GEOMETRY MID-TERM

CLAY SHONKWILER

1. Polar Coordinates
R2 can be considered as a regular surface S by setting S = {(x, y, z) ∈
R3 : z = 0}. Polar coordinates on S are given as follows. Let U =
{(ρ, θ) : ρ > 0, 0 < θ < 2π} and let φ : U → V ⊂ S be given by
φ(ρ, θ) = (ρ cos θ, ρ sin θ, 0), where V = S\{(x, 0, 0) ∈ R3 : x ≥ 0}.
(a) Compute the first fundamental form of S in polar coordinates.
Answer: First, we need to compute Φ1 = ∂φ ∂φ
∂ρ and Φ2 = ∂θ :

Φ1 = (cos θ, sin θ, 0)
Φ2 = (−ρ sin θ, ρ cos θ, 0).
Now, the coefficients of the first fundamental form are as follows:
E = hΦ1 , Φ1 i = cos2 θ + sin2 θ = 1
F = hΦ1 , Φ2 i = −ρ sin θ cos θ + ρ sin θ cos θ = 0
G = hΦ2 , Φ2 i = ρ2 sin2 θ + ρ2 cos2 θ = ρ2 .
♣
(b) Compute the Christoffel symbols of S in polar coordinates. How
does this compare with the computation in the usual coordinate system
ψ(x, y) = (x, y, 0)?
Answer: In order to compute the Christoffel symbols, first we need to
compute the partials of E, F and G:
Eρ = 0 Eθ = 0
Fρ = 0 Fθ = 0
Gρ = 2ρ Gθ = 0
Now, plugging these values into the system of equations that lets us deter-
mine the Christoffel symbols, we see
0 = 12 Eρ = hΦ11 , Φ1 i = Γ111 E + Γ211 F = Γ111
0 = Fρ − 12 Eθ = hΦ11 , Φ2 i = Γ111 F + Γ211 G = Γ211 ρ2
0 = 21 Eθ = hΦ12 , Φ1 i = Γ112 E + Γ212 F = Γ112
ρ = 21 Gρ = hΦ12 , Φ2 i = Γ112 F + Γ212 G = Γ212 ρ2
−ρ = Fθ − 21 Gρ = hΦ22 , Φ1 i = Γ122 E + Γ222 F = Γ122
0 = Gθ = hΦ22 , Φ2 i = Γ122 F + Γ222 G = Γ222 ρ2 .
From this, then, we conclude that
Γ111 = Γ211 = Γ112 = Γ222 = 0
1
2 CLAY SHONKWILER

and
1
Γ212 =, Γ122 = −ρ.
ρ
These differ somewhat from the Christoffel symbols in the usual coordinate
system, which are all zero, due to the fact that the partials of E, F and G
in the usual coordinate system are all zero.
♣

2. Maximal Integral Curves

On a previous HW we showed that on a compact manifold the maximal
integral curves of a smooth vector field are defined for all time t. Show that
on the open disk D = {(x, y) ∈ R2 : x2 + y 2 = 1} there exists a smooth
vector field X for which the maximal integral curves are not defined for all
time t. Hence, the compactness condition on our HW problem is necessary.
Example: Let Xq = q for all q ∈ D. Specifically, the vector in Tp D
associated with the point p = (x, y) is simply (x, y). Now, let p = (x0 , y0 ) ∈
D and let α : I → D be an integral curve through p of X. Then α(t) =
(x(t), y(t)) for t ∈ I,
α(0) = (x0 , y0 )
and
(x0 (t), y 0 (t)) = α0 (t) = Xα(t) = (x(t), y(t)) = α(t).
Hence, we see that α(t) = (x0 et , y0 et ). Since the integral curves are unique,
we see that curves of this form are the only integral curves on D; specifically,
any maximal integral curve must be of this form. Now, if p = 0, a maximal
integral curve can be defined for all time t. However, this curve is constant,
forever remaining fixed at the origin. On the other hand, if p 6= 0, then we
see that, for all t > t0 for some t0 ∈ R,
|(x0 et , yo et )| > 1,
meaning α is not defined for these values of t. Since, as we argued above,
all integral curves, including any maximal integral curves, must be of this
form, we see that the maximal integral curves of X on D are not defined for
all time t.
♣

