INTRODUCTION

TO
DIFFERENTIAL GEOMETRY

Exercises with solucions

Course 21/22

Master in Mathematics and Statistics

Faculty of Science - University of Helsinki

Ilkka Holopainen
Juan-Carlos Felipe-Navarro
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 1, Solutions
24.1.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, January 24, 12:00 o’clock.

∗1. Two metrics on a set are called equivalent if they induce the same
metric topology.
Let (X, d) be a metric space. Prove that there exists a metric d¯ on
¯ is bounded.
X, equivalent to d, such that (X, d)
¯
[Hint: Consider the function d : X × X → R,

¯ y) = d(x, y)
d(x, .
1 + d(x, y)

Solution. Before anything let us show that d¯ is a distance. In order to

do it let us define the auxiliary increasing function f (t) = t/(1 + t),
which gives a bijection between [0, +∞) and [0, 1). Then, d¯ = f (d).
From being d a distance and the properties of f , it is trivial that d¯
is positive definite and symmetric. Then, let us conclude that it is
a distance by checking the triangle inequality. That is, by triangle
inequality for d and the monotonicity of f we obtain

¯ y) = f (d(x, y)) ≤ f (d(x, z) + d(z, y)) = d(x, z) + d(z, y)

d(x,
1 + d(x, z) + d(z, y)
d(x, z) d(z, y)
= +
1 + d(x, z) + d(z, y) 1 + d(x, z) + d(z, y)
d(x, z) d(z, y) ¯ z) + d(z,
¯ y)
≤ + = d(x,
1 + d(x, z) 1 + d(z, y)
¯ is a metric space. In order to show
for any x, y, z ∈ X. Thus, (X, d)
that it is bounded it is enough to note that d(x,¯ y) < 1 for each
x, y ∈ X.
¯ induce the same topo-
Finally, let us show that (X, d) and (X, d)
logy. Since balls are a base in metric spaces, it is enough to check
that balls with one metric are also balls in the other one. That is, if
follows from the properties of f that
   
r ρ
Bd (x, r) = Bd¯ x, and Bd¯ (x, ρ) = Bd x,
1+r 1−ρ
for any x ∈ X, r ∈ [0, +∞) and ρ ∈ [0, 1).

∗2. Prove that a topological space X is Hausdorff if and only if, for
every x ∈ X, the intersection of all closed neighborhoods of x is the
2

singleton {x}. [Note: a closed neighborhood of a point is a closed set

containing some open neighborhood of the point.]
Solution. Given x ∈ X let us define the set
\
Nx := {W : W ⊂ X is a closed neighborhood of x}.
Suppose that X is Hausdorff. Let x, y ∈ X be distinct. Then,
there exist U and V disjoint neighborhoods of x and y, respectively.
In particular X \ V is a closed neighborhood of x not containing y.
Hence y ∈ / Nx . Since we can do it for any y 6= x and x ∈ Nx by
definition, we conclude that Nx = {x}.
Let x, y ∈ X be distinct. Since Nx = {x} there exists W , a closed
neighborhood of x not containing y. By definition of closed neigh-
borhood, it means the existence of U ⊂ W , an open neighborhood
of x. Therefore U and X \ W are disjoint neighborhoods of x and y,
respectively, and X is Hausdorff.

3. Let (X1 , T1 ), . . . , (Xk , Tk ) be topological spaces and let X = X1 ×

· · · × Xk be equipped with the product topology. Prove:
(a) If every (Xi , Ti ) is Hausdorff, then also X is Hausdorff.
(b) If every (Xi , Ti ) is N2 , then also X is N2 .
Solution.
(a) Let x = (x1 , . . . , xk ), y = (y1 , . . . , yk ) ∈ X be distinct. Then
xj 6= yj for some index j. Let Uj and Vj be disjoint neighbour-
hoods of xj and yj , respectively. Then the sets
U := X1 × · · · × Uj × · · · × Xk
and
V := X1 × · · · × Vj × · · · × Xk
are disjoint neighbourhoods of x and y, respectively.
(b) Let Uj be a countable base for Xj , j = 1, . . . , k. Define
U = {U1 × · · · × Uk : Uj ∈ Uj , j = 1, . . . , k}.
This set is countable. Moreover, we claim that it is also a base
for the product topology of X. That is, let us take an open set
U ⊂ X. For every x = (x1 , . . . , xk ) ∈ U we may find an open set
x ∈ V1x ×· · ·×Vkx ⊂ U by the definition of the product topology.
For each index j we may also find a base element Ujx ∈ Uj so
that xj ∈ Ujx ⊂ Vjx . Consequently x ∈ U1x × · · · × Ukx ⊂ U and
we may write U as a union of elements in U:
[
U= U1x × · · · × Ukx .
x∈U

4. Let Tcc = {U ⊂ R : R \ U is countable} ∪ {∅}.

(a) Prove that Tcc is a topology in R (co-countable topology).
 3

(b) Prove that (R, Tcc ) is not Hausdorff.

(c) Prove that limits of convergent sequences are unique in (R, Tcc ).
Solution.
(a) Clearly ∅, R ∈ Tcc (sinceS R \ R = ∅). Let {Uj }j∈J ⊂ Tcc . Since
we want to prove that j Uj ∈ Tcc , we may assume without loss
of generality that Uj 6= ∅ for all j ∈ J. Then R \ Uj is countable
for all j ∈ J. Since the intersection of an arbitrary family of
countable sets is also countable, we have that
!
[ \
R\ Uj = R \ Uj
j∈J j∈J
[
is a countable set. Hence Uj ∈ Tcc .
j∈J
Similarly if U, V ∈ Tcc \ {∅}, then R \ (U ∩ V ) = R \ U ∪ R \ V
is countable, implying U ∩ V ∈ Tcc . These arguments show that
Tcc is a topology.
(b) Any two non-empty sets U, V ∈ Tcc have uncountable inter-
section, since R \ (U ∩ V ) is countable. In particular no such
intersection can be empty and hence no pair of distinct points
can have disjoint neighborhoods.
(c) Suppose (xk ) ⊂ R is a sequence converging in Tcc to x and to
y. Suppose x 6= y. Then R \ {x} is a neighbourhood of y and
there exists n0 so that xk ∈ R \ {x} for all k ≥ k0 . In other
words xk 6= x for all k ≥ k0 . Thus the set R \ {xk0 , xk0 +1 , . . .}
is a neighbourhood of x so there should exist k1 so that
xk ∈ R \ {xk0 , xk0 +1 , . . .}
whenever k ≥ k1 . However this is clearly impossible. Thus we
conclude that x must equal y.

5. Let M be the set of all continuous functions f : [0, 1] → R. Define

d(f, g) = sup{|f (x) − g(x)| : 0 ≤ x ≤ 1}
for f, g ∈ M. Prove that:
(a) d is a metric in M .
(b) d(fn , f ) → 0 ⇐⇒ fn → f uniformly.
(c) (M, d) is complete.
Solution.
(a) Clearly d(f, f ) = 0 and conversely, if d(f, g) = 0 then
|f (t) − g(t)| ≤ d(f, g) = 0
for all t ∈ [0, 1] so that f = g.
4

Also
d(f, g) = sup |f (t) − g(t)| = sup |g(t) − f (t)| = d(g, f )
0≤t≤1 0≤t≤1

so that d is symmetric.
Finally, the triangle inequality holds:
d(f, g) = sup |f (t) − g(t)| ≤ sup (|f (t) − h(t)| + |h(t) − g(t)|)
0≤t≤1 0≤t≤1

≤ sup (|f (t) − g(t)| + d(h, g)) = d(f, h) + d(h, g)

0≤t≤1

(b) Suppose fn → f uniformly. Given ε > 0 there is n0 so that

|fn (t) − f (t)| < ε/2 for all t ∈ [0, 1] whenever n ≥ n0 . Thus
d(fn , f ) = sup |fn (t) − f (t)| ≤ ε/2 < ε
t∈[0,1]

whenever n ≥ n0 . Thus d(fn , f ) → 0 as n → ∞.

Next assume that d(fn , f ) → 0 as n → ∞. Let ε > 0 and take
n0 so that d(fn , f ) < ε whenever n ≥ n0 . Then
|fn (t) − f (t)| ≤ d(fn , f ) < ε
for all t ∈ [0, 1] whenever n ≥ n0 . In other words fn → f
uniformly.
(c) Let (fn ) be a Cauchy sequence in M . Then (fn (t)) is a Cauchy
sequence in R having limit f (t) for each t. We must prove that
f is continuous (i.e. belongs to M ) and that d(fn , f ) → 0 as
n → ∞. To accomplish these it suffices to show that fn → f
uniformly.
For this let ε > 0 and take n0 so that d(fn , fm ) < ε whenever
n, m ≥ n0 . Then
|fn (t) − f (t)| = lim |fn (t) − fm (t)| ≤ lim sup d(fm , fn ) ≤ ε
m→∞ m→∞

for all t ∈ [0, 1] whenever n ≥ n0 . Thus fn → f uniformly as

n → ∞ and we are done.

6. Let M be as above. Define a mapping T : M → M by setting (for

all f ∈ M and x ∈ [0, 1])
Z x
T (f )(x) = a + K(x, y)f (y) dy,
0

where a is a constant and K : [0, 1] × [0, 1] → R is continuous. Let

Z x 
k = sup |K(x, y)| dy : 0 ≤ x ≤ 1
0

and suppose that k < 1.

(a) Prove that T is a contraction, i.e. d T (f ), T (g) ≤ Ld(f, g),
where L < 1.
 5

(b) Prove that the integral equation

Z x
f (x) = a + K(x, y)f (y) dy
0
has a unique solution.
Solution. Before anything let us show that T f ∈ M for each f ∈ M
(i.e. T f is continuous whenever f is). Take a sequence xn converging
to x in [0, 1]. Then
lim χ[0,xn ] (y)K(xn , y) = χ[0,x] (y)K(x, y)
n→∞
pointwise except possibly at y = x. By the dominated convergence
theorem
Z 1
lim T f (xn ) = lim χ[0,xn ] (y)K(xn , y)f (y) dy + a
n→∞ n→∞ 0
Z 1
= χ[0,x] (y)K(x, y)f (y) dy + a = T f (x)
0
so that T f is, indeed, continuous.
(a) Let us estimate
Z x
d(T f, T g) = sup K(x, y)(f (y) − g(y)) dy
0≤x≤1 0
Z x
≤ sup |K(x, y)||f (y) − g(y)| dy
0≤x≤1 0
Z x
≤ sup |K(x, y)|d(f, g) dy
0≤x≤1 0
= kd(f, g).
Thus T is a contraction with L = k < 1.
(b) By the previous results T : M → M is a contractive self map-
ping of a complete metric space. The Banach fixed point theo-
rem implies the existence of a unique element f ∈ M so that
T (f ) = f . By definition this is the unique solution of the equa-
tion
Z x
f (x) = a + K(x, y)f (y) dy, x ∈ [0, 1].
0
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 2, Solutions
31.1.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, January 31, 12:00 o’clock.

1. Let M (m × n, R) denote the space of all (real) m × n-matrices. Give

M (m×n, R) a (natural) topology by identifying a matrix A = (aij ) ∈
M (m×n, R) with a point (a11 , a12 , . . . , a1n , a21 , . . . , a2n , . . . , am m
1 , . . . an ) ∈
Rmn and verify that M (m × n, R) is a topological mn-manifold.
We have defined a topology of M (m × n, R) also by a norm |A| =
sup{|Ah| : |h| = 1}. Show that these two topologies are the same.
Solution. Define ϕ : M (m × n, R) → Rnm , by setting

ϕ (aji ) = (a11 , a12 , . . . , a1n , a21 , . . . , a2n , . . . , am m

1 , . . . an ).

Then ϕ is a linear bijection. Let

T = {ϕ−1 U : U ⊂ Rnm open}.

Then T is a topology on M (m × n, R), more precisely the topology

induced by ϕ. Since ϕ is a bijection, it follows from the definition of
T that U ⊂ M (m × n, R) is open if and only if ϕU ⊂ Rnm is open.
Hence ϕ is a homeomorphism.
Since ϕ is a homeomorphism and Rnm is Hausdorff and N2 , also
M (m × n, R) is Hausdorff and N2 . It follows that M (m × n, R) is a
topological nm-manifold with a global chart ϕ.
Next we prove that T is determined by a norm. Define a norm |·|2
on the vector space M (m × n, R) by setting |A|2 = |ϕ(A)|, where
|ϕ(A)| is the standard norm of ϕ(A) ∈ Rnm .
Then the mapping
ϕ between normed spaces M (m × n, R), | · |2 and Rnm is norm
preserving, therefore ϕ is a linear isomorphism. Consequently, if T 0
is the topology on M (m × n, R) determined by the
norm | · |2 , the
mapping ϕ is a homeomorphism M (m × n,
R), T 0 → Rnm . Since ϕ
is also a homeomorphism M (m × n, R), T → Rnm , it follows that
T0=T.
Let us still prove that the operator norm |·| and |·|2 are equivalent,
so that the norm |·| determines the same topology T . Let A = (aij ) ∈
M (m × n, R) be arbitrary and let h ∈ Rn , with |h| = 1. Denoting
2

by Ai the ith row of A, we get

m m m
X X
2 X 2
|Ah|2 = (Ah)2i = (Ai )T · h ≤ (Ai )T |h|2
i=1 i=1 i=1
m m X
n
2
X X
= (Ai )T = (aij )2 = |A|22 .
i=1 i=1 j=1

hence |A| ≤ |A|2 . On the other hand,

m X
X n m X
X n n
X
|A|22 = (aij )2 = (Aej )2i = |Aej |2 ≤ n|A|2 ,
i=1 j=1 i=1 j=1 j=1
√
so |A| ≤ |A|2 ≤ n|A|. Hence the norms | · | and | · |2 are equivalent.
Note: In general, if E is a finite dimensional vector space, all
norms on E are equivalent.

