Math 225B: Differential Geometry, Homework 6

Ian Coley
February 13, 2014
Problem 8.7.
R
Let be a 1-form on a manifold M . Suppose that c = 0 for every closed curve c in M .
Show that is exact.
Solution.
We claim
that this condition is equivalent to the following: if c : [0, 1] M is any curve,
R
then c depends only on c(0) and c(1), that is, the integral of is path independent. We
will show only the forward direction, which we require. Suppose that c, d are two paths such
that c(0) = d(0) and c(1) = d(1). Then the composition c d is a closed path at c(0), where
d(t) = d(1 t). Hence
Z
= 0.
cd

But we also have

=
cd

= 0 =
d

Z
=

.
d

Therefore without loss of generality assume that M is connected, since the following proof
may be repeated for any number of connected components. Then fix p M and let, define
a function f : M R by
Z
f (x) =

where is a path from p to x. Since is path independent, this makes sense for any .
Further, consider the function g(x) constructed as above with basepoint q M and path
and 0 from that q to p. Then we have
Z
Z
Z
g(x) = =
+ = C + f (x),

where C is a constant. Therefore every choice of basepoint yields the same function up to
positive constant depending on basepoint.
We claim that df = . To see this, we will show they take the same value on any vector
field X at any p M . Let p M . Then consider a curve c through p so that c(0) = p and
df (X)p = Xp (f ) =


d

f (c(t)) ,
dt
t=0

which is always possible by results in previous sections. We parametrise the path between
c(0) and c(t) the path ct (s) by ct (s) = c(st). Then
Z 
Z

d
d 1 
d


 =
c 
df (X) = f (c(t)) =
dt
dt ct t=0 dt 0 t t=0
t=0
  
Z
d 1
d

=
(ct (s))
ds
dt 0
ds
t=0

 
Z 1
d
d 

=
(c(st)) tc(st)  ds
dt 0
ds s
t=0

 
Z t

d
d

=
(c(u)) cu  du
dt 0
dt u
t=0



d 
= (c(t)) ct  
dt t t=0


d 
= (c(0)) c0  = (p)(Xp ).
dt 0
This completes the proof.
Problem 8.8.
A manifold M is called simply connected if M is connected and if every smooth map f :
S 1 M is smoothly contractible to a point.
(a) If M is smoothly contractible to a point, then M is simply connected.
(b) S 1 is not simply connected.
(c) S n is simply connected for n > 1.
(d) If M is simply connected and p M , then any smooth map f : S 1 M is smoothly
contractible to p.
(e) If M = U V , where U and V are simply connected open subsets with U V simply
connected, then M is simply connected.
(f) If M is simply connected, then H 1 (M ) = 0.
Solution.
(a) Let p : M M be the smooth contraction of M to a point, i.e. there is a homotopy
Ht : M M so that H0 = id and H1 = {p}. Then let f : S 1 M be any map. Then
the composition f Ht = Ft : I S 1 M is a smooth homotopy with F0 = f and
F1 = {p}. Therefore f is smoothly contractible to a point, so M is simply connected.
(b) If S 1 were simply connected, then every map f : S 1 S 1 would be homotopic to a
constant map. In particular, the identity map would be homotopic to a constant map,
so S 1 would be contractible. But S 1 is a compact manifold of positive dimension, so
it is not contractible. Therefore S 1 is not simply connected.
2

(c) Let f : S 1 S k be a map. We would like to show that this is homotopic to a constant.
Taking the hint, Sards theorem allows us to choose p S k \ f (S 1 ) when k > 1.
Recall that we showed in 1.1.13 that S k \ {p} is diffeomorphic to Rk for k 1. Further,
Rk is contractible, so it is simply connected. Therefore f (S 1 ) is diffeomorphic to a loop
in Rk , which is homotopic to a constant. Therefore this homotopy can be pulled back
along the diffeomorphism so f (S 1 ) is simply connected in S k \ {p}, hence in S k as well.
This completes the proof.
(d) Since M is path connected and embeddable into Euclidean space, it is in particular
path connected via local charts. Therefore suppose that f : S 1 M contracts to a
point q M . Then there is a path : [0, 1] M so that (0) = q and (1) = p.
Therefore if Ht is the homotopy between f and the point {q}, we have a homotopy
Ft : M M given by

f (x)
t=0

H2t (x)
0 < t < 1/2

.
Ft (x) = q
t = 1/2

(2t 1) 1/2 < t < 1

p
t=1
Hence f is contractible to an arbitrary point p M .
(e) Let f : S 1 M be a smooth map. Firstly, if the image of f lands in either U or V
alone, then f is contractible to a point, so we are done. Therefore suppose that the
image of f lands in both U and V , and that the image of f is not contained in U V .
Let us parametrise S 1 by [0, 1] for simplicity. Then we may break up [0, 1] into disjoint
segments Xi = [ai , bi ] so that f (Xi ) U or V . From here, we may connect the points
f (ai ) and f (bi ) in U V by a path i to create a loop in a simply connected subset
(either U or V ). Therefore there is a homotopy of this loop to the loop consisting of
i on [0, 1/2] and i1 on [1/2, 1]. Deleting the backwards direction, we have taken the
section f (Xi ) of our loop and smoothly contacted it to lie completely in U V .
Repeat this process a finite number of times, so that the image of f lies completely in
U V . Thus we are reduced to the first case, so we are done.
(f) We verify this via the integral. Assume that is a closed 1-form. Note that every
closed loop c in M is the image of S 1 in M under some map f . By assumption, f
is homotopic to a constant map p, so by Theorem 8.13 their pullbacks are also equal.
Therefore for a 1-form ,
Z
Z
Z
Z