3. Covering Maps
Let S̃ and S be two regular surfaces. A map F : S̃ → S is said to be a
smooth covering map if
(a) F is smooth.
(b) For each p ∈ S there exists a neighborhood O of p such that F −1 (O) =
∪α∈J Vα is the disjoint union of open sets Vα such that for each α ∈ J we
have F : Vα → O is a diffeomorphism. Such a neighborhood O is said to be
evenly covered by F .
 GEOMETRY MID-TERM 3

Now suppose that F1 : S1 → S and F2 : S2 → S are smooth covering

maps and g : S1 → S2 is a homeomorphism such that F2 ◦ g = F1 and
F1 ◦ g −1 = F2 . Show that g : S1 → S2 is a diffeomorphism.
Proof. Let p ∈ S1 and let q = F1 (p) = (F2 ◦g)(p). Let O1 be a neighborhood
of q that is evenly covered by F1 and let O2 be a neighborhood of q that is
evenly covered by F2 . Then
F1−1 (O1 ) = ∪α∈J V1α
and
F2−1 (O2 ) = ∪α∈J V2α
where F1 : V1α → O1 and F2 : V2α → O2 are diffeomorphisms. Let V1β be
the V1α containing p and let V2β be the V2α containing g(p). Let
O = O1 ∩ O2 , V1 = V1β ∩ F1−1 (O), V2 = V2β ∩ F2−1 (O).
Then F1 : V1 → O and F2 : V2 → O are diffeomorphisms. Let
φ1 : U1 ⊂ R2 → W1 ⊂ V1
φ2 : U2 ⊂ R2 → W2 ⊂ V2
and
φ : U ⊂ R2 → W ⊂ O
be coordinate charts on p, g(p) and q, respectively, such that F1 (φ1 (U1 )) ⊆
φ(U ) and F2 (φ2 (U2 )) ⊆ φ(U ). Then
H1 := φ−1 ◦ F1 ◦ φ1 : U1 → U
and
H2 := φ−1 ◦ F2 ◦ φ2 : U2 → U
are differentiable. Furthermore, if
Ũ := H1 (U1 ) ∩ H2 (U2 ),
Ũ1 := H1−1 (Ũ )
and
Ũ2 := H2−1 (Ũ )
then H1 : Ũ1 → Ũ and H2 : Ũ2 → Ũ are diffeomorphisms. Hence, H1−1 ◦ H2
is differentiable, meaning that, when the domains of the F ’s and φ’s are
restricted appropriately,
H1−1 ◦ H2 = (φ−1 ◦ F1 ◦ φ1 )−1 (φ−1 ◦ F2 ◦ φ2 )
= φ−1 −1
1 ◦ F1 ◦ φ ◦ φ
−1 ◦ F ◦ φ
2 2
−1 −1
= φ1 ◦ F1 ◦ F2 ◦ φ2
= φ−1 −1
1 ◦ F1 ◦ F1 ◦ g
−1 ◦ φ
2
= φ−1
1 ◦ g −1 ◦ φ
2

is differentiable. This says precisely that g −1 is differentiable on the appro-

priate neighborhood of g(p).
4 CLAY SHONKWILER

A similar argument shows that g is differentiable on the corresponding

neighborhood of p. Since our choice of p was arbitary, we see that g is a
differentiable homeomorphism with differentiable inverse at all points in S1 :
that is to say, g is a diffeomorphism. 

