Sumersión Riemanniana
Sumersión Riemanniana
Sumersión Riemanniana
This course will offer an introduction to the elementary theory of smooth manifolds. Topics will include:
vector fields, differential forms, vector bundles, Lie derivatives, Stokes' theorem, and de Rham cohomology.
Math Tower 3-114. Phone: 632-8261. email: redden at math dot sunysb dot edu
Homework: Working homework problems is the only way to really learn the material. While you are
encouraged to work with others, you must write up all solutions on your own. Homework sets will usually be
collected in class on Thursdays.
Due 2/7: 2. 6, 8, 18
Midterm: Take Home midterm due Friday, March 28. Review/Practice HW, Review Solutions, MIDTERM
EXAM. Midterm Solutions
Final Exam: Thursday, May 15. 2:00p.m. Physics P129. Final Exam.
The midterm will be a timed, take-home exam (taken on the honor sys-
tem). You will have 3 hours to complete the exam. You may use your notes
and Spivak, but you may not obtain outside help from anyone. The problems
will be a mix of abstract theorems and concrete calculations. There will be
multiple problems, and you will be able to choose which ones you do (e.g.
work 2 out of 3 problems). Here are a few good problems to practice with:
1. Let f : R2 → R be defined by
f (x, y) = x3 + xy + y 3 + 1.
Calculate the Lie bracket on the vector fields X, Y, Z. What sort of structure
does this look like?
1
Math 531 Midterm Review Solutions
1. Let f : R2 → R be defined by
f (x, y) = x3 + xy + y 3 + 1.
f (x, y) = x3 + xy + y 3 + 1
f (0, 0) = 1
f (1/3, 1/3) = 32/27
f (−1/3, −1/3) = 28/27
h i
∂f ∂f
0 0 = Df = ∂x ∂y
= 3x + y x + 3y 2
2
The two points where Df has rank 0 (where Df does not have full rank)
occur at (0, 0) and (−1/3, −1/3). Therefore, since f has full rank at all points
M = f −1 (32/37), we know that M is a codimension 1 submanifold of R2 .
However, we do not yet know whether the other two sets are submani-
folds or not. The above process merely guarantees a submanifold structure
in certain situations; it does eliminate the possibility. In general, there is
no standard method for showing that something is not a manifold, though
usually you assume you have a submanifold and try to obtain some sort of
contradiction. We investigate the two unknown situations from above more
carefully.
First, let us consider, f −1 f (−1/3, −1/3). Some sneaky algebra (it helps
to use a computer) shows us that
x3 + xy + y 3 + 1 = 28/27
x3 + xy + y 3 − 1/27 = 0
1
(3x + 3y − 1)(9x2 + 3x − 9xy + 3y + 9y 2 + 1) = 0
27
We know that 3x + 3y − 1 = 0 is a submanifold of R2 . Further investigation
shows that the second factor
9x2 + 3x − 9xy + 3y + 9y 2 + 1 ≥ 0.
1
The above follows from a bit of Calc 3. We can show that this quadratic
function has a critical point at (−1/3, −1/3), f (−1/3, 1/3) = 0, and the
Hessian is positive definite (f is “concave up” in every direction). Therefore,
So, this particular subset is a submanifold, though you cannot simply use
the Implicit Function Theorem.
On the other hand, f −1 f (0, 0) is not a manifold because of the point
(0, 0). Seeing this is a little trickier, and I won’t expect you to do anything
like this on the exam. I’ll try to write up a clean solution to show this is not,
but this is not the important part of the problem. It is, though, good to see
that these things can happen.
S 2 = {(x, y, z) ∈ R3 |x2 + y 2 + z 2 = 1}
U = UN , V = US
x y
(u1 , u2 ) = ( , ) = u(x, y, z)
1−z 1−z
x y
(v 1 , v 2 ) = ( , ) = v(x, y, z)
1+z 1+z
2
Here, u and v are the standard coordinates induced from the stereographic
projection, and on overlap U ∩ V , we have functions u ◦ v −1 and v ◦ u−1 given
by the relations
ui
vi =
(u1 )2 + (u2 )2
vi
ui = 1 2
(v ) + (v 2 )2
The two trivializations u, v both give rise to a trivialization of the tangent
bundle over U and V , respectively. Explicitly, this is an equivalence of bun-
dles φU
φ
U × R2 ←−
U
TU
∂ ∂
(u, (a1 , a2 )) 7−→ a1 1
+ a2 2
∂u ∂u
along with the similarly defined equivalence φV . The transition function
φV ◦ φ−1
U is then a bundle equivalence
φV ◦ φ−1 2 2
U : (U ∩ V ) × R → (U ∩ V ) × R .