2. Prove that GL(n, R) is a disconnected topological n2 -manifold.

2
Solution. Let ϕ : M (n × n, R) → Rn be as in the previous exercise,
so ϕ is a homeomorphism between topological n2 -manifolds. Let us
prove that the determinant det : M (n × n, R) → R,

X
det (Aij ) = sgn(σ)A1σ(1) A2σ(2) · · · Anσ(n) ,
σ∈Sn

is continuous. Here Sn is the set of permutations of {1, 2, . . . , n} and

sgn : Sn → {−1, 1} is the sign of permutation. Let f = det ◦ϕ−1 and
2
x = (x1 , x2 , . . . , xn2 ) ∈ Rn . Then
X
f (x) = sgn(σ)xσ(1) xn+σ(2) x2n+σ(3) · · · x(n−1)n+σ(n) ,
σ∈Sn

and therefore f is continuous. Since ϕ is a homeomorphism, also

det = f ◦ ϕ is continuous.
Since det is continuous and
−1
GL(n, R) = det(R \ {0}),
we see that GL(n, R) is an open subset of the topological n2 -manifold
M (n × n, R). In particular, GL(n, R) equipped with the relative to-
pology is a topological n2 -manifold. The image of a connected set
under a continuous mapping is connected. Since R \ {0} is discon-
nected, the same holds true for GL(n, R).

3. Let sl(n, R) be the set of those n × n matrices whose trace vanishes:

n
X
sl(n, R) = {A = (aij ) ∈ M (n × n, R) : tr A = aii = 0}.
i=1

Prove that sl(n, R) is a topological manifold.

 3

Solution. Note that tr : M (m × n, R) → R is a linear map and

sl(n, R) = ker tr
is thus a closed linear subspace of M (m × n, R) of dimension
dim ker(tr) = dim M (n × n, R) − dim Im(tr) = n2 − 1.

Since all finite dimensional vector spaces over R are homeomorphic

to some Rk and hence topological manifolds it follows that sl(n, R),
in particular, is a topological manifold (of dimension n2 − 1).

4. Prove that in the implicit function theorem (Theorem 1.14 in the

lecture notes)

−1
ϕ0 (x0 ) = − u0 (y0 ) v 0 (x0 ),
where v(x) = f (x, y0 ).
Solution. Adopting the notation of Theorem 1.14 we have
f (x, ϕ(x)) = 0
in a neighbourhood of x0 . Let us denote h : Rm → Rn+m , h(x) =
(x, ϕ(x)). Differentiating this at x0 using the chain rule we obtain
0 = f 0 (x0 , ϕ(x0 ))h0 (x0 ).
Note that ϕ(x0 ) = y0 so that 0 = f 0 (x0 , y0 )h0 (x0 ). Inspecting the
Jacobian matrix f 0 (x0 , y0 ) we note that the first m columns make up
the Jacobian matrix for v 0 (x0 ) and the next n columns correspond to
u0 (y0 ). Similarly the first m rows of h0 (x0 ) form the identity matrix
Im×m and the remaining n rows the Jacobian matrix for ϕ0 (x0 ).
Using the matrix block shorthand we may write
 
0 0 Im×m
0 = (v (x0 ) u (y0 ))
ϕ0 (x0 )
from which we get
0 = v 0 (x0 ) + u0 (y0 )ϕ0 (x0 ).
Since u0 (y0 ) is invertible the claim follows from this by rearranging
terms and applying (u0 (x0 ))−1 to both sides.

5. Prove lemma 1.7 in the lecture notes: Let G ⊂ Rm be open and J ⊂

G be a closed line segment with endpoints a and b. Let f : G → Rn
be differentiable at each point of J. Then there exists, for all v ∈ Rn ,
a point xv ∈ J such that v · (f (b) − f (a)) = v · Df (xv )(b − a). In
particular, if |Df (x)| ≤ M for all x ∈ J, we have |f (b) − f (a)| ≤
M |b − a|.
Solution. Let γ : [0, 1] → G be the map γ(t) = a + t(b − a), in
other words γ([0, 1]) = J, and gv : Rn → R the map gv (x) = v · x
where v · x = v T x in terms of matrix multiplication. Now, by the
4

mean value theorem for real valued functions there exists a point
ξ ∈ (0, 1) such that
v · (f (b) − f (a)) = v · (f (γ(1)) − f (γ(0)))
= (gv ◦ f ◦ γ)(1) − (gv ◦ f ◦ γ)(0)
= (gv ◦ f ◦ γ)0 (ξ).
On the other hand,
(gv ◦ f ◦ γ)0 (ξ) = D(gv ◦ f )(γ(ξ))γ 0 (ξ)
= ((Dgv ◦ f )Df )(γ(ξ))γ 0 (ξ)
= v · Df ((γ(ξ)))(b − a).
Choose xv = γ(ξ), and the first claim follows. In particular, as this
holds for all v ∈ Rn , choose v = f (b) − f (a) and suppose that
|Df (x)| ≤ M for all x ∈ J. Now,
|v|2 = |v · v| = |v · (f (b) − f (a))| = |v · f 0 (xv )(b − a)|
= |v||f 0 (xv )||b − a| ≤ |v|M |b − a|
and
|f (b) − f (a)| ≤ M |b − a|.

6. Let f : R2 → R2 , f (x, y) = (x, g(x, y)) where


 y − x2 for y > x2
g(x, y) = 0 for |y| ≤ x2
 y + x2 for y < −x2

Prove that f is continuous everywhere and differentiable at the ori-

gin. Compute Jf (0). Is f locally injective at the origin?
Solution. Clearly g is continuous for |y| =6 x2 . Let |y| = x2 and
ε = (ε1 , ε2 ) such that ε1 , ε2 > 0. Now,
|g(x + ε1 , y + ε2 ) − g(x, y)| ≤ ||y + ε2 | − (x + ε1 )2 | −→ 0
as |ε| −→ 0. Hence g is continuous everywhere. Since the first com-
ponent of f is also continuous, f is continuous everywhere. Let ε be
as before. If |ε2 | ≤ ε21 , then |f (ε) − f (0) − ε| = |(ε1 , 0) − (ε1 , ε2 )| =
|ε2 | ≤ ε21 . If, on the other hand ε21 < |ε2 |, then |f (ε) − f (0) − ε| =
|(ε1 , ε2 ± ε21 ) − (ε1 , ε2 )| = ε21 . So, for all ε > 0
|f (ε) − f (0) − ε| ε2
≤ 1 ≤ |ε1 | −→ 0
|ε| |ε|
as |ε| −→ 0. Thus, f is differentiable at the origin. By simple subs-
titution, Jf (0) = det I = 1, however, since g(h, h) = g(h, −h2 ) for
all h > 0, f is not locally injective at the origin.
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 3, Solutions
7.2.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, February 7, 12:00 o’clock.

1. Let M be the following set

M = {(x, y, |x|) ∈ R3 : (x, y) ∈ R2 }.

Construct a C ∞ atlas A on M so that (M, Ā) becomes a diffe-

rentiable 2-manifold. Is it diffeomorphic with R2 ? If yes, what is a
diffeomorphism f : M → R2 .
Solution. Let ϕ : M → R2 , ϕ(x, y, |x|) = (x, y). The mapping ϕ is
a homeomorphism when M has the relative topology. Thus M is a
topological 2-manifold. Let A = {(M, ϕ)} and Ā be the maximal
C ∞ -atlas containing (M, ϕ). Then (M, Ā) is a 2-dimensional C ∞
manifold. It is diffeomorphic to R2 since f = ϕ is a diffeomorphic
between (M, Ā) and R2 . That is, each local representation of f is a
C ∞ -diffeomorphism.

2. Let M be a differentiable manifold, p ∈ M, I ⊂ R an open interval,

and 0 ∈ I.
(a) Let γ : I → M be a C ∞ -path such that γ(0) = p. Show that
γ̇0 ∈ Tp M.
(b) Let v ∈ Tp M. Show that there exists a C ∞ -path γ : I → M
such that γ̇0 = v.
Solution.
(a) We have to show that the functional γ̇0 : C ∞ (p) → R is linear
and satisfies the so called Leibniz rule. For the first we have

0
γ˙0 (af + bg) = (af + bg) ◦ f (0) = (af ◦ γ + bg ◦ γ)0 (0)
= a(f ◦ γ)0 (0) + b(g ◦ γ)0 (0) = aγ˙0 (f ) + bγ̇0 (g)

and for the second similarly

γ̇0 (f g) =(f ◦ γ · g ◦ γ)0 (0)

=f ◦ γ(0)(g ◦ γ)0 (0) + g ◦ γ(0)(f ◦ γ)0 (0)
=f (p)γ̇0 (g) + g(p)γ̇0 (f ),

given any f, g ∈ C ∞ (p) and a, b ∈ R.

2

(b) Let us take a chart (U, x) at p. Given a vector v ∈ Tp M , it may

be written as a sum of the basis elements, i.e.,
Xn
v= ai (∂i )p .
i

Set w = (a1 , . . . , an ) ∈ Rn and consider the path α : R → Rn ,

α(t) = x(p) + wt, t ∈ R.
For a sufficiently small neighborhood I of 0 we have α(I) ⊂ xU
so the composition γ(t) = x−1 ◦ α(t) yields a C ∞ -path defined
on I. Now the chain rule gives
γ̇0 (f ) = (f ◦ γ)0 (0) = (f ◦ x−1 ◦ α)0 (0)
= α0 (0) · ∇ f ◦ x−1 α(0)

Xn
= ai Di (f ◦ x−1 )(α(0))
i=1
Xn
= ai (∂i )p f = v(f )
i
∞
for every f ∈ C (p).

3. (a) Let M, N, and L be differentiable manifolds and let f : M → N

and g : N → L be C ∞ -mappings. Show that
(g ◦ f )∗p = g∗f (p) ◦ f∗p
for all p ∈ M.
(b) Let f : M → M be the identity mapping f = id. Show that
f∗p = id : Tp M → Tp M
for every p ∈ M.
Solution.
(a) Given any v ∈ Tp M and h ∈ C ∞ (g ◦ f )(p) ,

(g ◦ f )∗p v(h) = v(h ◦ g ◦ f ) = f∗p v(h ◦ g) = g∗f (p) (f∗p v)(h).

(b) Let us compute
id∗p v(h) = v(h ◦ id) = v(h), v ∈ Tp M, h ∈ C ∞ (p),
thus id∗p v = v for all v ∈ Tp M .

4. Let M m and N n be differentiable manifolds, p ∈ M m , and f : M m →

N n a smooth mapping such that f∗ : Tp M m → Tf (p) N n is a linear
isomorphism. Prove that f is a local diffeomorphism at p ∈ M m .
Solution. First, since f∗p is a linear isomorphism, the tangent spaces
Tp M m and Tf (p) N n are isomorphic and hence dim M = dim N .
 3

−1 ∞

at f (p). Now y ◦ f ◦
Let (U, x) be a chart at p and (V, y) a chart
−1 0
x : xU → yV is C and det y ◦ f ◦ x (x(p)) 6= 0 since f∗p is
invertible. By the inverse mapping theorem there exist a neighbor-
hood U0 of p and a neighborhood V0 of f (p) such that the mapping
y ◦ f ◦ x−1 |xU0 has a C 1 smooth inverse mapping g : yV0 → xU0 .
Since y ◦ f ◦ x−1
has continuous partial derivatives of every order
0
and det y ◦ f ◦ x−1 (q) 6= 0, q ∈ xU0 , we can conclude by using Cra-
mer’s rule that g has continuous partial derivatives of every order.
Hence f is a local C ∞ diffeomorphism.

∗5. (Embedding of RP 2 into R4 .) Let RP 2 be the real 2-dimensional

projective space (see Example 0.12.5 in lecture notes for the defini-
tion). Define a mapping F : R3 → R4 ,
F (x, y, z) = (x2 − y 2 , xy, xz, yz).
Let S2 ⊂ R3 be the unit sphere and let ϕ = F |S2 . Consider the
mapping ϕ̃ : RP 2 → R4 ,


ϕ [p] = ϕ(p), p = (x, y, z).
Prove that:
(a) ϕ̃ is well-defined,
(b) ϕ̃ is injective,
(c) ϕ̃ is an immersion.
Solution.
(a) By definition of RP 2 , the antipodal points of S 2 are identi-
fied: (x, y, z) ∼ (−x, −y, −z). Clearly ϕ̃ is well-defined since
F (x, y, z) = F (−x, −y, −z).
(b) Suppose F (x1 , y1 , z1 ) = F (x2 , y2 , z2 ). Then, one can easily conclu-
de that (x2 , y2 , z2 ) = (x1 , y1 , z1 ) or (−x1 , −y1 , −z1 ).
(c) Since the quotient map π : S 2 → RP 2 is a local diffeomorp-
hism, it suffices to check that the map F |S 2 : S 2 → R4 is an
immersion. Consider the differential of F :
 
2x −2y 0
y x 0
dF =  z
.
0 x
0 z y
Let p = (x, y, z) ∈ S2 and consider first the case x2 + y 2 6= 0.
Then rank dF = 3 since there exists non-singular 3 × 3 minor
of dF . If x = y = 0, then z 2 = 1 and rank dF = 2. However, in
that case  
0 0 0
 0 0 0
dF = z 0 0

0 z 0
4

and hence dF maps the xy-plane = Tp S2 to the zt-plane ⊂ R4 =

TF (p) R4 . Hence in any case dF is injective.
Since RP 2 is compact, the map ϕ̃ is closed (smooth injective),
which shows that ϕ̃ is a homeomorphism to its image and hence
an embedding.

