DIFFERENTIAL GEOMETRY HW 5

CLAY SHONKWILER

5
Show that the law of cosines in spherical geometry is

cos c = cos a cos b + sin a sin b cos θ.

Proof. Consider the spherical triangle depicted below:

Form radii from each of the vertices of the triangle to the origin of the sphere,
denoted by OA, OB and OC. Since this triangle is on a unit sphere, the
length of the arc a is given by the measure of the angle BOC in radians,
the length of b is the measure of AOC and the length of c is the measure
of AOB. Thus, we may as well think of BOC as angle a and similarly
for these other two angles. Now, θ is, in addition to the angle between a
and b, the dihedral angle between the plane P1 defined by OB and OC and
the plane P2 defined by OA and OC. Let Ni be the unit normal to Pi for
−−→ −−→
i = 1, 2; then θ is the angle between N1 and N2 . Now N1 = OB × OC and
−→ −−→
N2 = OA × OC. Therefore, by definition of the inner product,

−−→ −−→ −→ −−→

hN1 , N2 i = |OB × OC||OA × OC| cos θ
 −−→ −−→   −→ −−→ 
= |OB||OC| sin a |OA||OC| sin b cos θ
= sin a sin b cos θ.
1
2 CLAY SHONKWILER

−−→ −−→ −→ −−→

On the other hand, N1 = OB × OC and N2 = OA × OC, so
−−→ −−→ −→ −−→
hN1 , N2 i = hOB × OC, OA × OCi
−−→ −→ −−→ −−→ −−→ −−→ −−→ −→
= hOB, OAihOC, OCi − hOB, OCihOC, OAi
−−→ −→  −−→ −−→   −−→ −→ 
= |OB||OA| cos c − |OB||OC| cos a |OC||OA| cos b
= cos c − cos a cos b.

Equating these two, we see that

cos c − cos a cos b = sin a sin b cos θ,

so
cos c = cos a cos b + sin a sin b cos θ.


10
Let S ⊂ R3 be a smooth surface homeomorphic to S 2 . Suppose Γ ⊂ S
is a simple closed geodesic, and let A and B be the two regions on S which
have Γ as boundary. Let N : S → S 2 be the Gauss map. Prove that N (A)
and N (B) have the same area on S 2 .
R
Proof. Recall that, if R is a region on a surface, then R κd(area) is equal to
the signed area of N (R) on S 2 . Note that, since Γ is a simple closed curve
and S is homeomorphic to S 2 , A and B are both homeomorphic to a disk;
hence, χ(A) = χ(B) = 1. Now, using the Gauss-Bonnet Theorem,
Z Z k
X
2π = 2πχ(A) = κd(area) + κg (s)ds + θi
A ∂A i=0
Z
= Area N (A) + 0ds + 0
∂A
= Area N (A).

Similarly,
Z Z k
X
2π = 2πχ(B) = κd(area) + κg (s)ds + θi
B ∂B i=0
Z
= Area N (B) + 0ds + 0
∂B
= Area N (B).

Thus, we see that N (A) and N (B) have the same area on S 2 . 
 DIFFERENTIAL GEOMETRY HW 5 3

12
Let S ⊂ R3 be a surface homeomorphic to a cylinder and with Gaussian
curvature K < 0. Show that S has at most one simple closed geodesic.

Proof. Suppose there are two simple closed geodesics on S, γ1 and γ2 . Of

course, there are 3 possibilities, either γ1 and γ2 don’t intersect, they inter-
sect in one point, or they intersect in more than one point. If they intersect
in more than one point, then, looking at two adjacent intersection points
(as viewed traveling along, say, γ1 ), the region bounded by γ1 and γ2 is
homeomorphic to a disc, since γ1 and γ2 are simple closed curves. However,
by problem 11 (discussed in class), this is impossible, since K < 0 on S, so
we see that γ1 and γ2 can intersect in at most 1 point.
On the other hand, suppose γ1 and γ2 don’t intersect at all. Let S 0 be
the region bounded by γ1 and γ2 . Then, since γ1 and γ2 are simple closed
curves and S is homeomorphic to a cylinder, S 0 is also homeomorphic to a
cylinder. Hence, χ(S 0 ) = 0. Now, by the Gauss-Bonnet Theorem,

Z Z k
X
0
0 = 2πχ(S ) = κd(area) + κg (s)ds + θi
S0 ∂S 0 i=0
Z Z
= κd(area) + 0ds + 0
S0 ∂S 0
Z
= κd(area)
S0
< 0,

which is an obvious impossibility.

Thus, we conclude that γ1 and γ2 must intersect in exactly one point.
However, for this to be the case, γ1 and γ2 must be tangent at their point
of intersection, which, by the uniqueness of geodesics, implies that γ1 and
γ2 describe the same curve. Therefore, we conclude that S has at most one
simple closed geodesic. 

