Durham University Michaelmas 2014

Pavel Tumarkin

Riemannian Geometry IV, Term 1 (Sections 1–2)

1 Smooth manifolds
“Smooth” means “infinitely differentiable”, C ∞ .

Definition 1.1. Let M be a set. An n-dimensional smooth atlas on M is a collection of triples (Uα , Vα , ϕα ),
where α ∈ I for some indexing set I, s.t.

(a) Uα ⊆ M ; Vα ⊆ Rn is open ∀α ∈ I;
S
(b) α∈I Uα = M ;
(c) Each ϕα : Uα → Vα is a bijection;
(d) For every α, β ∈ I such that Uα ∩ Uβ 6= ∅ the composition ϕβ ◦ ϕ−1

α ϕα (Uα ∩Uβ ) : ϕα (Uα ∩ Uβ ) →

ϕβ (Uα ∩ Uβ ) is a smooth map for all ordered pairs (α, β), where α, β ∈ I.

The number n is called the dimension of M , the maps ϕα are called coordinate charts, the compositions
ϕβ ◦ ϕ−1
α are called transition maps or changes of coordinates.

Example 1.2. Two atlases on a circle S 1 ⊂ R2 .

Definition 1.3. Let M have a smooth atlas. A set A ⊆ M is open if for every α ∈ I the set ϕα (A ∩ Uα )
is open in Rn . If A ⊂ M is open and x ∈ A, A is called an open neighborhood of x.
Definition 1.4. M is called Hausdorff if for each x, y ∈ M , x 6= y, there exist open sets Ax 3 x and
Ay 3 y such that Ax ∩ Ay = ∅.
Example 1.5. An example of a non-Hausdorff space: a line with a double point.
Definition 1.6. M is called a smooth n-dimensional manifold if M has a countable n-dimensional smooth
atlas and M is Hausdorff
Example 1.7. Atlas for a square in R2 .
Example. Examples of smooth manifolds: torus, Klein bottle, 3-torus, real projective space.
Definition 1.8. Let U ⊆ Rn be open, m < n, and let f : U → Rm be a smooth map (i.e., all the partial
∂fi
derivatives are smooth). Let Df (x) = ( ∂x j
) be the matrix of partial derivatives at x ∈ U (differential or
Jacobi matrix). Then
(a) x ∈ Rn is a regular point of f if rk Df (x) = m (i.e., Df (x) has a maximal rank);
(b) y ∈ Rm is a regular value of f if the full preimage f −1 (y) consists of regular points only.
Theorem 1.9 (Corollary of Implicit Function Theorem). Let U ⊂ Rn be open, f : U → Rm smooth,
m < n. If y ∈ f (U ) is a regular value of f then f −1 (y) ⊂ U ⊂ Rn is an (n − m)-dimensional smooth
manifold.

Examples 1.10–1.11. An ellipsoid as a smooth manifold; matrix groups are smooth manifolds.

1
2 Tangent space
Definition 2.1. Let f : M m → N n be a map of smooth manifolds with atlases (Ui , ϕi (Ui ), ϕi )i∈I and
(Wj , ψj (Wj ), ψj )j∈J m n
 . The map f is smooth if it induces smooth maps between open sets in R and R ,
−1 
i.e. if ψj ◦ f ◦ ϕi ϕ (U ∩f −1 (W ∩f (U ))) is smooth for all i ∈ I, j ∈ J.
i i j i
If f is a bijection and both f and f −1 are smooth then f is called a diffeomorphism.

Definition 2.2. A derivation on the set C ∞ (M, p) of all smooth functions on M defined in a neighborhood
of p is a linear map δ : C ∞ (M, p) → R, s.t. for all f, g ∈ C ∞ (M, p) holds δ(f · g) = f (p)δ(g) + δ(f )g(p)
(the Leibniz rule).
The set of all derivations is denoted by D∞ (M, p). This is a real vector space (exercise).

Definition 2.3. The space D∞ (M, p) is called the tangent space of M at p, denoted Tp M . Derivations
are tangent vectors.

Definition 2.4. Let γ : (a, b) → M be a smooth curve in M , t0 ∈ (a, b), γ(t0 ) = p and f ∈ C ∞ (M, p).
Define the directional derivative γ 0 (t0 )(f ) ∈ R of f at p along γ by

0 f (γ(t0 + s)) − f (γ(t0 )) 0 d 
γ (t0 )(f ) = lim = (f ◦ γ) (t0 ) =  (f ◦ γ)
s→0 s dt t=t0

Directional derivatives are derivations (exercise).

Remark. Two curves γ1 and γ2 through p may define the same directional derivative.

Notation. Let M n be a manifold, ϕ : U → V ⊆ Rn a chart at p ∈ U ⊂ M . For i = 1, . . . , n define the

curves γi (t) = ϕ−1 (ϕ(p) + ei t) for small t > 0 (here {ei } is a basis of Rn ).
∂ 
= γi0 (0), i.e.