∗6. If v ∈ Tp M is a (non-zero) vector and f is a smooth function in

a neighborhood of p, the (directional) derivative Dv f := vf makes
sense. Suppose that we are given two linearly independent vectors
v, w ∈ Tp M . What problems (if any) do you face if you want to
2
define a nice “second order directional derivative”, say Dvw ?
Solution. Note that if we only use that v, w ∈ Tp M , then, wvf is not
well-defined since vf is only defined at p. If using two vector fields
V, W in a neighborhood of p, then W V f, V W f are well-defined, but
V W f 6= W V f in general.
Consider two linearly independent vectors v, w ∈ Tp M . Then we
may find a chart x = (x1 , . . . , xn ) at p such that v = ∂1 and w = ∂2
(verify this!). Now a second order partial derivative of a function
f , say ∂ij2 f (p) could be defined as the corresponding second order
partial derivative of the local representation of f , but this is not well-
defined! It depends on the choice of the chart. So it is not determined
solely by the vectors v, w. This is where we need connection!
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 4, Solutions
14.2.2022

1. Let M, M1 , M2 , N, N1 , and N2 be differentiable manifolds.

(a) Let π1 : M × N → M and π2 : M × N → N be the canonical
projections. For (p, q) ∈ M × N , define
τ : T(p,q) (M × N ) → Tp M ⊕ Tq N
by setting
τ v = π1∗ v + π2∗ v.
Show that τ is an isomorphism. Thus we may identify
T(p,q) (M × N ) = Tp M ⊕ Tq N . We use this identification in the
sequel.
(b) Let f : M1 × M2 → N be a smooth mapping. For (p, q) ∈ M1 ×
M2 , we define mappings fp : M2 → N and f q : M1 → N ,
fp (q) = f q (p) = f (p, q).
Show that
f∗(p,q) (v + w) = (f q )∗p v + (fp )∗q w
for all v ∈ Tp M1 and w ∈ Tq M2 .
(c) Let fi : M → Ni , i = 1, 2, be smooth mappings. Show that the
mapping (f1 , f2 ) : M → N1 × N2 is smooth and that
(f1 , f2 )∗p v = f1∗p v + f2∗p v
for all v ∈ Tp M, p ∈ M .
(d) Let fi : Mi → Ni , i = 1, 2, be smooth mappings. Show that
the mapping
f1 × f2 : M1 × M2 → N1 × N2 , (f1 × f2 )(p, q) =
f1 (p), f2 (q) , is smooth and that
(f1 × f2 )∗(p,q) (v + w) = f1∗p v + f2∗2 w
for all v ∈ Tp M, w ∈ Tq M2 , (p, q) ∈ M1 × M2 .
Solution. Denote mi = dim Mi , ni = dim Ni , i = 1, 2, and m =
dim M, n = dim N .
(a) Let (U, x) and (V, y) be charts at p ∈ M and q ∈ N respecti-
vely, and (W, z) = (U × V, (x ◦ π1 , y ◦ π2 )) be the induced chart on
M × N at (p, q). Then, it suffices to show that

∂



∂
  ∂x i p 
1≤i≤m
(1) τ =

∂
∂z i (p,q)  ∂yi−m m<i≤m+n
q
2

for all 1 ≤ i ≤ m + n since τ maps a basis of T(p,q) (M × N ) to a

basis of Tp M ⊕ Tq N and is therefore an isomorphism. Let us prove
(1). We assume that i ≤ m. If h ∈ C ∞ (p), then
  !  
∂ ∂
π1∗ h= (h ◦ π1 )
∂z i (p,q) ∂z i (p,q)
= Di (h ◦ π1 ◦ z −1 )(z(p, q))

0
= h ◦ π1 ◦ z −1 ◦ (t 7→ z(p, q) + tei ) (0)
1≤m
0
= (h ◦ x−1 ◦ (t 7→ x(p) + tei ) (0)
 
−1 ∂
= Di (h ◦ x )(x(p)) = h.
∂xi p
Similarly, if h ∈ C ∞ (q), then
  !
∂
0
π2∗ i
h = h ◦ π2 ◦ z −1 ◦ (t 7→ z(p, q) + tei ) (0)
∂z (p,q)
1≤m
0
= t 7→ h(q) (0) = 0.
Hence
       
∂ ∂ ∂ ∂
τ = π1∗ + π2∗ = +0
∂z i (p,q) ∂z i (p,q) ∂z i (p,q) ∂xi p
and (1) holds. The case i > m follows in a similar way.
(b) Denote the projections πi : M1 × M2 → Mi , i = 1, 2. Take
charts (U, x), (V, y) at p and q, respectively, and let (W, z) = (U ×
V, (x ◦ π1 , y ◦ π2 )) be the induced chart at (p, q). By linearity it is
enough to show that the claim holds if
 
∂
v+w = ,
∂z i (p,q)
where 1 ≤ i ≤ m1 + m2 . For 1 ≤ i ≤ m1 and h ∈ C ∞ f (p, q)


  !  
∂ ∂
f∗ i
h= i
(h ◦ f ) = Di (h ◦ f ◦ z −1 )(z(p, q))
∂z (p,q) ∂z (p,q)
=(h ◦ f ◦ z −1 ◦ (t 7→ z(p, q) + tei ))0 (0)
=(h ◦ f q ◦ x−1 ◦ (t 7→ x(p) + tei ))0 (0)
 
q −1 q ∂
=Di (h ◦ f ◦ x )(x(p)) = f∗ h.
∂xi p
The case m < i ≤ m + n is similar.
(c) Denote the projections πi : N1 × N2 → Ni , i = 1, 2. We write
f = (f1 , f2 ) : M → N1 × N2 . Then πi ◦ f = fi , i = 1, 2. To show
smoothness of f , let p ∈ M, (q1 , q2 ) ∈ N1 × N2 , and let (U, x) be a
chart on M at p ∈ M and (Vi , yi ) be a chart on Ni at qi , i = 1, 2.
 3

Then (W, z) = (V1 × V2 , (y1 ◦ π1 , y2 ◦ π2 )) is a chart on N1 × N2 at

(q1 , q2 ). Now
z ◦ f ◦ x−1 = (y1 ◦ π1 ◦ f ◦ x−1 , y2 ◦ π2 ◦ f ◦ x−1 )
= (y1 ◦ f1 ◦ x−1 , y2 ◦ f2 ◦ x−1 )
is smooth since f1 and f2 are. Hence f is smooth. Since πi ◦ f = fi
we use the identification τ = id from (a) to get
f∗ v = π1∗ f∗ v + π2∗ f∗ v = f1∗ v + f2∗ v.
(d) To show the smoothness, fix (p1 , p2 ) ∈ M1 × M2 and (q1 , q2 ) ∈
N1 ×N2 . Let (Ui , xi ) be a chart on Mi at pi and let (Vi , yi ) be a chart
of Ni at qi , i = 1, 2. Then (W, z) = (U1 × U2 , (x1 ◦ π1 , x2 ◦ π2 )) is a
chart at (p1 , p2 ) ∈ M1 × M2 and (R, w) = (V1 × V2 , (y1 ◦ π1 , y2 ◦ π2 ))
is a chart at (q1 , q2 ) ∈ N1 × N2 . Now
w ◦ (f1 × f2 ) ◦ z −1
= y1 ◦ π1 ◦ (f1 × f2 ) ◦ z −1 , y2 ◦ π2 ◦ (f1 × f2 ) ◦ z −1

= y1 ◦ f1 ◦ π1 ◦ z −1 , y2 ◦ f2 ◦ π2 ◦ z −1

= y1 ◦ f1 ◦ x−1 −1


1 ◦ π 1 , y2 ◦ f 2 ◦ x 2 ◦ π 2

is a smooth map. This shows that f1 × f2 is smooth. By the previous

results, we get
(b)
(f1 × f2 )∗(p,q) (v + w) = ((f1 × f2 )q )∗p v + ((f1 × f2 )p )∗q w



= s 7→ (f1 (s), f2 (q) ∗p v + t 7→ (f1 (p), f2 (t)) ∗q w
(c)
= f1∗p v + (s 7→ f2 (s))∗p v + f2∗q w + (t 7→ f1 (p))∗q w
= f1∗p v + f2∗q w.

2. Suppose that Mi is a submanifold of Ni , i = 1, 2. Prove that M1 ×M2

is a submanifold of N1 × N2 .
Solution. Denote by ik : Mk → Nk the immersions, k = 1, 2. Since
they are homeomorphisms onto their images it follows that i1 ×
i2 : M1 × M2 → N1 × N2 is a homeomorphism onto its image. It
suffices to see that (i1 × i2 )∗(p,q) is an injection. For this let (v, w) ∈
Tp M1 ⊕ Tq M2 be such that (i1 × i2 )∗(p,q) (v, w) = 0. Then
(1d)
0 = (i1 × i2 )∗(p,q) (v, w) = (i1∗p v, i2∗q w),
thus i1∗p v = 0 and i2∗q w = 0. Since i1∗p and i2∗q are injections we
have that (v, w) = 0.
3. Let G be a Lie group, e ∈ G the neutral element, and i : G →
G, i(g) = g −1 . Prove that i∗ e v = −v for all v ∈ Te G.
4

Solution. The idea is to “differentiate both sides” of the identity

gg −1 = e with respect to g. For g ∈ G, define Ig : G → G × G and
I g : G → G × G as
Ig (h) = (g, h), I g (h) = (h, g).
Next we define the multiplication m : G × G → G, m(g, h) = gh
and also the left-translation Lg : G → G and the right-translation
Rg : G → G by
Lg (h) = gh, Rg (h) = hg.
Let us compute the tangent map of m. If g, h ∈ G, v ∈ Tg G, and
w ∈ Th G, then by (1b) we have
(v, w) = idG×G ∗(g,h) (v, w) = (I h )∗g v + (Ig )∗h w.

Therefore
m∗(g,h) (v, w) = m∗(g,h) (I h )∗g v + (Ig )∗h w

= m∗(g,h) (I h )∗g v + m∗(g,h) (v, w) (Ig )∗h w

= m ◦ I h ∗g v + m ◦ Ig ∗h w

= (Rh )∗g v + (Lg )∗h w.

Since gg −1 = e for every g ∈ G, m ◦ (idg , i) is a constant map, so its
tangent map is zero. Combining the above with (1c) we get




0 = m ◦ (idG , i) ∗g v = m∗(g,g−1 ) (idG , i)∗g v = m∗(g,g−1 ) v, i∗g v



= Rg−1 ∗g v + Lg ∗g−1 (i∗g v).


Since Lg−1 ◦ Lg = idG , operating with Lg−1 ∗e on both sides of the
previous identity we get



0 = Lg−1 ∗e Rg−1 ∗g v + i∗g v.
Hence


i∗g v = − Lg−1 ∗e
Rg−1 ∗g
v,
and taking g = e we get the desired formula since Le = idG = Re .

4. Let V be a vector field on M . Show that the following conditions

are equivalent:
(a) V ∈ T (M ).
(b) If v i : U → R, i = 1, . . . , n, are the component functions of V
with respect to a chart (U, x), x = (x1 , . . . , xn ), i.e.
V |U = v i ∂i ,
then v i ∈ C ∞ (U ).
(c) If U ⊂ M is open and f : U → R is a smooth function, then
the function V f : U → R, (V f )(p) = Vp f, is smooth.
 5

Solution. Recall that if (U, x) is a chart on M , then (T U, x̄) is a

chart on T M . Here x̄ : T U → xU × Rn and

x̄(p, v) = (x1 (p), . . . , xn (p), vx1 , . . . , vxn )

for p ∈ U and v ∈ Tp M .
(a) ⇐⇒ (b): Let (U, x) be a chart. Note that given s ∈ Rn

x ◦ V ◦ x−1 (s)
= (s, Vx−1 (s) (x1 ), . . . , Vx−1 (s) (xn ))
= (s, v i ◦ x−1 (s)(∂i )x−1 (s) x1 , . . . , v i ◦ x−1 (s)(∂i )x−1 (s) xn )
= (s, v 1 ◦ x−1 (s), . . . , v n ◦ x−1 (s))

where n = dim M . Therefore x ◦ V ◦ x−1 is smooth if and only if

v i ◦ x−1 is for i = 1, . . . , n. Thus V ∈ T (M ) if and only if each v i |U
is smooth.
(b) =⇒ (c): Let U ⊂ M be open and f ∈ C ∞ (U ). Thus if (W, x)
is a chart at p ∈ U (with W ⊂ U ) it follows that f ◦ x−1 is a smooth
function, thus Di (f ◦ x−1 ) is a smooth map as well. We have

V f (p) = v i (p)(∂i )p f = v i (p)Di (f ◦ x−1 )(x(p))

and since, by assumption the functions v i are smooth it follows that

V f is smooth.

(c) =⇒ (b): Let (U, x) be a chart. Since xk is smooth, by as-

sumption we obtain that

V xk = v i ∂i xk = v k

is a smooth map, k = 1, . . . , dim M .

∗5. (a) Compute [X, Y ], [X, Z] and [Y, Z] when X, Y, Z ∈ T (R3 ) are
vector fields

∂ y ∂ ∂ x ∂ ∂
X= − , Y = + and Z = .
∂x 2 ∂z ∂y 2 ∂z ∂z

(b) Does there exists a smooth submanifold S of R3 such that the

vectors Xp and Yp form a basis of Tp S for all p ∈ S?
(c) Does there exists a smooth submanifold M of R3 such that the
vectors Xp and Zp form a basis of Tp M for all p ∈ M ?
6

Solution. (a):
[X, Y ]f = X(Y f ) − Y (Xf )
   
∂ ∂f x ∂f y ∂ ∂f x ∂f
= + − +
∂x ∂y 2 ∂z 2 ∂z ∂y 2 ∂z
   
∂ ∂f y ∂f x ∂ ∂f y ∂f
− − − −
∂y ∂x 2 ∂z 2 ∂z ∂x 2 ∂z
∂ 2f 1 ∂f x ∂ 2f y ∂ 2f xy ∂ 2 f
= + + − −
∂x∂y 2 ∂z 2 ∂x∂z 2 ∂z∂y 4 ∂ 2z
∂ 2f 1 ∂f y ∂ 2f x ∂ 2f xy ∂ 2 f
− + + − +
∂y∂x 2 ∂z 2 ∂y∂z 2 ∂z∂x 4 ∂ 2z
∂f
= = Zf.
∂z
Thus, [X, Y ] = Z. Similarly,
[X, Z]f = X(Zf ) − Z(Xf )
∂ 2f y ∂ 2f ∂ 2f y ∂ 2f
= − − + = 0,
∂x∂z 2 ∂ 2 z ∂z∂x 2 ∂ 2 z
and
[Y, Z]f = Y (Zf ) − Z(Y f )
∂ 2f x ∂ 2f ∂ 2f x ∂ 2f
= + − − = 0.
∂y∂z 2 ∂ 2 z ∂z∂y 2 ∂ 2 z
Therefore [X, Z] = [Y, Z] = 0.
(b): Since [X, Y ] = Z and Zp is not a linear combination of Xp and
Yp for any p, such submanifold S cannot exits (cf. Theorem 3.21 in
the lecture notes).
(c): Yes, for example
M = {(x, y, z) ∈ R3 : y = 0}
is a submanifold of R3 and Xp , Zp ∈ Tp M for every p ∈ M .
P4
∗6. Let S3 = {(x1 , x2 , x3 , x4 ) ∈ R4 : i=1 (x
i 2
) = 1} and
X = −x2 ∂1 + x1 ∂2 + x4 ∂3 − x3 ∂4 ,
Y = −x3 ∂1 − x4 ∂2 + x1 ∂3 + x2 ∂4 ,
Z = −x4 ∂1 + x3 ∂2 − x2 ∂3 + x1 ∂4 ,
V = x1 ∂1 + x2 ∂2 + x3 ∂3 + x4 ∂4 ,
where ∂i = ∂/∂xi for i = 1, . . . , 4, are the standard coordinate
vectors in R4 . Show that:
(a) Xp , Yp , Zp and Vp are mutually orthogonal unit vectors of R4
at each point p ∈ S3 .
(b) X, Y, Z ∈ T (S3 ).
 7