=
f =
p=
=0
c

S1

S1

{p}

This holds for every closed curve c. Therefore by Problem 8.7, is exact. Therefore
H 1 (M ) = 0 since every 1-form is exact.

Problem 8.12.
(a) Let M be {(x, y) R2 : |(x, y)| < 1} together with a proper portion of its boundary,
and let = x dy. Show that
Z
Z
d 6=
,
M

even though both sides make sense.

(b) Similarly, find an example to Stokes Theorem when M = (0, 1) and is a 0-form
whose support is not compact.
(c) Examine a partition of unity for (0, 1) by functions with compact support to see just
why the proof of Stokes Theorem breaks down in this case.
Solution.
(a) Let us examine D, the unit disc in R2 . We may apply Stokes Theorem to this manifold:
Z
Z
.
d =
S1

Now let c : [0, 1] S 1 be an appropriately-oriented parametrisation of S 1 . Then

Z
Z 1
d =
c .
D

Now under the assumption that we have only a proper portion of the boundary of the
unit disk, we take < 1 so that we have c([0, ]) = M . Hence
Z
Z
d =
c .
M

Now we compute directly. Let c(t) = (cos(2t + ), sin(2t + )), where is an

appropriate rotation factor. On one hand,

x dx dy +
x dy dy = dx dy.
x
y

d = d(x dy) =

Since this is a volume form on M , we obtain

Z
d = |M | = .
M

However,
Z

=
M

Z
(x c) d(y c) =

c (x dy) =
0

cos(2t + )(2 cos(2t + )) dt

0

Z
=

cos(2t + ) d(sin(2t + ))
0

cos2 (2t + )2dt

Using the substitution u = 2t + ,

2+

1
cos (u)du =
2
2

2+

2 cos2 (u) 1 + 1 du

2+

2+
1
1
cos(2u) + 1 du = (sin(u) cos(u) + u) 
2
2

1
= (sin(2t + ) cos(2t + ) + 2t + ) 0
2
1
= + (sin(2 + ) cos(2 + ) sin() cos())
2
1
= + (sin(4 + 2) sin(2)) .
4
R
It is clear that, for small values of , we have M < . Therefore we check the
maximum of this integral on [0, 1]. It can be verified by taking the derivative of this
integral with respect to or using Wolfram Alpha that the maximum of this integral
occurs at = 1. Hence we always have, for our choice of (0, 1),
Z
Z
d >
,
=

so in particular they are not equal.

(b) Let M = (0, 1), and let = x, the coordinate function. Since M = , the integral of
any form over the boundary is zero. Then
Z
Z 1
Z
dx =
dx = 1 6= 0
.
M

(c) The step of the proof where Stokes Theorem fails is that, given a partition of unity ,
the support of need not be contained in finitely many . Therefore we cannot
conclude
Z X
XZ

d =
d = 0
M

since the marked equality only holds when is supported in finitely many .

Problem 8.13.
Suppose M is a compact orientable n-manifold (with no boundary), and is an (n 1)-form
on M . Show that d is 0 at some point.
Solution.
First, since we may put a finite partition of unity on M , we reduce to the local case. We
know that
Z
Z
Z
d =
= = 0.
M

Further, we know that d is a volume form, so it is of the form

d = f dx1 dxn
for some function f . Since at every point the term dx1 dxn contributes positively to
the integral, if f were always positive or always negative, then the entire integral would be
always positive or always negative. Therefore f = 0 identically, or f is somewhere positive
and somewhere negative, hence is somewhere zero by the mean value theorem. Since the
sign of f determines the sign of d, we are done.
Problem 8.14.
Let M1 , M2 Rn be compact n-dimensional manifolds with boundary with M2 M1 \ M1 .
Show that for any closed (n 1)-form on M1 ,
Z
Z
=
.
M1

M2

Solution.
An n-dimensional submanifold of Rn takes the induced orientation from Rn , so Mi are
orientable. Consider the manifold M = M1 \ M2 t M2 , so that M = M1 t M2 . Then
applying Stokes theorem,
Z
Z
Z
Z
0=
d =
=

,
M

M1

M2

where the minus sign is introduced since the outward pointing normal of M2 with respect
to M is the negation of the outward pointing normal of M2 with respect to M . Therefore
Z
Z
=

M1

M2

as claimed.