4. Non-Orientable Surfaces
In do Carmo (and in class) Gaussian Curvature was developed only for
orientable surfaces. Is this necessary? Please discuss.
Answer: No, it is not necessary for a surface to be orientable in order
for us to determine the Gaussian Curvature at a point p on the surface.
To see why this is the case, we first note that every surface covered by a
single coordinate system is trivially orientable. Hence, any surface, even a
non-orientable surface like the Mobius strip is locally orientable. Therefore,
if we take our point p and consider some coordinate chart φ : U ⊂ R2 → V
on a neighborhood of p, then we see that we can assign an orientation to
this neighborhood of V .
Given this orientation N , then, we can certainly define the Gauss map
N : V → R3 on this neighborhood V of p. In fact, since V is itself a
regular surface, we can define the Gaussian curvature K, which is just the
determinant of dNp . Now, this determinant is not dependent on the choice
of orientation for V (remember, V is an orientable surface), so this notion
of Gaussian curvature at p is well-defined.
♣

5. Regular n-manifolds
Let M2 (R) be the set of all 2 × 2 matrices and SL(2, R) = {A ∈ M2 (R) :
det(A) = 1} the set of all real 2 × 2 matrices with determinant
  1.
a b
(a) M2 (R) can be identified with R4 via the map 7→ (a, b, c, d).
c d
Use this identification to show that SL(2, R) is a regular 3-surface.
 
a b
Proof. Let ∈ SL(2, R), which is identified with (a, b, c, d) ∈ R4 .
c d
Since the zero matrix is not in SL(2, R) we may assume, up to a change
of axes, that a 6= 0. Then let V = {(x, y, z, w) ∈ R4 : x 6= 0} and let
U = {(x, y, z) ∈ R3 : x 6= 0. Then V is certainly an open set of R4 and U is
an open set of R3 . Now, let φ : U → V be given by
 
1 + yz
(x, y, z) 7→ x, y, z, .
x
We see that
 
x y 1 + yz
det 1+yz =x − yz = 1 + yz − yz = 1,
z x x
 GEOMETRY MID-TERM 5

so the map φ is well-defined. Since x 6= 0, we see that φ is smooth since

each coordinate function is. Also,
 
1 0 0
 0 1 0 
dφ =  0 0

1 
−x(1+yz) z y
x2 x x

The top minor of dφ has determinant 1 for all points in U , so dφ is injective.

Hence, it only remains to show that φ is a homeomorphism. Certainly, since
φ is smooth, φ is continuous, so we need only show that φ−1 exists and is
continuous. If A = (a, b, c, d) ∈ SL( 2, R) (again, remember we may assume
a 6= 0), then
1 + bc
1 = det A = ad − bc ⇒ d =
a
so φ−1 (a, b, c, d) = (a, b, c). Certainly this is a smooth function and, therefore
continuous, so we see that φ is, indeed, a coordinate chart on V ∩ SL(2, R).
Since our choice of (a, b, c, d) was arbitrary (up to change of axes), we con-
clude that SL(2, R) is indeed a regular 3-surface. 
 
1 0
(b) Show that the tangent space of SL(2, R) at I = is given by
0 1
 
a b
TI (SL(2, R)) = {(I; ) : a + d = 0}.
c d
That is, TI (SL(2, R)) can be identified with the set of traceless 2×2 matrices
over R. This tangent space is an example of a Lie algebra.

Proof. By the theorem given in the statement of the problem, we know that
TI (SL(2, R)) = dφI (R3 ).
In part (a) above we gave the general form for dφ, we need only plug in the
coordinates of
φ−1 (1, 0, 0, 1) = (1, 0, 0)
to that formula. Specifically, we see that, if (x1 , x2 , x3 ) ∈ R3 ,
   
  1 0 0   x1
x1  0 1 0  x1  x2 
dφI  x2  =   0 0 1  x2
 =
 x3
.