(c) S3 is parallelizable.
Solution. (a): Computing the standard inner products (”dot pro-
ducts”) in R4 we immediately see that, for every p ∈ S3 :
Xp · Xp = Yp · Yp = Zp · Zp = Vp · Vp = 1
and
Xp · Yp = Xp · Zp = Xp · Vp = Yp · Zp = Yp · Vp = Zp · Vp = 0.
(b): First we observe that Tp S3 is a 3-dimensioal vector subspace

⊥
of R4 (= Tp R4 ) for every p ∈ S3 . Hence Tp S3 , the orthogonal
complement with respect to the standard inner product in R4 , is 1-
dimensional.
Let f : R4 → R be f (x) = 4i=1 (xi )2 − 1, for x = (x1 , . . . , x4 ). By
P
direct computation we have Xp f = Yp f = Zp f = 0 and Vp f = 2 for
every p ∈ S3 .
To prove that X, Y, Z ∈ T S3 we decompose orthogonally an ar-
bitrary W ∈ Tp R4 into Wp = Wp> + Wp⊥ , where Wp> ∈ Tp S3 and

⊥
Wp⊥ ∈ Tp S3 . Since f |S3 = 0, we have Wp> f = 0 by Lemma
3.22 in the lecture notes. Above we noticed that Vp f = 2 for eve-

⊥
ry p ∈ S3 , so Vp 6∈ Tp S3 , and therefore the Tp S3 -component of
Vp , Vp⊥ , is non-zero (in fact, soon we will see that Vp = Vp⊥ ). Since

⊥
Tp S3 is 1-dimensional, it is spanned by Vp⊥ 6= 0 for every p ∈ S3 .
Furthermore,
2 = Vp f = (Vp> + Vp⊥ )f = Vp> f +Vp⊥ f,
|{z}
=0

so Vp⊥ f 3
= 2 for every p ∈ S . Above we noticed that Xp f = 0, so
0 = Xp f = (Xp> + Xp⊥ )f = Xp> f +Xp⊥ f,
| {z }
=0

where Xp> f = 0 by Lemma 3.22. Hence also Xp⊥ f = 0. Since Xp⊥ ∈

⊥
⊥
Tp S3 and Tp S3 is spanned by Vp⊥ , we have
Xp⊥ = λ(p) Vp⊥
|{z}
∈R

for every p ∈ S3 . Then

0 = Xp⊥ f = λ(p)Vp⊥ f = 2λ(p),
so λ(p) = 0 and consequently Xp⊥ = 0 for every p ∈ S3 . This proves
that Xp ∈ Tp S3 ∀p ∈ S3 . Similarly, we can prove that Yp , Zp ∈
Tp S3 ∀p ∈ S3 .
Finally, X, Y, Z ∈ T (S3 ), i.e. are smooth vector fields on S3 , by
Lemma 3.3.(b).
8

(c): By (a) and (b), the smooth vector fields X, Y, Z ∈ T (S3 ) form
a global frame on S3 , hence S3 is parallelizable.
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 5, Solution
21.2.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, February 21, 12:00 o’clock.

1. Let X, Y, T ∈ T (R3 ) be as in Example 3.12, that is

∂ ∂ ∂ ∂
X= , Y = +x , T = .
∂x ∂y ∂t ∂t

Define paths αi , β : R → R3 , i = 1, . . . , 4, as follows: For each (fixed)

t > 0, let α1 be the integral curve of X starting at 0 ∈ R3 , i.e.
α1 (0) = 0, let α2 be the integral curve of Y starting at α1 (t) (α2 (0) =
α1 (t)), let α3 be the integral curve of −X starting at α2 (t) (α3 (0) =
α2 (t)), and let α4 be the integral curve √ of −Y starting at α3 (t)
(α4 (0) = α3 (t)). Finally, let β(t) = α4 ( t). (The values t ≤ 0 are
treated in an obvious way). Verify that β̇0 = T (= [X, Y ]).
Solution. Fix t > 0. Then one can easily deduce that
(
α10 (s) = Xα1 (s)
=⇒ α1 (s) = (s, 0, 0) =⇒ α1 (t) = (t, 0, 0);
α1 (0) = (0, 0, 0)

(
α20 (s) = Yα2 (s)
=⇒ α2 (s) = (t, s, ts) =⇒ α2 (t) = (t, t, t2 );
α2 (0) = (t, 0, 0)

(
α30 (s) = −Xα3 (s)
=⇒ α3 (s) = (t − s, t, t2 ) =⇒ α3 (t) = (0, t, t2 );
α3 (0) = (t, t, t2 )

(
α40 (s) = −Yα4 (s)
2
=⇒ α4 (s) = (0, t − s, t2 ) =⇒ α4 (t) = (0, 0, t2 ).
α4 (0) = (0, t, t )

√
Hence β(t) = α4 ( t) = (0, 0, t) and therefore β̇0 = β 0 (0) =
(0, 0, 1) = T .
2

Moreover, one can check that [X, Y ] = T . That is, given any
smooth function f we have
[X, Y ](f ) = X(Y (f )) − Y (X(f ))
     
∂ ∂f ∂f ∂ ∂f ∂ ∂f
= +x − −x
∂x ∂y ∂t ∂y ∂x ∂t ∂x
∂ 2f ∂f ∂ 2f ∂ 2f ∂ 2f
= + +x − −x
∂x∂y ∂t ∂x∂t ∂y∂x ∂t∂x
∂f
= = T (f ).
∂t

2. A path γ : R → M is periodic if there exists T > 0 such that

γ(t + kT ) = γ(t) for all t ∈ R and k ∈ Z. Let V ∈ T (M ) and
let γ be a maximal integral curve of V . Prove that γ has exactly one
of the following properties:
(a) γ is a constant path,
(b) γ is injective,
(c) γ is periodic and non-constant.
Solution. It is clear that if the maximal integral curve has one of
the listed properties it cannot have another. Hence, it is sufficient to
show that if it is non-constant and non-injective, it is periodic. That
is, suppose γ : I → M is non-constant and non-injective. Hence,
there are a, b ∈ I, a < b, such that γ(a) = γ(b). If we define γa :
I − a → M, γa (t) = γ(t + a), it follows that it is a maximal integral
curve of V starting at γ(a). Similarly γb : I −b → M, γb (t) = γ(t+b)
is a maximal integral curve of V starting at γ(b). Since γ(a) = γ(b)
we obtain by uniqueness that γa = γb . In particular, I − a = I − b,
so I = I + a − b means that I = R. Let us denote T = b − a > 0.
Then, for all t ∈ R
γ(t) = γa (t − a) = γb (t − a)
= γb (t − a + b − b) = γ(t − a + b)
= γ(t + T ).
Finally, let us show by induction that the previous identity implies
periodicity. Let
S = {k ∈ N | γ(t) = γ(t + kT ) ∀t}.
It is trivial that 0 ∈ S. Moreover, we proved above that 1 ∈ S. Now,
let us suppose that k ∈ S, then
γ(t + (k + 1)T ) = γ((t + T ) + kT ) = γ(t + T ) = γ(t),
so k + 1 ∈ S. Hence γ(t) = γ(t + kT ) for all k ∈ N. But now also
γ(t) = γ(t − kT + kT ) = γ((t − kT ) + kT ) = γ(t − kT ) for every
k ∈ N, so γ is periodic.
 3

3. Let G be a Lie group and X ∈ T (G) a left-invariant vector field.

Prove that X is complete.
Solution. Let e ∈ G denote the neutral element and Lg : G → G,
h 7→ gh the left multiplication. Let us recall that X being left-
invariant means that (Lg )∗h Xh = Xgh for all g, h ∈ G.
Given g ∈ G, let θg : Ig → G be the maximal integral curve of X
starting at g. Let us define γ = Lg ◦ θe : Ie → G. First, it satisfies
γ(0) = gθe (0) = gθ(0, e) = ge = g,
so γ is a path starting at g. Moreover, for all t ∈ Ie and f ∈ C ∞ , by
left-invariance,
Xγ(t) f = Xgθ(t,e) = (Lg )∗θ(t,e) Xθ(t,e) f
= Xθ(t,e) (f ◦ Lg ) = θ˙te (f ◦ Lg ) = (f ◦ Lg ◦ θe )0 (t)
= (f ◦ γ)0 (t) = γ̇t f.
Hence, Xγ(t) = γ̇t and γ is an integral curve of X starting at g.
By maximality of θg , Ie ⊂ Ig for all g ∈ G. It remains to show
that Ie = R. By part (b) of Theorem 3.36, Ie = t + Iθ(t,e) , for all
t ∈ Ie . Thus, since Ie ⊂ Iθ(t,e) it follows that t + Ie ⊂ t + Iθ(t,e) = Ie ,
and so Ie can not be bounded from below or above. Since it is an
interval, Ie = R.

4. Prove that θ : R × R2 → R2 ,
1
θ t, (x, y) = t2 + yt + x, t + y ,



2
is a flow and find its infinitesimal generator.
Solution. Clearly θ is a smooth mapping with flow domain R × R2 .
Let p = (x0 , y0 ) ∈ R2 , and t, s ∈ R. It is clear that θ(0, p) = p.
Moreover,
θ(t, θ(s, p)) = θ(t, (s2 /2 + y0 s + x0 , s + y0 ))
= ((t + s)2 /2 + y0 (t + s) + x0 , (t + s) + y0 )
= θ(t + s, p),
and hence θ is a flow on R2 .
We know that θp is the integral curve of V starting at p if
V ((θp (0))i ) (∂i )p = ((θp )i )0 (0) (∂i )p = y0 (∂x )p + (∂y )p .
Hence, the infinitesimal generator of θ is V = y ∂x + ∂y .

∗5. Let V ∈ T (M ) be a smooth compactly supported vector field, that

is, the set
suppV = {p ∈ M : Vp 6= 0}
is compact. Prove that V is complete.
4

Solution. Let V be as above. By the “Fundamental Theorem of

Flows”, Theorem 3.37 of the lecture notes, let θ be its maximal
smooth flow and given p ∈ M , θp : Ip → M the corresponding maxi-
mal integral curve of V starting from p. Now, either θp (Ip ) ⊆ suppV
or θp (Ip ) * suppV. We will show that in both scenarios Ip = R. On
the one hand, if θp (Ip ) ⊆ suppV, it follows by the Escape lemma,
Lemma 3.47 of the lecture notes, by compactness of suppV, that
Ip = R. On the other hand, assume that θp (Ip ) * suppV. Then, the-
re exists e t ∈ Ip such that q = θp (e
t) ∈
/ suppV. Let γ : R → M be the
constant path γ(t) = q. Thus, γ(0) = q and γ̇(t) = 0 = Vq = Vγ(t) .
Hence, γ is the maximal integral curve of V starting from q, and by
uniqueness γ = θq . Furthermore, now
p = θp (0) = θ(0, p) = θ(−t, θ(t, p))
= θ(−t, q) = θq (−t) = γ(−t)
= q,
so p = q and, hence, θp = θq = γ implying that Ip = R. Since
we have proven that Ip = R for all p ∈ M we conclude that V is
complete.

∗6. Let V ∈ T (R2 ) be the vector field

∂ ∂
V(x,y) = (1 + x2 )
+ .
∂x ∂y
Determine the flow θ : D(V ) → R2 of V . Is V complete?
Solution. Fix p = (x0 , y0 ) ∈ R2 . Let γ : Jp → R2 , γ(t) = (x(t), y(t)),
be an integral curve of V starting from p. Then, γ̇t = Vγ(t) , or equi-
valently, (
x0 (t) = 1 + x(t)2 , x(0) = x0
y 0 (t) = 1, y(0) = y0 .
Thus, by integration
(
x(t) = tan(t + arctan(x0 )) = xcos 0 cos t+sin t
t−x0 sin t
y(t) = t + y0 .
for all t s.t. |t + arctan(x0 )| < π/2.
Let Jp = (−π/2 − arctan(x0 ), π/2 −
arctan(x0 )). Then γ : Jp →
2
R , γ(t) = tan(t + arctan(x0 )), t + y0 is an integral curve starting
at p such that x(t) → −∞ as t → −π/2−arctan(x0 ), and x(t) → ∞
as t → π/2 − arctan(x0 ). Hence γ can not be continuously extended
to a larger interval than Jp , so it is maximal, and by uniqueness
γ = θp . In particular, as Ip = Jp 6= R, it follows that θ : D(V ) =
∪p∈R2 (Ip × {p}) → R2 , θ(t, (x, y)) = (tan(t + arctan(x)), t + y) is not
a global flow.
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 6, Solution
28.2.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, February 28, 12:00 o’clock.

1. Let X, Y ∈ T (R3 ) be smooth vector fields defined by

∂ ∂ ∂
Xp = , Yp = + x , p = (x, y, z) ∈ R3 .
∂x ∂y ∂z
Compute the Lie derivative LX Y directly by using the definition
(3.61) in the lecture notes. Verify then that LX Y = [X, Y ] as it
should be.
Solution. First, it is easy to check that the flow of X starting at
p = (x0 , y0 , z0 ) is the mapping θ : R × R3 → R3 ,


θ t, (x0 , y0 , z0 ) = (x0 + t, y0 , z0 ).
Then, θ−t : R3 → R3 is the mapping
θ−t (p) = (x0 − t, y0 , z0 )
∞
Next, given f ∈ C (p) we have
θ(−t)∗θ(t,p) Yθ(t,p) f = Yθ(t,p) (f ◦ θ−t ) = Yθ(t,p) (f (x − t, y, z))
    !
∂ ∂
= + (x0 + t) f (x − t, y, z)
∂y θ(t,p) ∂z θ(t,p)
∂f ∂f
= (x0 , y0 , z0 ) + (x0 + t) (x0 , y0 , z0 ).
∂y ∂z
Hence,
θ(−t)∗θ(t,p) Yθ(t,p) f − Yp f
(LX Y )p f = lim
t→0 t
∂  
t ∂z ∂f ∂
= lim = (x0 , y0 , t0 ) = f,
t→0 t ∂z ∂z p
∂
which means that LX Y = ∂z .
On the other hand, we computed in Example 3.12 (lecture notes)
that [X, Y ] = ∂/∂z. Thus, LX Y = [X, Y ].