13
Let S ⊂ R3 be a smooth closed surface of positive curvature, and thus
homeomorphic to S 2 . Show that if Γ1 and Γ2 are two simple closed geodesics
on S, then they must intersect each other.

Proof. Suppose Γ1 and Γ2 don’t intersect each other. Then, since Γ1 and
Γ2 are simple closed curves and S is homeomorphic to S 2 , the region S 0
bounded by Γ1 and Γ2 is homeomorphic to a cylinder. Therefore, since
4 CLAY SHONKWILER

χ(S 0 ) = 0,
Z Z k
X
0
0 = 2πχ(S ) = κd(area) + κg ds + θi
S0 ∂S 0 i=0
Z Z
= κd(area) + 0ds + 0
S0 ∂S 0
Z
= κd(area)
S0
> 0,
which is impossible. Therefore, we conclude that Γ1 and Γ2 must intersect
each other. 

(b)
Summarize the proof of the Gauss-Bonnet Theorem.
Theorem 13.1. Let U be an open set in R2 and X : U → S ⊂ R3 a regular
local parametrization of the surface S. Let R ⊂ X(U ) be a region homeo-
morphic to a disk. Let α : I → ∂R be a piecewise smooth parametrization of
∂R by arc length, with vertices at α(s0 ), α(s1 ), . . . , α(sk ) and with exterior
angles θ0 , θ1 , . . . , θk at these vertices. Then
Z Z k
X
Kd(area) + κg (s)ds + θi = 2π.
R ∂R i=0

Proof. (sketch) We may as well assume X gives an orthogonal parametriza-

tion of S. Let W (s) = α0 (s), which is unit length and well-defined on the
Xu
smooth pieces of ∂R. If V (s) = |X u|
, then φ is the angle from V to W .
Now, the rate of change of W with respect to the frame (Xu , Xv ) is equal
to the rate of change of V with respect to the same frame plus the rate
of change of the angle between W and V . Since the algebraic value of a
covariant derivative of a vector field tells us how that vector field changes
with respect to (Xu , Xv ), this means that
   
DW DV dφ
= +
ds ds ds
√ dX √
along each smooth arc of ∂R. Now, since dV ds = E dsu + d dsE Xu , only the
first term in this sum contributes to
dV √
       
DW DV dφ dV dφ dφ
= = ,N × V + = , GXv + ,
ds ds ds ds ds ds ds
so a simple computation shows that
   
DW 1 dv du dφ
= (EG)−1/2 Gu − Ev + .
ds 2 ds ds ds
 DIFFERENTIAL GEOMETRY HW 5 5

Hence, along each smooth arc,

Dα0
   
1 −1/2 dv du dφ
κg (s) = = (EG) Gu − Ev + .
ds 2 ds ds ds
Pk
Furthermore, ∂R dφ
R
ds ds+ i=0 θi gives the total change in angle of inclination
with respect to the frame (Xu , Xv ) along the simple closed curve ∂R; that
is, 2π. Hence,
Z k Z   Z k
X 1 −1/2 dv du dφ X
κg (s)ds + θi = (EG) Gu − Ev ds + ds + θi
∂R i=0 ∂R 2 ds ds ∂R ds i=0
Z  
1 −1/2 dv du
= (EG) Gu − Ev ds + 2π.
∂R 2 ds ds
Now, using Green’s Theorem to convert the remaining integral to a surface
integral, we see that
Z   Z  
1 −1/2 dv du 1 −1/2 dv 1 −1/2 du
(EG) Gu − Ev ds = (EG) Gu − (EG) Ev ds
∂R 2 ds ds ∂R 2 ds 2 ds
Z     
1 1
(1) = (EG)−1/2 Gu + (EG)−1/2 Ev dudv.
R 2 u 2 v
R
With a little manipulation, we can see that this is − R Kd(area). Using the
Gauss equation
−EK = Γ112 Γ211 + Γ212,u + Γ212 Γ212 − Γ211 Γ212 − Γ211,v − Γ211 Γ222
and making the appropriate substitutions for the Christoffel symbols, we
get an equation equivalent to
−1 h i
K= (EG)−1/2 (Ev (EG)−1/2 )v + (Gu (EG)−1/2 )u .
2
√
Since EG is the volume multiplier (as per problem 4 below),
Z Z Z     
1/2 1 −1/2 1 −1/2
− Kd(area) = − K(EG) dudv = − − (EG) Gu + (EG) Ev dudv
R R R 2 u 2 v
Z     
1 1
= (EG)−1/2 Gu + (EG)−1/2 Ev dudv,
R 2 u 2 v
which is the integral in (1). Therefore, we see that
Z k
X Z
κg (s)ds + θi = − Kd(area) + 2π,
∂R i=0 R

so
Z Z k
X Z Z
Kd(area) + κg (s)ds + θi = Kd(area) − Kd(area) + 2π = 2π.
R ∂R i=0 R R


6 CLAY SHONKWILER

4
(a): Let A and B be two vectors in the plane, with an angle θ be-
tween them. Then the area of the parallelogram that they span is
|A||B| sin θ.