Definition 2.5. Define ∂x i p


∂  d ∂
(f ) = (f ◦ γi )0 (0) = (f ◦ ϕ−1 )(ϕ(p) + tei )t=0 = (f ◦ ϕ−1 )(ϕ(p)),

∂xi p
 dt ∂xi

∂
where ∂xi on the right is just a classical partial derivative.

By definition, we have
∂ ∂
h ,..., i ⊆ {Directional derivatives} ⊆ D∞ (M, p)
∂x1 ∂xn

Proposition 2.6. h ∂x∂ 1 , . . . , ∂x∂n i = {Directional derivatives} = D∞ (M, p).

Lemma 2.7. Let ϕ : U ⊆ M → Rn be a chart, ϕ(p) = 0. Let γ̃(t) = ( ni=1 ki ei )P t : R → Rn be a line,

P
where {e1 , . . . , en } is a basis, ki ∈ R. Define γ(t) = ϕ ◦ γ̃(t) ∈ M . Then γ (0) = ni=1 ki ∂x
−1 0 ∂
i
.

Example 2.8. For the group SLn (R) = {A ∈ Mn | det A = 1}, the tangent space at I is the set of all
trace-free matrices: TI (SLn (R)) = {X ∈ Mn (R) | tr X = 0}.

Proposition 2.9. (Change of basis for Tp M ). Let M n be a smooth manifold, ϕα : Uα → Vα a

 Pn ∂xβi ∂
chart, (xα1 , . . . , xαn ) the coordinates in Vα . Let p ∈ Uα ∩ Uβ . Then ∂x∂α  = i=1 ∂xα ∂xα , where

j p j i

∂xβ ∂(ϕiβ ◦ϕ−1

α )
i
∂xα = ∂xα (ϕ(p)), ϕiβ = πi ◦ ϕβ .
j j

2
Definition 2.10. Let M, N be smooth manifolds, let f : M → N be a smooth map. Define a linear map
Df (p) : Tp M → Tf (p) N called the differential of f at p by Df (p)γ 0 (0) = (f ◦ γ)0 (0) for a smooth curve
γ ∈ M with γ(0) = p.

Remark. Df (p) is well defined (exercise).


∂  ∂
Lemma 2.11. (a) If ϕ is a chart, then Dϕ(p) : Tp M → Tϕ(p) Rn is the identity map taking ∂xi p to ∂xi

f g
(b) For M → N → L holds D(g ◦ f )(p) = Dg(f (p)) ◦ Df (p).

Example 2.12. Differential of a map from a disc to a sphere.

Tangent bundle and vector fields

Definition 2.13. Let M be a smooth manifold. A disjoint union T M = ∪p∈M Tp M of tangent spaces to
each p ∈ M is called a tangent bundle.
There is a canonical projection Π : T M → M , Π(v) = p for every v ∈ Tp M .

Proposition 2.14. The tangent bundle T M has a structure of 2n-dimensional smooth manifold, s.t.
Π : T M → M is a smooth map.

Definition 2.15. A vector field X on a smooth manifold M is a smooth map X : M → T M such that
∀p ∈ M X(p) ∈ Tp M
The set of all vector fields on M is denoted by X(M ).

Remark 2.16. (a) X(M ) has a structure of a vector space.

(b) Vector fields can be multiplied by smooth functions.

(c) Taking
Pn a coordinate chart (U, ϕ = (x1 , . . . , xn )), any vector field X can be written in U as X(p) =
∂
i=1 fi (p) ∂xi ∈ Tp M , where {fi } are some smooth functions on U .

Examples 2.17–2.18. Vector fields on R2 and 2-sphere.

P ∂
Remark 2.19. Observe that for X = ai (p) ∂x i
∈ X(M ) we have X(p) ∈ Tp M , i.e. X(p) is a directional
Pp ∈ M
derivative at . Thus, we can use the vector field to differentiate a function f ∈ C ∞ (M ) by
∂f 
(Xf )(p) = ai (p) ∂xi p , so that we get another smooth function Xf ∈ C ∞ (M ).

Proposition 2.20. Let X, Y ∈ X(M ). Then there exists a unique vector field Z ∈ X(M ) such that
Z(f ) = X(Y (f )) − Y (X(f )) for all f ∈ C ∞ (M ).
This vector field Z = XY − Y X is denoted by [X, Y ] and called the Lie bracket of X and Y .

Proposition 2.21. Properties of Lie bracket:

(a) [X, Y ] = −[Y, X];

(b) [aX + bY, Z] = a[X, Z] + b[Y, Z] for a, b ∈ R;

(c) [[X, Y ], Z] + [[Y, Z], X] + [[Z, X], Y ] = 0 (Jacobi identity);

(d) [f X, gY ] = f g[X, Y ] + f (Xg)Y − g(Y f )X for f, g ∈ C ∞ (M ).

3
Definition 2.22. A Lie algebra is a vector space g with a binary operation [·, ·] : g × g → g called the
Lie bracket which satisfies first three properties from Proposition 2.21.

Proposition 2.21 implies that X(M ) is a Lie algebra.

Theorem 2.23 (The Hairy Ball Theorem). There is no non-vanishing continuous vector field on an
even-dimensional sphere S 2m .