2. Let V, W, X ∈ T (M ) and h ∈ C ∞ (M ). Prove that:

(a) LV W = −LW V.
(b) LV [W, X] = [LV W, X] + [W, LV X].
(c) L[V,W ] X = LV LW X − LW LV X.
2

(d) LV (hW ) = (V h)W + hLV W.

(e) If f : M → N is a diffeomorphism, then f∗ (LV W ) = Lf∗ V (f∗ W ).
Solution. Rewriting the Lie derivative as the Lie bracket (Theorem
3.62) and applying Lemma 3.10 we get:
(a) Using the antisymmetry of the Lie bracket we have
LV W = [V, W ] = −[W, V ] = −LW V.
(b) From Jacobi identity and the antisymmetry of the Lie bracket
we get
LV [W, X] = [V, [W, X]] = −[W, [X, V ]] − [X, [V, W ]]
= [W, [V, X]] + [[V, W ], X]
= [W, LV X] + [LV W, X]
(c) Again by applying the Jacobi identity and the antisymmetry of
the Lie bracket we conclude
L[V,W ] X = [[V, W ], X] = [X, [W, V ]]
= −[W, [V, X]] − [V, [X, W ]]
= −[W, [V, X]] + [V, [W, X]]
= −LW [V, X] + LV [W, X]
= LV LW X − LW LV X.
(d) From property (d) in Lemma 3.10 we deduce
LV (hW ) = [V, hW ] = h[V, W ] + (V h)W = hLV W + (V h)W.
(e) Let us define Ve = f∗ V and W
f = f∗ W . By Lemma 3.12 we know
∞
that for any g ∈ C (N )
V (g ◦ f ) = (Ve g) ◦ f and W (g ◦ f ) = (W
f g) ◦ f.
Then, given p ∈ M we can apply the previous identities to get
Vp (W (g ◦ f )) = Vp (W
f (g) ◦ f ) = Vef (p) (W
f g)
and
Wp (V (g ◦ f )) = Wp (Ve (g) ◦ f ) = W
ff (p) (Ve g).
Next, by using the definition of the tangent mapping we deduce
(f∗p [V, W ]p ) g = [V, W ]p (g ◦ f )
= Vp (W (g ◦ f )) − Wp (V (g ◦ f ))
f g) − W
= Vef (p) (W ff (p) (Ve g)

= [Ve , W
f ]f (p) g = [f∗ V, f∗ W ]f (p) g.
Finally, we arrive at
f∗ (LV W ) = f∗ [V, W ] = [f∗ V, f∗ W ] = Lf∗ V (f∗ W ).
 3

3. Let (U, x), x = (x1 , . . . , xn ), be a chart at p and let ∂1 , . . . , ∂n be

the corresponding coordinate vector fields.
(a) Prove that dxip (∂j ) = δji .
(b) Prove that dfp = (∂i )p f dxip is the differential of a function f ∈
C ∞ (p) at p.
Solution.
(a) Since xi ◦ x−1 is the ith component of the identity mapping we
obtain
dxip (∂j ) = (∂j )p xi = Dj (xi ◦ x−1 )(x(p)) = δji .
(b) Given v ∈ Tp M we know that it can be written as
n
X
v= v i (∂i )p ,
i=1

where v i = vxi = dxip v. Then,

n
X n
X
dfp v = vf = (vxi (∂i )p )f = (dxip v)(∂i )p f.
i=1 i=1

Hence dfp = (∂i )p f dxip .

4. Let f, g ∈ C ∞ (M ). Prove that:

(a) d(af + bg) = adf + bdg if a, b ∈ R.
(b) d(f g) = f dg + gdf.
(c) d(f /g) = (gdf − f dg)/g 2 in the set where g 6= 0.
(d) Let J ⊂ R be an open interval such that f (M ) ⊂ J and let
h : J → R be a smooth function. Then d(h ◦ f ) = (h0 ◦ f )df.
(e) If f is a constant function, then df = 0.
Solution.
(a) For all p ∈ M and for all v ∈ Tp M ,
d(af + bg)p v = v(af + bg) = avf + bvg = adfp v + bdgp v.
Hence, d(af + bg) = adf + bdg.
(b) For all p ∈ M and for all v ∈ Tp M ,
d(f g)p v = v(f g) = g(p)vf + f (p)vg
= g(p)dfp v + f (p)dgp v.
Hence, d(f g) = f dg + gdf.
(c) Let p ∈ M such that g(p) 6= 0, then there is a neighbourhood A
such that p ∈ A ⊂ g −1 (R \ {0}). By apliying (b) in A we obtain
df = d((f /g)g) = (f /g)dg + gd(f /g).
Hence, d(f /g) = (1/g)df − (f /g 2 )dg = (gdf − f dg)/g 2 in A.
4

(d) For all p ∈ M and for all v ∈ Tp M , let γ be a smooth path such
that γ˙0 = v. Then,
d(h ◦ f )p v = v(h ◦ f ) = γ˙0 (h ◦ f ) = (h ◦ f ◦ γ)0 (0)
= h0 (f (γ(0)))(f ◦ γ)0 (0) = (h0 ◦ f )(p)γ˙0 f
= (h0 ◦ f )(p)vf = (h0 ◦ f )(p)dfp v
Thus, d(h ◦ f ) = (h0 ◦ f )df .
(e) For all p ∈ M and for all v ∈ Tp M , if f is constant, then
dfp v = vf = 0,
and so df = 0.

5. Suppose that X, Y ∈ T (M ) are smooth vector fields and that θ is

the flow of X and ψ the flow of Y . Suppose, furthermore, that X
and Y are complete, that is D(X) = D(Y ) = R × M . Prove that
the following are equivalent:
(a) [X, Y ] = 0,
(b) (θt )∗ Yp = Yθ(t,p) for all (t, p) ∈ D(X),
(c) (ψt )∗ Xp = Xψ(t,p) for all (t, p) ∈ D(Y ),
(d) θt ◦ ψs = ψs ◦ θt .
Solution.
(b)(c) ⇒ (a)
If (b) holds, then (θ−t )∗ Yθ(t,p) = Yp . Thus, we get from the defini-
tion of Lie derivative that

(θ−t )∗ Yθ(t,p) − Yp
[X, Y ]p = LX Y p = lim = 0.
t→0 t
The same reasoning applied to (c) yields [Y, X]p = 0.
(d) ⇒ (b)(c)
Consider the path
γ : t 7→ θt ◦ ψs (p).
On the one hand we have
ψ (p)
γ̇0 = θ̇0 s = Xψs (p) .
On the other hand, using (d) we have alternatively that γ(t) =
ψs ◦ θt (p). Hence
γ̇0 = ψs∗ θ̇0p = ψs∗ Xp .
Thus we get (c). The same reasoning applied to the path
σ : s 7→ ψs ◦ θt (p)
yields (b).
(b) ⇒ (d)
Let us take t ∈ R and consider the flow
σ(s, p) = θ−t ◦ ψs ◦ θt (p),
 5

whose infinitesimal generator is

θ (p)
σ̇0p = (θ−t )∗ ψ̇0t = (θ−t )∗ Yθt (p) .
Next, using (b) we get
σ̇0p = (θ−t )∗ (θt )∗ Yp = Yp
which, by the uniqueness of flows, means that σ = ψ. This proves
(d).
(a) ⇒ (b)
Let p ∈ M and V : R → Tp M , V (t) = (θ−t )∗ Yθ(t,p) . Since V (0) =
Yp it is enough to show that V (t) is constant. That is, let t0 ∈ R
and s = t − t0 , then
d d
V 0 (t0 ) =



(θ−t )∗ Yθ(t,p) |t=t0 = (θ−t0 −s )∗ Yθ(s+t0 ,p) |s=0
dt ds
d

= (θ−t0 )∗ (θ−s )∗ Yθ(s,θ(t0 ,p)) |s=0
ds
= (θ−t0 )∗ (LX Y )θ(t0 ,p) = 0.
Thus (b) follows.

∗6. Prove Lemma 5.5 in the lecture notes: If α ∈ Λk (V ) and β ∈ Λ` (V ),

then
α ∧ β(v1 , . . . , vk+` ) =
X


(sgn σ)α vσ(1) , . . . , vσ(k) β vσ(k+1) , . . . , vσ(k+`)
σ∈Sh(k,`)

for every v1 , . . . , vk+` ∈ V .

Solution. Let us define the equivalence relation ∼ on Sk,l by setting
σ ∼ σ̃ if
{σ(1), ..., σ(k)} = {σ̃(1), ..., σ̃(k)}
and
{σ(k + 1), ..., σ(k + l)} = {σ̃(k + 1), ..., σ̃(k + l)}.
It is clear that the number of elements in each class is k!l!. Mo-
reover, there exists a unique shuffle in each class. Thus, we can take
them as the representatives of the classes.
Next, we define h : Sk,l → R as
h(σ) = sgn(σ)α(vσ(1) , . . . , vσ(k) )β(vσ(k+1) , . . . , vσ(k+l) ).
Let us check that h(σ) = h(σ̃) if σ ∼ σ̃. That is, given σ ∼ σ̃ we
know that there exits π, τ ∈ Sk,l such that σ̃ = πτ σ with
τ (σ(i)) = σ(i) ∀i ∈ {1, ..., k}
and
π(σ(i)) = π(i) ∀i ∈ {k + 1, ..., k + l}.
6

Then,
h(σ 0 ) = sgn(σ 0 )α(vσ0 (1) , . . . , vσ0 (k) )β(vσ0 (k+1) . . . , vσ0 (k+l) )
= sgn(πτ σ)α(vπσ(1) , . . . , vπσ(k) )β(vτ σ(k+1) . . . , vτ σ(k+l) )
= sgn(σ)α(vσ(1) , . . . , vσ(k) )β(vσ(k+1) , . . . , vσ(k+l) )
= h(σ).
Here we are using the fact that α ∈ Λk (V ) and β ∈ Λ` (V ) are
alternating as well as the properties of the sign function.
Finally we obtain
α ∧ β(v1 , . . . , vk+l )
1 X
= (sgn σ)α(vσ(1) , . . . , vσ(k) )β(vσ(k+1) , . . . , vσ(k+l) )
k!l! σ∈S
k+l

1 X 1 X X
= h(σ) = h(σ)
k!l! σ∈S k!l! σ̃∈S
k+l k+l [σ]=[σ̃]
1 X X
= h(σ̃)
k!l! σ̃∈S
k+l [σ]=[σ̃]
1 X X
= h(σ̃) {σ ∈ Sk,l : [σ] = [σ̃]} = h(σ̃)
k!l! σ̃∈S σ̃∈Sk+l
k+l
X
= (sgn σ̃)α(vσ̃(1) , . . . , vσ̃(k) )β(vσ̃(k+1) , . . . , vσ̃(k+l) ).
σ∈Sk+l

∗7. Let M be a smooth manifold and let g ∈ T 2 (M ) be a Riemannian

metric tensor field; see Example 4.18 in the lecture notes. The gra-
dient of a smooth function f ∈ C ∞ (U ), U ⊂ M open, is the vector
field ∇f ∈ T (U ) defined by
gp (∇f (p), Xp ) = Xp f
for every Xp ∈ Tp M and for every p ∈ U . Show that, for any p ∈ U ,
among all unit vectors Xp ∈ Tp M , the directional derivative Xp f is
the greatest when Xp is parallel to ∇f (p) and the norm of ∇f (p) is
equal to the value of the directional derivative in that direction.
Solution. Let us take p ∈ U ⊂ M . We distinguish two cases:
• If ∇f (p) = 0, then for any Xp ∈ Tp M
Xp f = gp (∇f (p), Xp ) = 0,
and both claims follow.
• If ∇f (p) 6= 0, let us take the unitary vector parallel to ∇f (p)
∇f (p)
Yp = .
|∇f (p)|
 7

On the one hand we have

 
∇f (p)
Yp f = gp (∇f (p), Yp ) = gp ∇f (p), = |∇f (p)|.
|∇f (p)|
On the other hand, for any unitary vector Xp ∈ Tp M we obtain
by Cauchy-Schwarz inequality that
Xp f = gp (∇f (p), Xp ) ≤ |Xp ||∇f (p)| = |∇f (p)|.
Thus, both claims are proved.
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 7, Solution
14.3.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, March 14, 12:00 o’clock.

1. Let V be a finite dimensional vector space.

(a) Prove that a sequence (αj ) in T k (V ) converges to α ∈ T k (V )
(with respect to the topology induced by any norm of T k (V ))
if and only if
αj (v1 , . . . , vk ) → α(v1 , . . . , vk )
for all v1 , . . . , vk ∈ V .
(b) Prove that the wedge product is continuous in the following
sense: if αj → α in T k (V ) and βj → β in T l (V ), then αj ∧ βj →
α ∧ β in T k+l (V ).
Solution.
(a) First, note that without loss of generality we can assume that
α = 0. Let (e1 , . . . , , en ) be a basis for V and (ε1 , . . . , εn ) the dual
basis. Then, given any ν ∈ T k (V ), it can be written as
X
ν= νi1 ,...,ik εi1 ⊗ · · · ⊗ εik ,
1≤ij ≤n
P
where νi1 ,...,ik = ν(ei1 , . . . , eik ). We set kνk = 1≤ij ≤n |νi1 ,...,ik |, which
clearly defines a norm in T k (V ).
On the one hand, suppose that αj → 0, i.e., kαj k → 0. It means
that αj i1 ,...,ik = αj (ei1 , . . . , eik ) → 0. Hence, by multilinearity, given
any v1 , . . . , vk ∈ V we conclude that αj (v1 , . . . , vk ) → 0.
On the other hand, if αj (v1 , . . . , vk ) → 0 for all v1 , . . . , vk ∈ V ,
then in particular αj i1 ,...,ik = αj (ei1 , . . . , eik ) → 0. Thus, kαj k → 0.
(b) Given v1 , . . . , vk+l ∈ V ,
(αj ∧ βj )(v1 , . . . , vk+l )
X
= (sgnσ)αj (vσ(1) , . . . , vσ(k) )βj (vσ(k+1) , . . . , vσ(k+l) )
σ∈Sh(k,l)
X
→ (sgnσ)α(vσ(1) , . . . , vσ(k) )β(vσ(k+1) , . . . , vσ(k+l) )
σ∈Sh(k,l)