x3 −(1+0) 0 0 x 3
1 1 1
−x1

Now, we identify this vector with the matrix

 
x1 x2
.
x3 −x1
6 CLAY SHONKWILER

We note that x1 + (−x1 ) = 0; when we vary (x1 , x2 , x3 ) over all of R3 , we

see that, indeed,
 
a b
TI (SL(2, R)) = dφI (R3 ) = {(I; ) : a + d = 0}.
c d


6. Vector Fields as Differential Operators

Let Xp ∈ Tp S be a tangent vector and f : U ⊆ S → R be a smooth
function defined on an open neighborhood of p. Then
d
Xp f ≡ (f ◦ α)|t=0
dt
where α : (−, ) → S is a smooth curve such that α(0) = p and α0 (0) = Xp .
(a) Let X be a vector field on an open set U of S. Show that X is a
smooth vector field if and only if for any function f : U → R the function
q 7→ Xq f is smooth.
Proof. Suppose X is a smooth vector field. This means that, if φ : W ⊆
R2 → V subseteqU is a coordinate chart on some point p ∈ U ,
Xq = a(q)Φ1q + b(q)Φ2q
where a, b : V → R are smooth maps. Let alpha : (−, ) → S be a smooth
curve such that α(0) = q and α0 (0) = Xq . If f : U → R is any smooth map,
then, by the chain rule,
d
Xq f = (f ◦ α)|t=0 = dfα(0) α0 (0) = dfα(0) (a(q)Φ1q + b(q)φ2q ),
dt
which is certainly a smooth function, since df, a, b, Φ1 and Φ2 are all smooth.
On the other hand, suppose X is a vector field and q 7→ Xq f is smooth.
Then
Xq = a(q)Φ1q + b(q)Φ2q
where φ : W ⊆ R2 → V ⊆ U is a coordinate chart on a neighborhood of
p ∈ U . We want to show that a, b : V → R are smooth. Let α : (−, ) → U
be a smooth curve such that α(0) = q and α0 (0) = Xq . Then,
d
Xq f = (f ◦α)|t=0 = dfα(0) α0 (0) = dfα(0) (a(q)Φ1q +b(q)Φ2q = dfq a(q)Φ1q +dfq b(q)Φ2q
dt
since df is a linear map. Since q 7→ Xq f is smooth, it follows that a and b
are smooth, meaning that X is smooth. 
(b) Let λ ∈ R and f, g : U → R be smooth functions. Show that for any
p ∈ S we have
Xp (λf + µg) = λXp f + µXp g
and
Xp (f g) = Xp (f )g(p) + f (p)Xp (g).
 GEOMETRY MID-TERM 7

Proof. Let α : (−, ) → S be a smooth curve such that α(0) = p and

α0 (0) = Xp . Since λf + µg is smooth, we can use the chain rule to see that
d
Xp (λf + µg) = dt ((λf + µg) ◦ α)|t=0
= d(λf + µg)α(0) α0 (0)
= (d(λf )α(0) + d(µg)α(0) )α0 (0)
= (λdfα(0) + µdgα(0) )α0 (0)
= λdfα(0) α0 (0) + µdgα(0) α0 (0)
d d
= λ dt (f ◦ α)|t=0 + µ dt (g ◦ α)|t=0
= λXp f + µXp g.
On the other hand, using the product rule for differentials and the chain
rule,
d
Xp (f g) = dt (f g ◦ α)|t=0
= d(f g)α(0) α0 (0)
= (dfα(0) g(α(0)) + f (α(0))dgα(0) )α0 (0)
= dfα(0) g(p)α0 (0) + f (p)dgα(0) α0 (0)
= dfα(0) α0 (0)g(p) + f (p)dgα(0) α0 (0)
d d
= dt (f ◦ α)|t=0 g(p) + f (p) dt (g ◦ α)|t=0
= Xp (f )g(p) + f (p)Xp (g).


(c) Show that for any smooth function f : W ⊆ S → R on a neighborhood

of p ∈ S and any tangent vector Xp ∈ Tp S we have
dfp (Xp ) = Xp f.

Proof. Let α be a smooth curve such that α(0) = p and α0 (0) = Xp . Now,
since f is smooth we may use the chain to see that
d
Xp f = (f ◦ α)|t=0 = dfα(0) · α0 (0) = dfp Xp .
dt
Hence, we conclude that
dfp (Xp ) = Xp f.


(d) Let X, Y, Z be smooth vector fields on S. Then the function q 7→

hYq , Zq i is a smooth function on S. Show that for any p ∈ S we have
Xp hY, Zi = h∇X Y (p), Zp i + hYp , ∇X Z(p)i.

