for Lemma 9.4. Exercise 9.5. Let g be the Euclidean metric on C m given by

Exercise 9.1. Let (M g) be a Riemannian manifold. Prove that

g (z w ) =

m X k=1

Re(zk wk ):

Let T m be the m-dimensional torus fz 2 C m j jz1 j = ::: = jzmj = 1g with the induced metric g ~. i. Find an isometric immersion : R m ! T m. ii. Determine all geodesics on (T m g). iii. Prove that (T m g) is at. Exercise 9.6. Find a proof for Proposition 9.13. Exercise 9.7. Let the Lie group S 3 = SU(2) be equipped with 1 the metric hX Y i = 2 Reftrace(X t Y )g. i. Find an orthonormal basis for TeSU(2). ii. Show that (SU(2) g) has constant sectional curvature +1. Exercise 9.8. Let S m be the unit sphere in R m+1 equipped with the standard Euclidean metric h iRm+1 . Use the results of Corollaries 8.24, 9.15 and Exercise 9.7 to prove that (S m h iRm+1 ) has constant sectional curvature +1 Exercise 9.9. Let H m = (R + R m;1 x121 h iRm ) be the m-dimensional hyperbolic space. On H m we de ne the operation by ( x) ( y) = ( y + x). For k = 1 : : : m de ne the vector eld @ . Prove that, Xk 2 C 1(TH m) by (Xk )x = x1 @x k i. (H m ) is a Lie group, ii. the vector elds X1 : : : Xm are left-invariant, iii. the metric g is left-invariant, iv. (H m g) has constant curvature ;1.

CHAPTER 10

Curvature and Local Geometry


This chapter is devoted to the study of the local geometry of a Riemannian manifold and how that is controlled by its curvature tensor. We are interested in understanding the spreading of geodesics that all go through the same given point. Using Jacobi elds we obtain a fundamental comparison result describing the curvature dependence of local distances. De nition 10.1. Let (M g) be a smooth Riemannian manifold. By a smooth 1-parameter family of geodesics we mean a C 1-map : (; ) I ! M such that the curve t0 : I ! M given by t0 : s 7! (t0 s) is a geodesic for all t0 2 (; ). We call t 2 (; ) the family parameter of . Proposition 10.2. Let (M g) be a Riemannian manifold and : (; ) I ! M be a 1-parameter family of geodesics. Then the vector eld Jt0 : I ! C 1 (TM ) along t0 given by
t0

Along we put X (t s) = @ =@s and J (t s) = @ =@t. Then @=@t @=@s] = 0 leads to J X ] = d (@=@t) d (@=@s)] = d ( @=@t @=@s]) = 0: The de nition of the curvature tensor now implies that R(J X )X = r Jr XX ; r Xr JX ; rJ X ]X = ;r Xr JX = ;r Xr XJ: Hence for each t0 2 (; ) we have r_ t0r_ t0Jt0 + R(Jt0 _ t0 ) _ t0 = 0:
Proof.

satis es the second order ordinary di erential equation r_ t r_ t Jt0 + R(Jt0 _ t0 ) _ t0 = 0:


0 0

Jt0 (s) = @ @t (t0 s)

69

70

10. CURVATURE AND LOCAL GEOMETRY

if

De nition 10.3. Let (M g) be a Riemannian manifold and : I ! M be a geodesic. A vector eld J along is called a Jacobi eld

along where X = _ . We denote the space of all Jacobi elds along by J (TM ). Lemma 10.4. Let (M g) be a Riemannian manifold and : I ! M be a geodesic. Then the space of Jacobi elds J (TM ) is a vector space. Proof. This follows directly from the fact that the Jacobi eld equation is linear in J . Proposition 10.5. Let : I ! M be a geodesic, t0 2 I , p = (t0) and X = _ along . If v w 2 Tp M are two tangent vectors at p then there exists a unique Jacobi eld J along , such that Jp = v and (r XJ )p = w. Proof. Let fX1 : : : Xm g be an orthonormal frame of parallel vecP aX tor elds along . If J is a vector eld along , then J = m i=1 i i where ai = hJ Xii are smooth functions P onmI . The vector elds X : Xm are parallel so we have r _ iXi andP rXrXJ = XJ = i=1 a P1m: :a m k i=1 i Xi : For the curvature tensor we have R(Xi X )X = k=1 bi Xk , where bk i = hR(Xi X )X Xk i are smooth functions on I depending on the geometry of (M g). This means that R(J X )X is given by P m R(J X )X = i k=1 aibk i Xk . We now see that J is a Jacobi eld if and only if m m X X (ai + ak bik )Xi = 0: This is equivalent to the second order system
i=1 k=1