= (α ∧ β)(v1 , . . . , vk+l ),
by (a). This means that αj ∧ βj → α ∧ β.
2

2. Prove Lemma 4.8 in the lecture notes: If (v1 , . . . , vn ) is a basis for V

and (ω 1 , . . . , ω n ) the corresponding dual basis for V ∗ (ω i (vj ) = δji ),
then tensors
ω i1 ⊗ · · · ω ik ⊗ vj1 ⊗ · · · ⊗ vjl , 1 ≤ jp , iq ≤ n,
form a basis for Tlk (V ). In particular, dim Tlk (V ) = nk+l .
Solution. Denote
B = {ω i1 ⊗ · · · ω ik ⊗ vj1 ⊗ · · · ⊗ vjl : 1 ≤ jp , iq ≤ n}.
First we claim that B spans T`k (V ). Suppose that T ∈ T`k (V ), then
for every (k + `)-tuple (i1 , . . . , ik , j1 , . . . , j` ) we define
···j`
Tij11···i k
= T (vi1 , . . . , vik , ω j1 , . . . , ω j` ).
We claim
X ···j`
T = Tij11···i k
ω i1 ⊗ · · · ⊗ ω ik ⊗ vj1 ⊗ · · · ⊗ vj` .
1≤jp ,iq ≤n

For each (k + `)-tuples (vα1 , . . . , vαk , ω β1 , . . . , ω β` ) we have

X ···j` i1
Tij11···i k
ω ⊗ · · · ⊗ ω ik ⊗ vj1 ⊗ · · · ⊗ vj` (vα1 , . . . , vαk , ω β1 , . . . , ω β` )
1≤jp ,iq ≤n
X ···j` i1
= Tij11···i k
ω (vα1 ) · · · ω ik (vαk )vj1 (ω β1 ) · · · vj` (ω β` )
1≤jp ,iq ≤n
X ···j` i1
= Tij11···i δ · · · δαikk δjβ11 · · · δjβ``
k α1
1≤jp ,iq ≤n
···β`
= Tαβ11···α k

= T (vα1 , . . . , vαk , ω β1 , . . . , ω β` ),
so the claim follows from multilinearity since a (k, `)-tensor is de-
termined by its action on (k + `)-tuples of basis vectors.
Next, to prove that B is independent suppose that there exist
···j`
coefficients aji11···i k
such that
X ···j`
aji11···i k
ω i1 ⊗ · · · ⊗ ω ik ⊗ vj1 ⊗ · · · ⊗ vj` = 0.
1≤jp ,iq ≤n

Applying this to (any) (vα1 , . . . , vαk , ω β1 , . . . , ω β` ) gives aβα11···β

···αk = 0,
`

i.e. every coefficient is zero and B is independent.

3. Let M and N be smooth manifolds, f : M → N and h : N → R

smooth mappings, σ ∈ T k (N ), and τ ∈ T l (N ). Prove that:
(a) f ∗ dh = d(h ◦ f ).
(b) f ∗ (σ ⊗ τ ) = f ∗ σ ⊗ f ∗ τ.
Solution.
 3

(a) Given any v ∈ Tp M , by using the definition of the pull-back,

the differential and the tangent map we get
f ∗ dh(v) = dh(f∗ v) = f∗ v(h) = v(h ◦ f ) = d(h ◦ f )(v).
(b) Again, for any v1 , · · · , vk+l ∈ Tp M , we can use the definition of
the pull-back and the tensor product to show
f ∗ (σ ⊗ τ )(v1 , . . . , vk+l ) = σ ⊗ τ (f∗ v1 , . . . , f∗ vk+l )
= σ(f∗ v1 , . . . , f∗ vk )τ (f∗ vk+1 , . . . , f∗ vk+l )
= f ∗ σ(v1 , . . . , vk )f ∗ τ (vk+1 , . . . , vk+l )
= f ∗ σ ⊗ f ∗ τ (v1 , . . . , vk+l )

4. Let f : M → N be a smooth mapping and σ ∈ T k (N ). Show that

f ∗ σ ∈ T k (M ), that is, a smooth k-covariant tensor field on M .
Solution. Let us take p ∈ M and let (V, y i ) be a chart (of N ) around
f (p). We know that σ can be written in local coordinates as
X
σ= σi1 ···ik dy i1 ⊗ · · · ⊗ dy ik .
1≤iq ≤n

Then, by the previous exercise

X
f ∗σ = (σi1 ···ik ◦ f ) d(y i1 ◦ f ) ⊗ · · · ⊗ d(y ik ◦ f ).
1≤iq ≤n

Since f is smooth, the differentials p 7→ d(y ij ◦ f )p , 1 ≤ j ≤ k,

are smooth. One way to see this is to use Lemma 4.11 from the
lecture notes: if X is a vector field defined on an open set of V Then
d(y ij ◦ f )p Xp = Xp (y ij ◦ f ) is a smooth map. Hence f ∗ σ is smooth.

5. Let M = R3 . Determine which of the following differential forms

are closed and which are exact.
(a) α = yzdx + xzdy + xydz.
(b) β = xdx + x2 y 2 dy + yzdz.
(c) η = 2xy 2 dx ∧ dy + zdy ∧ dz.
Solution.
(a) We see that α = d(xyz) so it is both closed and exact.
(b) Computing
dβ = dx ∧ dx + d(x2 y 2 ) ∧ dy + d(yz) ∧ dz
= (2xy 2 dx + 2yx2 dy) ∧ dy + (zdy + ydz) ∧ dz
= 2xy 2 dx ∧ dy + zdy ∧ dz
we see that β is not closed. Therefore it cannot be exact either.
(c) From the previous computations we have η = dβ so that η is
exact and thus closed.
4

6. Let σ ∈ T k (M ) and Y1 , . . . , Yk ∈ T (M ). Prove that

LX σ(Y1 , . . . , Yk ) = LX σ (Y1 , . . . , Yk ) + σ(LX Y1 , Y2 , . . . , Yk )+
· · · + σ(Y1 , . . . , Yk−1 , LX Yk ).
Prove the claim in detail in case k = 2 and then discuss the general
case.
Solution. Let σ ∈ T k (M ) and
X, Y, Z ∈ T (M ). Note that p 7→
σp (Yp , Zp ) and p 7→ LX (Y, Z) p are C ∞ -functions. If θ is the flow
of X, then


1h ∗


i
LX σ(Y, Z) p = lim θt σ(Y, Z) p − σ(Y, Z) p
t→0 t
1h
i
σ(Y, Z) ◦ θtp − σ(Y, Z) p


= lim
t→0 t
1

= lim σθtp Yθtp , Zθtp − σp (Yp , Zp )
t→0 t
1
= lim θt∗ σ θ−t∗ Yθtp , θ−t∗ Zθtp − σp (Yp , Zp )


t→0 t
1h
= lim θt∗ σ θ−t∗ Yθtp , θ−t∗ Zθtp − σp θ−t∗ Yθtp , θ−t∗ Zθtp



t→0 t



+ σp θ−t∗ Yθtp , θ−t∗ Zθtp − σp Yp , θ−t∗ Zθtp

i
+ σp Yp , θ−t∗ Zθtp − σp (Yp , Zp )
1h
i
= lim θt∗ σ θ−t∗ Yθtp , θ−t∗ Zθtp − σp θ−t∗ Yθtp , θ−t∗ Zθtp


t→0 t
1h

i
+ lim σp θ−t∗ Yθtp , θ−t∗ Zθtp − σp Yp , θ−t∗ Zθtp
t→0 t
1h
i
+ lim σp Yp , θ−t∗ Zθtp − σp (Yp , Zp )
t→0 t




= LX σ p (Yp , Zp ) + σp (LX Y )p , Zp + σp Yp , (LX z)p .
We can obtain the general case in a similar way by adding and
subtracting terms corresponding to each variable Yi .
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 8, Solution
21.3.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, March 21, 12:00 o’clock.

1. Prove Lemma 5.13 of the lecture notes:

If f : M → N is smooth, then:
(a) f ∗ : Ak (N ) → Ak (M ) is linear.
(b) f ∗ (α ∧ β) = (f ∗ α) ∧ (f ∗ β).
(c) If (U, y), y = (y 1 , . . . , y n ), is a chart in N , then
 X 
f∗ ωi1 ···ik dy i1 ∧ · · · ∧ dy ik
i1 <···<ik
X
ωi1 ···ik ◦ f d y i1 ◦ f ∧ · · · ∧ d y ik ◦ f .




=
i1 <···<ik

Solution.
(a) First, we know from the previous lists of exercises that the
pullback f ∗ : Ak (N ) → Ak (M ). Next, given a, b ∈ R and α, β ∈
Ak (N ), it satisfies
(f ∗ (aα + bβ))p (v1 , . . . , vk ) = (aα + bβ)f (p) (f∗ v1 , . . . , f∗ vk )
= aαf (p) (f∗ v1 , . . . , f∗ vk ) + bβ(f∗ v1 , . . . , f∗ vk )
= a(f ∗ α)p (v1 , . . . , vk ) + b(f ∗ β)(v1 , . . . , vk )
= (af ∗ α + bf ∗ β)p (v1 , . . . , vk )
for all p ∈ M and v1 , . . . , vk ∈ Tp M . Therefore, f ∗ is linear.
(b) Given p ∈ M and v1 , . . . , vk+l ∈ Tp M , and using the definitions
of pullback, wedge product and tensor product, we obtain
(f ∗ (α ∧ β))p (v1 , . . . , vk+l ) = (α ∧ β)f (p) (f∗ v1 , . . . , f∗ vk+l )
X
= (sgn σ)(αf (p) ⊗ βf (p) )(f∗ vσ(1) , . . . , f∗ vσ(k+l) )
σ∈Sh(k,l)
X
= (sgn σ) αf (p) (f∗ vσ(1) , . . . , f∗ vσ(k) ) βf (p) (f∗ vσ(k+1) , . . . , f∗ vσ(k+l) )
σ∈Sh(k,l)
X
= (sgn σ) (f ∗ α)p (vσ(1) , . . . , vσ(k) ) (f ∗ β)p (vσ(k+1) , . . . , vσ(k+l) )
σ∈Sh(k,l)
X
= (sgn σ)(f ∗ α ⊗ f ∗ β)p (vσ(1) , . . . , vσ(k+l) )
σ∈Sh(k,l)
∗ ∗
= (f α ∧ f β)p (vσ(1) , . . . , vσ(k+l) ).
2

(c) First, observe that

f ∗ ωi1 ...ik dy i1 ∧ · · · ∧ dy ik = (ωi1 ...ik ◦ f )f ∗ (dy i1 ∧ · · · ∧ dy ik )

= (ωi1 ...ik ◦ f )(f ∗ dy i1 ) ∧ · · · ∧ (f ∗ dy ik )

= (ωi1 ...ik ◦ f )d(y i1 ◦ f ) ∧ · · · ∧ d(y ik ◦ f ),
where the first equality follows by Lemma 4.15(b) of the lecture
notes, the second by (b) above, and the last by Exercise 7/3 (a).
The result now follows by linearity of the pullback.

2. (a) Let ω 1 , ω 2 , . . . , ω 5 ∈ Λ1 (R5 ),

α = 2ω 1 ∧ ω 3 + ω 2 ∧ ω 3 − 3ω 3 ∧ ω 4 ∈ Λ2 (R5 ),
and
β = −ω 1 ∧ ω 2 ∧ ω 5 + 2ω 1 ∧ ω 3 ∧ ω 4 ∈ Λ3 (R5 ).
Compute α ∧ β.
(b) Let α = dx − xdy ∈ A1 (R3 ) and β = ydx ∧ dz − dy ∧ dz ∈
A2 (R3 ) Compute α ∧ β.
Solution.
(a) By Theorem 5.6 (c) of the lecture notes ω i ∧ ω j = −ω j ∧ ω i
for 1 ≤ i, j ≤ 5. In particular ω i ∧ ω i = 0. Together with the fact
that ∧ is bilinear and associative this gives
α ∧ β = 3ω 3 ∧ ω 4 ∧ ω 1 ∧ ω 2 ∧ ω 5 = 3ω 1 ∧ ω 3 ∧ ω 4 ∧ ω 2 ∧ ω 5
= 3ω 1 ∧ ω 2 ∧ ω 3 ∧ ω 4 ∧ ω 5 .
(b) Reasoning as before, ’
α ∧ β = (dx − xdy) ∧ (ydx ∧ dz − dy ∧ dz)
= −dx ∧ dy ∧ dz − xy dy ∧ dx ∧ dz
= (xy − 1) dx ∧ dy ∧ dz.

3. Let f : R3 → R2 , f (x, y, z) = (x2 , yz), α = y 2 dx+dy ∈ A1 (R2 ), and

β = xy dx ∧ dy ∈ A2 (R2 ). Compute α ∧ β, f ∗ α, f ∗ β, and f ∗ (α ∧ β).
Solution. As usual, let (x, y) 7→ (x, y) denote the standard coor-
dinates on R2 . Since α ∧ β ∈ A3 (R2 ) = {0}, α ∧ β = 0, and so
f ∗ (α ∧ β) = f ∗ 0 = 0 by linearity of the pullback. By Exercise 1 (c),
f ∗ α = f ∗ (y 2 dx) + f ∗ (dy) = (y 2 ◦ f )d(x ◦ f ) + d(y ◦ f )
= y 2 z 2 d(x2 ) + d(yz) = 2xy 2 z 2 dx + zdy + ydz.
Similarly,
f ∗ β = f ∗ (xydx ∧ dy) = (xy ◦ f )d(x ◦ f ) ∧ d(y ◦ f )
= x2 yz d(x2 ) ∧ d(yz) = 2x3 y 2 zdx ∧ dz + 2x3 yz 2 dx ∧ dy.
Note that one can also check directly that f ∗ (α) ∧ f ∗ (β) = 0.
 3

4. Let α ∈ A1 (R2 ),
α = f (x, y)dx + g(x, y)dy.
Compute dα and verify the connection between the Green’s theorem
(in vector calculus) and the Stokes’ theorem.
Solution. First, we have
dα = d(f dx) + d(gdy) = df ∧ dx + dg ∧ dy
= (∂x f dx + ∂y f dy) ∧ dx + (∂x gdx + ∂y gdy) ∧ dy
= ∂y f dy ∧ dx + ∂x g dx ∧ dy = (∂x g − ∂y f ) dx ∧ dy
Given M a smooth 2-manifold, Stokes’ theorem implies
Z Z
dα = α.
M ∂M

In the particular case α = f (x, y)dx + g(x, y)dy we get

Z Z
(∂x g − ∂y f ) dx ∧ dy = f (x, y)dx + g(x, y)dy.
M ∂M
2
Hence, if we take M = Ω ⊂ R a smooth bounded domain we recover
Green’s Theorem in vector calculus.