Proof. If A and B define two sides of the parallelogram, then, as

h
in the below drawing, sin θ = |B| where h is the altitude of the
parallelogram.

Hence, h = |B| sin θ, so the area of the parallelogram is

|A|h = |A||B| sin θ.


(b): If A1 , . . . , An are vectors in Rn , and

gij = hAi , Aj i,
show that q
det(gij ) = vol(A1 , . . . , An ).

Proof. Let A be the matrix with Ai as its ith row. Then A is a

linear transformation that takes the standard oriented basis of Rn
to (A1 , . . . , An ). The volume of the solid spanned by (A1 , . . . , An )
is given by vol(e1 , . . . , en ) · det A = 1 · det A = det A. Now, if Ai =
(Ai1 , . . . , Ain ), then
    
A11 · · · A1n A11 · · · An1 hA1 , A1 i · · · hA1 , An i
AAt =  ... .. ..   .. .. ..  =  .. .. ..
 = (gij ).
 
. .  . . .   . . .
An1 · · · Ann A1n · · · Ann hAn , A1 i · · · hAn , An i

Therefore, det(gij ) = det A · det At = (det A)2 = (vol(A1 , . . . , An ))2 ,

so we see that
q
det(gij ) = vol(A1 , . . . , An ).


 DIFFERENTIAL GEOMETRY HW 5 7

(c): Let x : U → M be a coordinate neighborhood in a Riemannian

manifold M , and R ⊂ x(U ) an open set with compact closure. Show
that
Z q
vol R = det(gij )dx1 · · · dxn ,
x−1 R

where gij = h∂/∂xi , ∂/∂xj i.

Proof. This is theD definition

E of the volume of R. To see why it makes
∂ ∂
p
sense, let gij = ∂xi , ∂xj . Then, by part (b), det(gij ) gives the
 
volume of the box ∂x∂ 1 , . . . , ∂x∂n . Then pulling back to Rn and
adding over all these boxes gives
Z q
vol R = det(gij )dx1 · · · dxn .
x−1 R

(d): Double check that the right hand side is invariant under change
of coordinates.

Proof. Suppose y : V → M is another

D coordinate
E neighborhood
∂ ∂
containing the region R. Let hij = ∂yi , ∂yj . Then
 
∂ ∂
q
vol ,..., = det(hij )
∂y1 ∂yn

and so, by the argument in (c),

Z q
vol R = det(hij )dy1 · · · dyn .
y −1 R

On the other hand,

   
∂ ∂ ∂ ∂
q q
det(gij ) = vol ,..., = Jvol ,..., = J det(hij ),
∂x1 ∂xn ∂y1 ∂yn
 
∂xi
where J = det ∂y j
is the jacobian of the change of coordinate
function y −1 ◦ x. Hence,
p
det(gij )
Z q Z Z q
det(hij )dy1 · · · dyn = Jdx1 · · · dxn = det(gij )dx1 · · · dxn ,
y −1 R x−1 R J x−1 R

so we see that the computation of the volume of R is independent

of the choice f coordinates. 
8 CLAY SHONKWILER

4
A function g : R → R given by g(t) = yt + x, t, x, y ∈ R, y > 0, is called a
proper affine function. The subset of all such functions with respect to the
usual composition law forms a Lie group G. As a differentiable manifold G
is simply the upper half-plane {(x, y) ∈ R2 : y > 0} with the differentiable
structure induced from R2 . Prove that:
(a): The left-invariant Riemannian metric of G which at the neutral
element e = (0, 1) coincides with the Euclidean metric (g11 = g22 =
1, g12 = 0) is given by g11 = g22 = y12 , g12 = 0 (this is the metric of
the non-euclidean geometry of Lobatchevski).