rXrXJ + R(J X )X = 0

ai +

m X k=1

ak bik = 0 for all i = 1 2 : : : m,

of linear ODEs in a = (a1 : : : am). A global solution will always exist and is uniquely determined by a(t0 ) and a _ (t0 ). This implies that J exists globally and is uniquely determined by the initial conditions J (t0 ) and (r XJ ) (t0 ). Corollary 10.6. Let : I ! M m be a geodesic. Then the vector space J (TM ) of all Jacobi elds along is 2m-dimensional.

10. CURVATURE AND LOCAL GEOMETRY

71

Lemma 10.7. Let (M g) be a Riemannian manifold, : I ! M be a geodesic and J be a Jacobi eld along . Let 2 R and : I ! I be given by : t 7! t= , then : I ! M is a geodesic and J is a Jacobi eld along . Proof. See Exercise 10.1. This means that when proving results about Jacobi elds along a geodesic we can always without loss of generality assume that j _ j = 1. Proposition 10.8. Let (M g) be a Riemannian manifold, : I ! M be a geodesic with j _ j = 1 and J be a Jacobi eld along . Let J T be the tangential part of J given by J T = hJ _ i _ and J N = J ; J T be the normal part. Then J T and J N are Jacobi elds along and there exist a b 2 R such that J T (s) = (as + b) _ (s) for all s 2 I . Proof. We now have r_ r_ J T + R(J T _ ) _ = r_ r_ (hJ _ i _ ) + R(hJ _ i _ _ ) _ = hr_r_ J _ i _ = ;hR(J _ ) _ _ i _ = 0: This shows that the tangential part J T of J is a Jacobi eld. The fact that J (TM ) is a vector space implies that the normal part J N = J ; J T of J also is a Jacobi eld. By di erentiating hJ _ i twice along we obtain
d2 hJ _ i = hr r J _ i = ;hR(J _ ) _ _ i = 0 _ _ ds2 so hJ _ i(s) = (as + b) for some a b 2 R . Note that the last statement of Proposition (10.8) implies that we now know all the tangential Jacobi elds along . At this point we remind the reader of the following classical fact: Remark 10.9. If is a real number then the unique solution to the initial value problem f + f = 0 f (0) = a and f_(0) = b is the function f : R ! R , f (s) = ac (s) + bs (s) where c : R ! R and s : R ! R are given by 8 p > <cosh( j js) if < 0 c (s) = >1 if = 0 :cos(p s) if > 0:

72

10. CURVATURE AND LOCAL GEOMETRY

and

8 p p > <sinh( j js)= j j s (s) = >s :sin(p s)=p


2

if < 0 if = 0 if > 0:

rotations about the origin produce a 1-parameter family of geodesics it t : s 7! se . Along the geodesic 0 : s 7! s we get the Jacobi eld J0(s) = @ t =@t(0 s) = is with jJ0(s)j = jsj = js (s)j2. Example 10.11. Let S 2 be the unit sphere in the standard Euclidean 3-space C R with the induced metric of constant sectional curvature = +1. Rotations about the R -axis produce a 1-parameter family of geodesics t : s 7! (sin(s)eit cos(s)). Along the geodesic 0 : s 7! (sin(s) cos(s)) we get the Jacobi eld J0 (s) = @ t =@t(0 s) = (isin(s) 0) with jJ0(s)j2 = sin2 (s) = js (s)j2. Example 10.12. Let B12 (0) be the open unit disk in the complex plane with the hyperbolic metric 4=(1 ;jzj2)2h iR2 of constant sectional curvature = ;1. Rotations about the origin produce a 1-parameter family of geodesics t : s 7! tanh(s)eit . Along the geodesic 0 : s 7! tanh(s) we get the Jacobi eld J0 (s) = i tanh(s) with 2 (s) 2 2 jJ0(s)j2 = 14;tanh tanh2 (s) = sinh (s) = js (s)j : We will now see that when the manifold (M g) has constant sectional curvature we can completely solve the Jacobi eld equation

Example 10.10. Let C be the complex plane with the standard Euclidean metric h iR of constant sectional curvature = 0. The

rXrXJ + R(J X )X = 0 along any given geodesic : I ! M , where X = _ . Let (M g) be a Riemannian manifold of constant sectional curvature and : I ! M be a geodesic with jX j = 1 where X = _ .