5. Let ω ∈ A2 (R3 ),
ω = xdy ∧ dz + ydz ∧ dx + zdz ∧ dy.
Write ω in spherical coordinates (ρ, ϕ, θ),

x = ρ sin ϕ cos θ,

y = ρ sin ϕ sin θ,

z = ρ cos θ.

Solution. First, let us compute dx, dy and dz. That is,

dx = sin ϕ cos θ dρ + ρ cos ϕ cos θ dϕ − ρ sin ϕ sin θ dθ,
dy = sin ϕ sin θ dρ + ρ cos ϕ sin θ dϕ + ρ sin ϕ cos θ dθ
and
dz = cos ϕ dρ − ρ sin ϕ dϕ.
Next, let us compute dy ∧ dz, dz ∧ dx and dx ∧ dy
dy ∧ dz = −ρ cos2 ϕ sin θ dρ ∧ dϕ + ρ cos2 ϕ sin θ dϕ ∧ dρ
+ ρ sin ϕ cos ϕ cos θ dθ ∧ dρ − ρ2 sin2 ϕ cos θ dθ ∧ dϕ
= −ρ sin θ dρ ∧ dϕ + ρ2 sin2 ϕ cos θ dϕ ∧ dθ
+ ρ sin ϕ cos ϕ cos θ dθ ∧ dρ,
4

dz ∧ dx = ρ cos2 ϕ cos θ dρ ∧ dϕ − ρ sin ϕ cos ϕ sin θ dρ ∧ dθ

− ρ sin2 ϕ cos θ dϕ ∧ dρ + ρ2 sin2 ϕ sin θ dϕ ∧ dθ
= ρ cos θ dρ ∧ dϕ + ρ2 sin2 ϕ sin θ dϕ ∧ dθ
+ ρ sin ϕ cos ϕ sin θ dθ ∧ dρ
and

dx ∧ dy = ρ cos ϕ sin ϕ cos θ sin θ dρ ∧ dϕ + ρ sin2 ϕ cos2 θ dρ ∧ dθ

+ ρ cos ϕ sin ϕ cos θ sin θ dϕ ∧ dρ + ρ2 sin ϕ cos ϕ cos2 θ dϕ ∧ dθ
− ρ sin2 ϕ sin2 θ dθ ∧ dρ − ρ2 sin ϕ cos ϕ sin2 θ dθ ∧ dϕ
= ρ2 sin ϕ cos ϕ dϕ ∧ dθ − ρ sin2 ϕ dθ ∧ dρ.
Hence,
ω = xdy ∧ dz + ydz ∧ dx + zdz ∧ dy
= −ρ2 sin ϕ sin θ cos θ + ρ2 sin ϕ sin θ cos θ dρ ∧ dϕ

+ ρ3 sin3 ϕ cos2 θ + ρ3 sin3 ϕ sin2 θ + ρ3 sin ϕ cos2 ϕ dϕ ∧ dθ

+ ρ2 sin2 ϕ cos ϕ cos2 θ + ρ2 sin2 ϕ cos ϕ sin2 θ − ρ2 sin2 ϕ cos ϕ dθ ∧ dρ

= ρ3 sin ϕ dϕ ∧ dθ.

6. Let X ∈ T (M ) be a smooth vector field and let σ and τ be smooth

covariant tensors fields. Prove that
LX (σ ⊗ τ ) = (LX σ) ⊗ τ + σ ⊗ (LX τ ).

Solution. Let p ∈ M and denote the flow of X by θ. Then,

(θ∗ (σ ⊗ τ ))p − (σ ⊗ τ )p
(LX (σ ⊗ τ ))p = lim t
t→0 t
(θt∗ σ)p ⊗ (θt∗ τ )p − σp ⊗ τp
= lim
t→0 t
((θ σ)p − σp ) ⊗ (θt∗ τ )p + σp ⊗ ((θt∗ τ )p − τp )
∗
= lim t
t→0
 ∗ t 
((θt∗ τ )p − τp )
 
(θt σ)p − σp ) ∗
= lim ⊗ (θt τ )p + lim σp ⊗
t→0 t t→0 t
=(LX σ)p ⊗ τp + σp ⊗ (LX τ )p .
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 9, Solution
28.3.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, March 28, 12:00 o’clock.

1. Let X, Y ∈ T (M ) and ω ∈ Ak (M ). Prove, by using the definition

of Lie derivative, that

LX (iY ω) = i(LX Y ) ω + iY (LX ω).

Solution. Let us use the other notation Y y ω = iY ω for the contrac-

tion. Let θ be the flow of X. By direct computation we have
1 
LX (Y y ω) = lim θt∗ (Y y ω) − Y y ω
t→0 t
1 ∗

= lim θ−t∗ Y y θt ω − Y y ω
t→0 t
1 ∗ ∗ ∗

= lim θ−t∗ Y y θt ω − Y y θt ω + Y y θt ω − Y y ω
t→0 t
1

θ−t∗ Y − Y y θt∗ ω + Y y θt∗ ω − ω


= lim
t→0 t
1 1
= lim θ−t∗ Y − Y y θt∗ ω + Y y lim θt∗ ω − ω



t→0 t t→0 t
= (LX Y )y ω + Y y (LX ω).

Another proof by using Theorem 5.28(g) of the lecture notes: We

have

LX (σ(Z1 , . . . , Zl )) = (LX σ)(Z1 , . . . , Zl )

+ σ(LX Z1 , Z2 , . . . , Zl ) + · · · + σ(Z1 , . . . , Zl−1 , LX Zl ),

where σ ∈ T l (M ) and Z1 , . . . , Zl are smooth vector fields. Since

σ(Z1 , . . . , Zl ) is a smooth real valued function, solving for LX σ gives
the useful identity

(LX σ)(Z1 , . . . , Zl )
l
X
= X(σ(Z1 , . . . , Zl )) − σ(Z1 , . . . , [X, Zi ], . . . , Zl ),
i=1

where we have used Theorem 3.61 from the lecture notes.

2

Let ω be given as in the statement. First, if k = 0 the claim is

trivial. Next, if k > 0 then iY ω ∈ Ak−1 (M ) and
LX (iY ω)(Y1 , . . . , Yk−1 ) = X((iY ω)(Y1 , . . . , Yk−1 ))
k−1
X
− (iY ω)(Y1 , . . . , [X, Yi ], . . . , Yk−1 )
i=1
k−1
X
= X(ω(Y, Y1 , . . . , Yk−1 )) − ω(Y, Y1 , . . . , [X, Yi ], . . . , Yk−1 ).
i=1

On the other hand,

iY (LX ω)(Y1 , . . . , , Yk−1 ) = LX ω(Y, Y1 , . . . , Yk−1 )
= X(ω(Y, Y1 , . . . , Yk−1 )) − ω([X, Y ], Y1 , . . . , Yk−1 )
k−1
X
− ω(Y, Y1 , . . . , [X, Yi ], . . . , Yk−1 ).
i=1

Subtracting iY (LX ω) from LX (iY ω) proves the claim.

2. Let α ∈ Ak (M ) be a closed differential k-form (that is, dα = 0) and

X ∈ T (M ) such that diX α = 0. Prove that
(θt∗ α)p = αp
for all (t, p) ∈ D(X). [We say that α is invariant under the flow of
X.]
Solution. By assumption, it follows from the Cartan’s magic formu-
la that LX α = iX dα + diX α = 0. Given p ∈ M , let Ip = {t ∈
R |(t, p) ∈ D(X)} and F : Ip → T k (Tp M ), t 7→ (θt∗ α)p , where θ is
the unique maximal flow of X. Since F (0) = αp , the claim follows if
F is constant for all t ∈ Ip :
d ∗ d d
F 0 (t) = (θs α)p |s=t = ((θs−t ◦ θt )∗ α)p |s=t = (θt∗ (θs−t ∗
α))p |s=t
ds ds ds
d d
= (θt∗ (θr∗ α))p |r=0 = (θt∗ ( (θr∗ α)|r=0 ))p = (θt∗ (LX α))p = 0.
dr dr

3. Prove that every smooth manifold M admits a Riemannian metric,

i.e. a smooth symmetric 2-covariant tensor field g that is positive
definite at each point p ∈ M .
Solution. Let us take a chart {(Uj , xj )} in M . First, let us use the
coordinates xj to construct a 2-covariant tensor field gj in each coor-
dinate domain Uj . That is, let us define
gj = δkl dxk ⊗ dxl ,
 3

which is clearly a smooth 2-tensor field on Uj . Moreover, given p ∈ Uj

and X, Y ∈ Tp M , it is satisfied
gj p (X, Y ) = δkl dxk (X) ⊗ dxl (Y ) = δkl X k Y l = gj p (Y, X),
and
gj p (X, X) = Xk X k > 0
if X is nonzero. Thus, gj is symmetric and positive definite.
Now, let us use these tensor fields in the coordinate domains to
construct the global one. By Theorem 6.7 of the lecture notes, M
admits a smooth partition of unity (hj )j∈J subordinate to {Uj }j∈J .
Then, let us define a 2-tensor field on M by
X
g= g˜j hj ,
J

where g˜j = gj in Uj and 0 otherwise. Since gj is symmetric, one can

easily deduce that g is symmetric as well. Moreover, given p ∈ M
and X ∈ Tp M nonzero, then
X
gp (X, X) = g˜j |p (X, X)hj (p) > 0
J

as each term is nonnegative and at least one hj (p) is strictly positive

and gj |p (X, X) > 0. Thus, g is a symmetric 2-tensor field on M that
is positive definite at each point p ∈ M .
It only remains to show that the tensor field g we have built is
smooth. Since g˜j hj is smooth in Uj for each j ∈ J and g˜j hj ≡ 0
in M \ spt hj , it is clear that g˜j hj is smooth in M . Next, as the
partition of unity is locally finite, given any p ∈ M there exists a
neighborhood W of p for which J 0 = {j ∈ J | W ∩ spt hj 6= ∅} is
finite. Thus,
X X
g|W = g̃j hj |W = g̃j hj |W
J J0

is smooth in W , and the claim follows.

4. Let M be an oriented smooth manifold and ω an orientation form

that determines the orientation. Let f ∈ C ∞ (M ) such that f (p) 6= 0
for all p ∈ M . Prove that ω and f ω determine the same orientation
on M if and only if f (p) > 0 for all p ∈ M .
Solution. Let us suppose that f > 0. The result follows by repeating
the proof of Theorem 7.2 of the lecture notes for the n-form f ω
noting that ω α > 0 if and only if f ω α > 0 as f was assumed strictly
positive.
Suppose now that ω and f ω determine the same orientation and
that M is connected. Hence M has exactly two orientations deter-
mined by ω and −ω. If it would have k components, it would have
4

2k orientations. Since f is continuous, then f < 0 or f > 0. Let us

assume by contradiction that f < 0. Then by the result already pro-
ved ω and −ω would determine the same orientation, which is not
possible. Thus, f > 0. For the general case suppose p ∈ M and let
U be a connected neighborhood. Now ω|U and f ω|U define the same
orientation on U . In particular by the previous f |U > 0 so f (p) > 0.

Recall that the divergence of a smooth vector field X with respect

to an orientation n-form ω is the function Divω X : M → R s.t.
(Divω X)ω = LX ω.
∗5. Let α ∈ A3 (R3 \{0}) be given in spherical coordinates (%, ϕ, ϑ), % >
0, ϕ ∈ [0, π], ϑ ∈ [0, 2π], (see Exercise 8/5) by
α = %2 sin ϕ d% ∧ dϕ ∧ dϑ.
∂
Let ∂% ∈ T (R3 \{0}) be the smooth coordinate vector field associated
to the spherical coordinates and let V = %−2 ∂% ∂
∈ T (R3 \ {0}).
Compute Divα V in R3 \ {0}.
Solution. We note that dα = 0, in fact, α = dω where
1
ω = %3 sin ϕ dϕ ∧ dθ.
3
By Cartan’s magic formula and the use of the properties of the
contraction (Theorem 5.27 in the lecture notes) we obtain
LV α = iV dα + diV α = diV α
= di%−2 ∂ (%2 sin ϕ d% ∧ dϕ ∧ dϑ)
∂ρ
 
−2 2
= d % i ∂ (% sin ϕ d% ∧ dϕ ∧ dϑ)
∂%
 
= d sin ϕ i ∂ (d% ∧ dϕ ∧ dϑ)
∂%
 h
= d sin ϕ (i ∂ d%) dϕ ∧ dϑ − (i ∂ dϕ) d% ∧ dϑ
∂% ∂%
i
+ (i ∂ dϑ) d% ∧ dϕ
∂%

= d (sin ϕ dϕ ∧ dϑ)
= cos ϕ dϕ ∧ dϕ ∧ dϑ
=0
in R3 \ {0}. Since (Divα V )α = LV α = 0 we conclude that
Divα V = 0 in R3 \ {0}.
Let us note that α is the standard volume form of R3 in spherical
coordinates and V = ∇g with g(x) = −|x|−1 . Then, since α is the
 5

volume form and g is harmonic in R3 \ {0} one obtains

Divα V = div∇g = ∆g = 0 in R3 \ {0}.

∗6. Using the spherical coordinates as above, let ω = 31 %3 sin ϕ dϕ ∧

dϑ ∈ T (R3 \ {0}). Apply Stokes’s theorem with ω and dω in the
closed ball M = B̄ 3 (0, R) ⊂ R3 .
Solution. On the one hand we have
R3 2π π
Z Z Z Z
1 3
ω= % sin ϕ dϕ ∧ dϑ = sin ϕ dϕ dϑ
∂M ∂B(0,R) 3 3 0 0
4
= πR3 .
3
On the other hand, since dω = %2 sin ϕ d% ∧ dϕ ∧ dϑ, we obtain
Z Z
dω = %2 sin ϕ d% ∧ dϕ ∧ dϑ
M B(0,R)
Z R Z 2π Z π
4
= %2 sin ϕ dϕ dϑ d% = πR3 .
3
R R 0 0 0
Hence, M dω = ∂M ω as it should be. Let us point out that since
the differential forms involved are defined just in R3 \ {0}, in order
to be precise we should integrate over the annulus B(0, R) \ B(0, ε)
and let ε → 0.
We also note that 43 πR3 is the measure (volume) of the ball
B(0, R) since dω is the volume form in R3 in spherical coordina-
tes.
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 10
4.4.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, April 4, 12:00 o’clock.