Proof. Again, let α be a smooth curve such that α(0) = p and α0 (0) = Xp .
Using the result proved in class that, if X1 , X2 are vector fields along α,
then
d DX2 DX2
hX1 (t), X2 (t)i = h , X2 i + hX, i,
dt dt dt
8 CLAY SHONKWILER

we see that
d
Xp hY, Zi = dt (hY, Zi ◦ α)|t=0
d
= dt (hY (t), Z(t)i)|t=0
= h DY DZ
dt , Zp i + hYp , dt i
= h∇X Y (p), Zp i + hYp , ∇X Z(p)i.


7. Vector Fields & Coordinate Charts

Let X be a smooth vector field on S and suppose p ∈ S is such that
Xp 6= 0. Show that there exists a coordinate chart φ : U → V containing p
such that Φ1q = Xq for all q ∈ V .

Proof. We show this in two parts. First we prove the result on an open set
contained in R2 , then extend that result to any regular surface.
Lemma 7.1. Let X be a vector field on an open set W ⊂ R2 and let p ∈ U
such that Xp 6= 0. Then there exists a coordinate chart φ : U → V ⊆ W on
a neighborhood of p such that Φ1q = Xq for all q ∈ V .
Proof. Choose a coordinate system on R2 such that p = (0, 0) and Xp is
in the direction of the x axis. Let α : Ṽ × I → W be the local flow at p,
Ṽ ⊂ W , t ∈ I and let φ̃ be the restriction of α to the rectangle
(Ṽ × I) ∩ {(x, y, t) ∈ R3 ; x = 0}.
We can identify {(x, y, t) ∈ R3 : x = 0} with R2 in such a way that the t
direction is identified with the u1 direction and the y direction is identified
with the u2 direction. The above rectangle, then, is a subset of the plane and
we see that φ̃ is a map from a subset of R2 into W . By the definition of local
flow, dφ̃p maps the unit vector in the u1 direction into X and and maps the
unit vector of the u2 axis into itself. Therefore, dφ̃p is nonsingular. By the
inverse function theorem, then, it follows that there exists a neighborhood
V ⊂ W of p where φ̃−1 is defined and differentiable. Now, let U = φ̃−1 (V )
and let φ be the restriction of φ̃ to U . Since φ is a differentiable function
with differentiable inverse, bijective and with nonsingular differential, we
see that φ : U → V is a coordinate chart on a neighborhood of p. It only
remains to show that Φ1q = Xq for all q ∈ V .
However, this follows immediately from how we have defined φ. Since the
u1 direction in U is identified with the t direction in the original rectangle
lying in the yt-plane,
∂φ ∂α
(u1 , u2 ) = (q, t) = Xα(q,t) .
∂u1 ∂t
Hence, if q ∈ V and (u0 , v0 ) ∈ U such that φ(u0 , v0 ) = q (that is to say, u0
and v0 are identified with t0 ∈ I and q0 is the unique point on the y-axis
 GEOMETRY MID-TERM 9

such that the trajectory through q0 passes through q), then

∂φ ∂α
φ1q = (u0 , v0 ) = (q0 , t0 ) = Xα(q0 ,t0 ) = Xq .
∂u1 ∂t

Now, we turn to the general case.
Let S, X and p be as stated in the problem and let ψ : U 0 → V 0 ⊂ S be a
coordinate chart on a neighborhood of p. Since X is smooth, we know that
in the coordinate system,
Xq = a(q)Ψ1q + b(q)Ψ2q
where a, b : V 0 → R are smooth. Consider the vector field X̃ = ãe1 + b̃e2
where
ã = a ◦ ψ, b̃ = b ◦ ψ : U 0 → R.
Then, as in the proof of the local first integral lemma for surfaces that we did
in class, X = dψ(X̃). Now, by the above lemma, there exists a coordinate
chart φ̃ : U → Ṽ ⊂ U 0 such that Φ̃1v = X̃v for all v ∈ Ṽ . Define
φ := ψ ◦ φ̃ : U → V = ψ(Ṽ ).
Then, by the chain rule,
dφ = dψ · dφ̃.
Specifically,
Φ1q = dψv (Φ̃1v ) = dψv (X̃v ) = Xq
where v = ψ −1 (q). 

DRL 3E3A, University of Pennsylvania

E-mail address: [email protected]