Further let P1 P2 : : : Pm;1 be parallel vector elds along such that g(Pi Pj ) = ij and g(Pi X ) = 0. Any vector eld J along may now be written as m ;1 X J (s) = fi(s)Pi(s) + fm (s)X (s):
i=1

We now see that J is a Jacobi eld if and only if

10. CURVATURE AND LOCAL GEOMETRY

73

m;1 i=1

fi(s)Pi(s) + fm (s)X (s) = r Xr XJ = ;R(J X )X = ;R(J N X )X = ; (g(X X )J N ; g(J N X )X ) = ; JN m ;1 X = ; fi (s)Pi(s):


i=1

This implies that J is a Jacobi eld if and only if the following system of ODEs is satis ed: (5) fi (s) + fi(s) = 0 for all i = 1 2 : : : m ; 1 and fm(s) = 0: It is clear thatP for the initial values ;1 vi Pi(s0) + vm X (s0), i. J (s0 ) = m i=1P m;1 ii. (r XJ )(s0) = i=1 wiPi(s0) + wm X (s0) or equivalently fi(s0 ) = vi and f_i (s0) = wi for all i = 1 2 ::: m we can solve the system (5) explicitly on the whole of I . Corollary 10.13. Let (M g) be a Riemannian manifold, : I ! M be a geodesic and J be a Jacobi eld along . If g(J (t0) _ (t0 )) = 0 and g((r_J )(t0 ) _ (t0 )) = 0 for some t0 2 I , then g(J (t) _ (t)) = 0 for all t 2 I . Proof. The conditions vm = wm = 0 imply that fm = 0. Example 10.14. Let S 2 be the unit sphere in the standard Euclidean 3-space C R with the induced metric of constant curvature = +1 and : R ! S 2 be the geodesic given by : s 7! (eis 0). Then _ (s) = (ieis 0) so it follows from Proposition (10.8) that all Jacobi elds tangential to are given by J(Ta b) (s) = (as + b)(ieis 0) for some a b 2 R : The vector eld P : R ! TS 2 given by s 7! (eis 0) (0 1)] satis es hP _ i = 0 and jP j = 1. The sphere S 2 is 2-dimensional and _ is parallel along so P must be parallel. This implies that all the Jacobi elds orthogonal to _ are given by J(N a b) (s) = (0 a cos s + b sin s) for some a b 2 R :

74

10. CURVATURE AND LOCAL GEOMETRY

We will now see how Jacobi elds can be constructed in the general situation when the curvature not necessarily is constant: Let (M g) be a complete Riemannian manifold, p 2 M and v w 2 TpM . Then s 7! s(v + tw) de nes a 1-parameter family of lines in the tangent space TpM which all pass through the origin 0 2 TpM . Remember that the exponential map m ! exp(B m ) (exp)pjB"m : B" "p (0) p (0) p (0) maps lines in TpM through the origin onto geodesics on M . Hence the map t : s 7! (exp)p (s(v + tw)) is a 1-parameter family of geodesics through p 2 M , as long as s(v + tw) m . This means that J : s 7! (@ t =@t)(0 s) is a is an element of B" p (0) Jacobi eld along the geodesic : s 7! 0 (s) with (0) = p and _ (0) = w. It is easily veri ed that J satis es the initial conditions J (0) = 0 and (r XJ )(0) = w. Lemma 10.15. Let (M g) be a Riemannian manifold with sectional curvature K uniformly bounded above by and : 0 ] ! M be a geodesic on M with jX j = 1 and X = _ . Further let J : 0 ] ! TM be a Jacobi eld along such that g(J X ) = 0 and jJ j 6= 0 on (0 ). Then d2 (jJ j) + i. ds jJ j 0, 2 ii. if f : 0 ] ! R is a C 2 -function, such that a. f + f = 0 and f > 0 on (0 ), b. f (0) = jJ (0)j, and c. f_(0) = jr XJ (0)j, then f (s) jJ (s)j on (0 ), iii. if J (0) = 0, then jr XJ (0)j s (s) jJ (s)j for all s 2 (0 ). Proof. i. Using the facts that jX j = 1 and hX J i = 0 we obtain d2 (jJ j) = d2 phJ J i = d ( hr XJ J i ) ds2 ds2 ds jJ j hrXrXJ J i jrXJ j2jJ j2 ; hrXJ J i2 = jJ j + jJ j3 hrXrXJ J i jJ j = ; hR(J X )X J i jJ j ; jJ j:

(s)j ii. De ne the function h : 0 ) ! R by h : s 7! jJ f (s) for s 2 (0 ) (s)j and h(0) = lims!0 jJ f (s) = 1. Then

10. CURVATURE AND LOCAL GEOMETRY

75

_ (s) = 21 ( d (jJ (s)j)f (s) ; jJ (s)jf_(s)) h f (s) ds Z s d2 1 = f 2(s) ( dt2 (jJ (t)j)f (t) ; jJ (t)jf (t))dt Z0 s d2 1 = f 2(s) f (t)( dt2 (jJ (t)j) + jJ (t)j)dt 0 0: _ (s) 0 so f (s) jJ (s)j for all s 2 (0 ). This implies that h iii. The function f (s) = j(r XJ )(0)j s (s) satis es f (s)+ f (s) = 0, f (0) = jJ (0)j = 0 and f_(0) = j(r XJ )(0)j so it follows from (ii) that j(rXJ )(0)j s (s) = f (s) jJ (s)j. Let (M g) be a Riemannian manifold of sectional curvature which is uniformly bounded above, i.e. there exists a 2 R such that Kp(V ) for all V 2 G2(TpM ) and p 2 M . Let (M g ) be another Riemannian manifold which is complete and of constant sectional curvature K . Let p 2 M , p 2 M and identify TpM = R m = Tp M . Let U be an open neighbourhood of R m around 0 such that the exponential maps (exp)p and (exp)p are di eomorphisms from U onto their images (exp)p(U) and (exp)p (U ), respectively. Let (r p q) be a geodesic triangle i.e. a triangle with sides which are shortest paths between their end points. Furthermore let c : a b] ! M be the side connecting r and q and v : a b] ! TpM be the curve de ned by c(t) = (exp)p(v(t)). Put c (t) = (exp)p (v(t)) for t 2 a b] and then it directly follows that c(a) = r and c(b) = q. Finally put r = c (a) and q = c (b).
for the distance function is satis ed: d(q r ) d(q r).

Theorem 10.16. For the above situation the following inequality

Proof. De ne a 1-parameter family s 7! s v (t) of straight lines in TpM through p. Then t : s 7! (exp)p(s v(t)) and t : s 7! (exp)p (s v(t)) are 1-parameter families of geodesics through p 2 M , and p 2 M , respectively. Hence Jt = @ t =@t and Jt = @ t =@t are Jacobi elds satisfying the initial conditions Jt(0) = Jt (0) = 0 and

76

10. CURVATURE AND LOCAL GEOMETRY

(r _ (t). Using Lemma (10.15) we now obtain XJ t )(0) = (r XJt )(0) = v jc_ (t)j = jJt (1)j = j(rXJt )(0)j s (1) = j(r _(t)j XJt )(0)j s (1) jJt (1)j = jc The curve c is the shortest path between r and q so we have d(r q ) L(c ) L(c) = d(r q): We now add the assumption that the sectional curvature of the manifold (M g) is uniformly bounded below i.e. there exists a 2 R such that Kp(V ) for all V 2 G2 (TpM ) and p 2 M . Let (M g ) be a complete Riemannian manifold of constant sectional curvature . Let p 2 M and p 2 M and identify TpM = R m = Tp M . Then a similar construction as above gives two pairs of points q r 2 M and q r 2 M and shows that d(q r) d(q r ): Combining these two results we obtain locally d(q r ) d(q r) d(q r ):

10. CURVATURE AND LOCAL GEOMETRY

Exercises

77

Exercise 10.1. Find a proof for Lemma (10.7). Exercise 10.2. Let (M g) be a Riemannian manifold and : I ! M be a geodesic such that X = _ = 6 0. Further let J be a non-vanishing
Jacobi eld along with g(X J ) = 0. Prove that if g(J J ) is constant along then (M g) does not have strictly negative curvature.

También podría gustarte