1. Let G be a Lie group. A differential form α is called left invariant

if α = L∗p α for every p ∈ G. More precisely, (L∗p α)q = αq for every
p, q ∈ G. Prove that there exists an orientation form on G and that
a left invariant orientation form is unique up to a multiplicative
constant. [Here Lp : G → G is the left translation Lp (q) = pq.]
Solution. Let be n = dim G and e the neutral element of G. Choose
ω1 , . . . , ωn ∈ T 1 (Te G) s.t. ωe := ω1 ∧ ω2 ∧ · · · ∧ ωn 6= 0 (i.e. spans
Λn (Te G) from being a one-dimensional linear space). Then, we define
ω ∈ Λn G as
ωp = (L−1 ∗
p ) ωe , p ∈ G.
We will show that ω is a left invariant smooth orientation form.
First, given p, q ∈ G we have
(L∗p ω)q (v1 , . . . , vn ) = ωpq Lp∗q v1 , . . . , Lp∗q vn

= (L−1 ∗


pq ) ωe Lp∗q v1 , . . . , Lp∗q vn

= ωe (L−1 −1


pq )∗ L p∗q v 1 , . . . , (L pq )∗ Lp∗q vn
−1 −1


= ωe (Lq )∗q v1 , . . . , (Lq )∗q vn
= ωq (v1 , . . . , vn )
for every v1 , . . . , vn ∈ Tq G. Hence ω is left invariant. Moreover, since
ωe 6= 0, there are v1 , . . . , vn ∈ Te G (e.g. duals to ω1 , . . . , ωn ) such
that ωe (v1 , . . . , vn ) 6= 0. Then, by left invariance
ωp Lp∗ v1 , . . . , Lp∗ vn = (L−1 ∗



p ) ωe Lp∗ v1 , . . . , L p∗ v n

= ωe (v1 , . . . , vn ) 6= 0,
i.e. ωp 6= 0.
Next, we show that ω is smooth. To see this, we note that
ωp = (L−1 ∗ −1 ∗ −1 ∗
p ) ωe = (Lp ) ω1 ∧ · · · ∧ (Lp ) ωn .

Now each p 7→ (L−1 ∗

p ) ωi , i = 1, . . . , n, is a smooth 1-form. This can
be seen by taking an arbitrary smooth vector field X ∈ T (G) and

noticing that the function p 7→ (L−1 ∗ −1
p ) ωi (Xp ) = (ωi )e (Lp )∗ Xp
is smooth as a composition of the linear functional (ωi )e and the
smooth map p 7→ (L−1 n
p )∗ Xp ∈ Te G. Hence, ω ∈ A (G), and in
particular is an orientation form.
2

Finally, if ω̃ is another left invariant orientation from, then it is

clear that ω̃e = cωe for some constant c 6= 0. From this relation and
the left invariance (of both ω̃ and ω), we conclude that ω̃p = cωp for
every p ∈ G.

2. Let M and N be connected smooth oriented manifolds and f :

M → N a local diffeomorphism. Prove that f is either orientation
preserving or orientation reversing.
Solution. Let ω be an orientation form on N associated to the orien-
tation of N . Then f ∗ ω is a smooth n-form on M . Let η be an
orientation form on M associated to the orientation of M . Now
f ∗ ω = gη, where g ∈ C ∞ (M ). Since f is a local diffeomorphism, we
will show that the function g never vanishes. Let us take {(Uα , xα )}
and {(Vβ , yβ )} orientations in M and N respectively. Given p ∈ M
we take i ∈ I and j ∈ J such that p ∈ Uαi and f (p) ∈ Vβj . Moreo-
ver, note that we can always assume that f (Uαi ) = Vβj . Then, we
can write ω and η in coordinates as
ω|Vβj = ωj dyβ1j ∧ dyβ2j ... ∧ dyβnj
and
η|Uαi = ηi dx1αi ∧ dx2αi ... ∧ dxnαi
with ωj and ηi positive smooth functions. By using the relation
f ∗ ω = gη we obtain that

0
gηi = (ωj ◦ f ) det y ◦ f ◦ x−1 .

0
Since f is a local diffeomorphism we have det y ◦ f ◦ x−1 never
vanishes. Thus from the previous identity we conclude that g never
vanishes. Finally, from being M connected, g is either positive eve-
rywhere (and f sense preserving) or g is negative everywhere (and
f sense reversing).

3. Let M be an oriented n-manifold with boundary and an orienta-

tion form ω, and let V be an outward-pointing smooth vector field
on ∂M . Prove that the induced orientation on ∂M , determined by
(iV ω)|∂M , is independent of the choice of the orientation form ω
and the outward-pointing vector field V .
Solution. Let ω 1 , ω 2 ∈ An (M ) be two orientation forms on M and
V1 , V2 : ∂M → T M be two smooth outward-pointing vector fields.
We will show that iV1 ω 1 and iV2 ω 2 determine the same orientation
restricted to ∂M . To this end, let p ∈ ∂M and (U, x) a chart at p
where x = (x1 , . . . , xn ) : U →
PH̄
n
. Then, we can write
P the vector
fields in coordinates as V1 p = i=1 V1i (∂i )p and V2 p = ni=1 V2i (∂i )p .
n

By Lemma 7.7 in the lecture notes we know that V1n , V2n < 0. Hence
there exists a constant c > 0 such that V1n = cV2n . On the other
 3

hand, by Exercise 9/4 we know that ω 1 = f ω 2 for some f > 0 as

can be seen writing out the orientation forms locally. Hence,
(iV1 ω 1 )p ((∂1 )p , . . . , (∂n−1 )p ) = (iV1 (f ω 2 ))p ((∂1 )p , . . . , (∂n−1 )p )
= f (p)ωp2 (V1 , (∂1 )p , . . . , (∂n−1 )p )
= f (p)ωp2 (V1n (∂n )p , (∂1 )p , . . . , (∂n−1 )p )
= cf (p)ωp2 (V2n (∂n )p , (∂1 )p , . . . , (∂n−1 )p )
= cf (p)ωp2 (V2 , (∂1 )p , . . . , (∂n−1 )p )
= cf (p)(iV2 ω 2 )p ((∂1 )p , . . . , (∂n−1 )p ).
As cf (p) > 0 for all p ∈ ∂M the result follows by Exercise 9/4.

∗4. Let M be a smooth manifold, α ∈ Ak (M ), X, Y ∈ T (M ) and

f ∈ C ∞ (M ). Prove that
(a)
Lf X α = f LX α + df ∧ iX α,
(b)
i[X,Y ] α = LX iY α − iY LX α.

Solution.
(a) Given Z1 , ..., Zk−1 , if we use Exercise 7/6 and the properties of
the Lie bracket, on the one hand we obtain
Lf X α(Z1 , ..., Zk )
k
X
= Lf X (α(Z1 , ..., Zk )) − α(Z1 , ..., [f X, Zi ], ..., Zk )
i=1
k
X
= f LX (α(Z1 , ..., Zk )) − f α(Z1 , ..., [X, Zi ], ..., Zk )
i=1
k
X
+ Zi (f ) α(Z1 , ..., X, ..., Zk )
i=1
k
X
= f LX α(Z1 , ..., Zk ) + Zi (f ) α(Z1 , ..., X, ..., Zk )
i=1

On the other hand,

df ∧iX α(Z1 , ..., Zk )
k
X
= (−1)i−1 df (Zi ) iX α(Z1 , ..., Zi−1 , Zi+1 , ..., Zk )
i=1
k
X
= Zi (f ) α(Z1 , ..., X, ..., Zk )
i=1
4

Hence, the identity follows.

(b) Given Z1 , ..., Zk , if we use again Exercise 7/6 and the properties
of the Lie bracket, on the one hand we obtain

LX iY α(Z1 , ..., Zk−1 )

k−1
X
= LX (iY α(Z1 , ..., Zk−1 )) − iY α(Z1 , ..., [X, Zi ], ..., Zk−1 )
i=1
k−1
X
= X (α(Y, Z1 , ..., Zk−1 )) − α(Y, Z1 , ..., [X, Zi ], ..., Zk−1 )
i=1

On the other hand,

iY LX α(Z1 , ..., Zk−1 ) = LX α(Y, Z1 , ..., Zk−1 )

= LX (α(Y, Z1 , ..., Zk−1 )) − α([X, Y ], Z1 , ..., Zk−1 )
k−1
X
− α(Y, Z1 , ..., [X, Zi ], ..., Zk−1 )
i=1
= X(α(Y, Z1 , ..., Zk−1 )) − i[X,Y ] α(Z1 , ..., Zk−1 )
k−1
X
− α(Y, Z1 , ..., [X, Zi ], ..., Zk−1 )
i=1

Hence, the identity follows by subtracting both expressions.

Recall that the divergence of a smooth vector field X with respect

to an orientation n-form ω is the function Divω X : M → R s.t.
(Divω X)ω = LX ω.
∗5. Let M be a smooth oriented n-manifold and let ω be an orientation
form (volume form) on M . Suppose that X ∈ T (M ) and f, g ∈
C ∞ (M ), with f > 0. Prove that

Divω (gX) = gDivω X + Xg

and
Xf
Divf ω X = Divω X + .
f

Solution. By definition,

LgX ω = Divω (gX)ω.

 5

We compute the left hand side by using Cartan’s magic formula:

LgX ω = igX dω + digX ω = digX ω


= d giX ω = dg ∧ iX ω + gdiX ω
(∗)
= iX dg ∧ ω + gdiX ω
= (Xg)ω + gLX ω

= Xg + dDivω (X) ω.
Hence
Divω (gX) = Xg + gDivω (X)
= LX gdigX ω.
Equality (∗) follows from


0 = iX dg ∧ ω = iX dg ∧ ω − dg ∧ iX ω.
| {z }
≡0
For the other equation, we compute


Divf ω X f ω = LX (f ω) = f LX ω + (Xf )ω

Xf
= f Divω X ω + (f ω)
f
Xf

= Divω X + (f ω).
f
Department of Mathematics and Statistics
Introduction to differential geometry
Exercise 11, Solution
11.4.2022

Return your written solutions to the bonus problems (marked by ∗) to

the Moodle area by Monday, April 11, 12:00 o’clock.

These are the last exercises!

1. Let M be a smooth oriented n-manifold and let ω be an orientation

form (volume form) on M . Suppose that X, Y ∈ T (M ). Prove that



Divω [X, Y ] = X Divω Y − Y Divω X .

Solution. By using Exercise 2 below and Theorem 5.28 (c)(a) from

the lecture notes we obtain


Divω [X, Y ] ω = L[X,Y ] ω
= LX LY ω − LY LX ω



= LX (Divω Y )ω − LY (Divω X)ω



= X Divω Y ω + Divω Y LX ω



− Y Divω X ω − Divω X LY ω




= X Divω Y ω + Divω Y Divω X ω




− Y Divω X ω − Divω X Divω Y ω


= X(Divω Y ) − Y (Divω X) ω.

∗2. Let M be a smooth manifold, α ∈ Ak (M ) and X, Y ∈ T (M ).

Prove that
L[X,Y ] α = LX LY α − LY LX α.

Solution. By Cartan’s magic formula, Exercise 10/4(b) and Theorem

5.28 (b) from the lecture notes we get
L[X,Y ] α = di[X,Y ] α + i[X,Y ] dα
= d (LX iY α − iY LX α) + LX iY dα − iY LX dα
= LX diY α − diY LX α + LX iY dα − iY LX dα
= LX (diY α + iY dα) − (diY LX α + iY dLX α)
= LX LY α − LY LX α.
2

3. Suppose M and N are oriented n-manifolds and f : M → N an

orientation preserving diffeomorphism. Prove: If ω ∈ An (N ) has
compact support, then f ∗ ω ∈ An (M ) has compact support and
Z Z
ω= f ∗ ω.
N M

Solution. Let us take {(Vα , yα )} an orientation atlas of N . Since f

is a sense preserving diffeomorphism, then {(Uα := f −1 Vα , xα :=
yα ◦ f )} is an orientation atlas of M . Furthermore, let {λi } be a
smooth partition of unity subordinate to {Vα }. This clearly means
that {λi ◦ f } is a smooth partition of unity subordinate to {Uα }.
Now,
Z Z !
X XZ XZ
ω= λi ω = λi ω = λi ω
N N i i N i V αi
XZ XZ
= (yα−1i )∗ (λi ω) = (f ◦ x−1 ∗
αi ) (λi ω)
i Rn i Rn
XZ XZ
= (x−1
αi )
∗ ∗
(f (λi ω)) = f ∗ (λi ω)
i Rn i U αi
XZ XZ
= (λi ◦ f )f ∗ ω = (λi ◦ f )f ∗ ω
Uα M
Zi X i Z i
= (λi ◦ f )f ∗ ω = f ∗ ω.
M i M

Note that the smoothness of f ω is clear, so f ∗ ω ∈ An (M ). Since

∗

f is a diffeomorphism and ω is compactly supported, also f ∗ ω is

compactly supported.

Recall from Real Analysis: Let M be an oriented n-manifold and

ω ∈ An (M ) an orientation (volume) form. By Riesz representation
theorem, there exists a unique measure mω (defined in the σ-algebra
of all Borel subsets of M ) such that
Z Z
u dmω = uω
M M
for every compactly supported continuous function u.

Let M and N be oriented n-manifolds with orientation forms ωM

and ωN , respectively. Let f : M → N be a diffeomorphism. We define
the Jacobian of f , denoted by Jf , with respect to ωM and ωN by the
formula
f ∗ ωN = Jf ωM .
 3

∗4. Let M be an oriented n-manifold with an orientation form ω ∈

An (M ) and the corresponding Borel measure mω . Let X ∈ T (M )
be a complete vector field and θ its flow. Prove that the following
are equivalent:
(i) Divω X ≡ 0,
(ii) Jθt ≡ 1 ∀t ∈ R,
(iii) θt∗ ω = ω,
(iv) Z Z
u dmω = (u ◦ θt ) dmω
M M
for all compactly supported continuous functions u on M and
for all t ∈ R.
Solution. First, note that by definition of divergence and Jacobian,
and Exercise 9/2 we obtain directly the equivalence of (i), (ii) and
(iii). Furthermore, these are equivalent to (iv) since by Exercise 3
we haveZ Z Z Z
udmω = uω = θt∗ (uω) = (u ◦ θt )θt∗ ω
M
ZM M
Z M

= (u ◦ θt )ω = (u ◦ θt )dmω .
M M
Let us point out that this is a far reaching generalization of the
translation invariance of the Lebesgue measure in Rn . Note that a
translation x 7→ x + a, where a ∈ Rn , is the flow of the parallel
vector field Xp = a for all p ∈ Rn .