Proof. g11 = g22 = y12 , g12 = 0 certainly defines a metric h, ip on each

point p ∈ G; since gij is smooth for each i, j ∈ {1, 2}, this defines a
Riemannian metric on G. Furthermore, at (0, 1), y12 = 11 = 1, so this
metric coincides with the Euclidean metric at the neutral element of
G. Thus, it only remains to show that this metric is left-invariant
on G.
To that end, let g0 (t) = y0 t + x0 be an element of G. Then left
translation by g0 is given by
Lg0 (g) = g0 ◦ g(t) = y0 (yt + x) + x0 = y0 yt + (y0 x + x0 ),
which is represented by the element (y0 x + x0 , y0 y) ∈ R2 . Therefore,
∂Lg ∂Lg
!
0 0
∂x ∂y
dLg0 = ∂Lg0 ∂Lg0
∂x ∂y
 
y0 0
= .
0 y0
∂ ∂ ∂ ∂
If v = a ∂x + b ∂y ∈ Tp G and w = c ∂x + d ∂y ∈ Tp G, then
 
∂ ∂ ∂ ∂ ac + bd
hv, wig = a + b ,c +d = acg11 +(ad+bc)g12 +bdg22 = .
∂x ∂y ∂x ∂y (x,y) y2
On the other hand,
h(dLg0 )g v, (dLg0 )g wiLg0 (g) = hy0 v, y0 wi(y0 x+x0 ,y0 y)
= y02 hv, wi(y0 x+x0 ,y0 y)
ac + bd
= y02
(y0 y)2
ac + bd
=
y2
= hv, wig0 ,
so the metric is left-invariant. 
 DIFFERENTIAL GEOMETRY HW 5 9
√
(b): Putting (x, y) = z = x + iy, i = −1, the transformation z 7→
z 0 = az+b
cz+b , a, b, c, d ∈ R, ad − bc = 1 is an isometry of G.

Proof. Let f denote the above Möbius transformation. Note that f

is defined on all of C = R2 except when z = −d
c ∈
/ G, so f is defined
on all of G. Now, if z = x + iy,
a(x + iy) + b
f (z) = f (x + iy) =
c(x + iy) + d
ax + b + iay cx + d − icy
= ·
cx + d + icy cx + d − icy
acx2 + (ad + bc)x + acy 2 + bd + i(acxy − acxy + (ad − bc)y)
=
(cx + d)2 + c2 y 2
acx2 + (ad + bc)x + acy 2 + bd y
= +i ,
(cx + d)2 + c2 y 2 (cx + d)2 + c2 y 2
which is in G, since (cx+d)y2 +c2 y2 > 0. Therefore, we see that f : G →
G. Furthermore, we know from complex analysis that Möbius trans-
formations are biholomorphic from their domain onto their image,
so we see that f : G → G is actually a diffeomorphism. Therefore, if
we can show that f preserves the first fundamental form, that will
suffice to show that f is an isometry.
Recall that the first fundamental form is given by
dx2 dy 2 dx2 + dy 2
ds2 = g11 dxdx + 2g12 dxdy + g22 dydy = + = .
y2 y2 y2
Now, letting z = x + iy, we know that dz = dx + idy and dz̄ =
dx − idy. Hence,
dzdz̄ = (dx + idy)(dx − idy) = dx2 + idxdy − idxdy − i2 dy 2 = dx2 + dy 2 .
z−z̄
Also, z − z̄ = (x + iy) − (x − iy) = 2iy, so y = 2i . Therefore,
dx2 + dy 2 dzdz̄ −4dzdz̄
ds2 = =
= .
y2 z−z̄ 2 (z − z̄)2
2i

Since z and z̄ are the coordinates of G, f (z) and f (z̄) are the coordi-
nates on the image, which is also G. Hence, since the first fundamen-
tal form on the domain is given by ds2 = −4dzdz̄
(z−z̄)2
, the fundamental
form on the range is given by
−4df (z)df (z̄) −4f 0 (z)dzf 0 (z̄)dz̄
ds2 = = .
(f (z) − f (z̄))2 (f (z) − f (z̄))2
Now, using the quotient rule,
a(cz + d) − c(az + b) ad − bc 1
f 0 (z) = = =
(cz + d)2 (cz + d)2 (cz + d)2
10 CLAY SHONKWILER

and
a(cz̄ + d) − c(az̄ + b) ad − bc 1
f 0 (z̄) = 2
= 2
= .
(cz̄ + d) (cz̄ + d) (cz̄ + d)2
Hence,
dzdz̄
f 0 (z)dzf 0 (z̄)dz̄ = .
(cz + d)2 (cz̄ + d)2
On the other hand,
az + b az̄ + b (az + b)(cz̄ + d) − (az̄ + b)(cz + d)
f (z) − f (z̄) = − =
cz + d cz̄ + d (cz + d)(cz̄ + d)
(ad − bc)z + (bc − ad)z̄
=
(cz + d)(cz̄ + d)
z − z̄
= .
(cz + d)(cz̄ + d)
Hence,
−4df (z)df (z̄) −4 (cz+d)dzdz̄
2 (cz̄+d)2 −4dzdz̄
2
ds = 2
=  2 = .
(f (z) − f (z̄)) z−z̄ (z − z̄)2
(cz+d)(cz̄+d)

Hence, f preserves the first fundamental form on G, so we see that

f is an isometry. 

DRL 3E3A, University of Pennsylvania

E-mail address: [email protected]