Since φV ◦ φ−1
U is a bundle map, we can consider this as a smooth map
φV ◦ φ−1
U : U ∩ V → Gl(2, R).
φV ◦φ−1
(a1 , a2 ) 7−→U (b1 , b2 ),
3
A simple calculation now shows us that
4
Now, we notice that the map
(x, θ) 7→ (x + 1, −θ),
π∗|x : Tx X → Tp M
(π∗ ◦ g∗ )|x : Tx X → Tp M
Then,
[X, Y ] = Z, [Y, Z] = X, [Z, X] = Y.
5
This is a straight-forward calculation using the definition of the Lie bracket.
∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂
[X, Y ] = z −y −z +x − −z +x z −y
∂y ∂z ∂x ∂z ∂x ∂z ∂y ∂z
∂ ∂
=y −x =Z
∂x ∂y
Notice that, since we know the answer will be vector field, we can ignore
terms with higher-order derivatives. Of course, if you just write everything
out, you will see that these cancel out. Similar calculations show [Z, X] =
Y, [Y, Z] = X.
One might notice that this structure looks a bit like the cross-product
structure on R3 (replacing X, Y, Z with i, j, k). In fact, this can be made ex-
plicit. The 3-dimensional vector space spanned by the vector fields {X, Y, Z},
together with the Lie bracket, form a Lie algebra. Likewise, R3 with the
cross-product structure also forms a Lie algebra. These two Lie algebras are
isomorphic under the map
aX + bY + cZ 7−→ abi + bb
j + cb
k.
The above map is linear and invertible, and the two brackets commute with
each other (check this).
Furthermore, it is not too difficult to show that the flow generated by
aX + bY + cZ is actually a 1-parameter subgroup of SO(3, R). This can be
seen through the work done in homework problem Ch. 5, number 6, as well
as a bit of what we talked about in class concerning matrix groups.
6
Math 531 Midterm
• This test is open notes and open Spivak. You may not consult any additional sources (or
people)
• You are to answer 2 of the first 3 questions and 1 of the last 2 questions.
The test is not meant to be confusing. Any extra information contained in the questions is
designed to clarify and not obscure, and the questions are not designed to trick you. Please let me
know if anything is unclear.
Math 531 Midterm
∂f
In other words, it is vector field with component functions .
∂xi
Consider the more general situation where M is a smooth manifold, and f : M → R a smooth
function.
∂f
a. Show that the functions i (where x is a local coordinate system) do not form the component
∂x
functions of a vector field. Show, however, that they do naturally form the component functions of
a 1-form.
b. Suppose we have a tensor g ∈ C ∞ (M, T M ∗ ⊗ T M ∗ ), and further suppose that at each point
x ∈ M , the tensor gx is symmetric and non-degenerate. In other words,
gx (v1 , v2 ) = gx (v2 , v1 ),
x = r cos θ
y = r sin θ
∂ ∂
Compute and in polar coordinates .
∂x ∂y
d2 x dx
= −kx − c ,
dt2 dt
where x(t) is the displacement of a mass at time t, and k, c are constants related to the strength of
the spring force and resistance force, respectively.
dx
a. Let y = , and rewrite the above equation in terms of a vector field on M = R2 .
dt
b. Show that flow induced by the vector field in (a) is a 1-parameter subgroup of Gl(2, R).
Solve 1 of the following 2 problems:
4. Implicit differentiation is a useful tool in calculus. Essentially, if we have
f (x, y) = c,
dy
then we solve for by implicitly differentiating and obtaining
dx
dy ∂f /∂x
=− .
dx ∂f /∂y
∂f
In other words, it is vector field with component functions .
∂xi
Consider the more general situation where M is a smooth manifold, and f : M → R a smooth
function.
∂f
a. Show that the functions i (where x is a local coordinate system) do not form the component
∂x
functions of a vector field. Show, however, that they do naturally form the component functions of
a 1-form.
b. Suppose we have a tensor g ∈ C ∞ (M, T M ∗ ⊗ T M ∗ ), and further suppose that at each point
x ∈ M , the tensor gx is symmetric and non-degenerate. In other words,
gx (v1 , v2 ) = gx (v2 , v1 ),
Solution:
Notice that we can define df ∈ C ∞ (M, T M ∗ ) in a coordinate-free way. That is, df (X) = X(f ),
where X is a vector field. Then, a simple calculation shows that
∂f i
df = dx
∂xi
∂f
in local coordinates. This implies ∂x i are the components of a (non-trivial) 1-form, and hence not
a vector field.
More explicitly, suppose ai , bi are the component functions of a vector field in x, y coordinates,
respectively. Then these are related by the equation
∂y j
b j = ai .
∂xi
However, we see that
∂f ∂f ∂xi
= .
∂y j ∂xi ∂y j
This does not satisfy the transformation property for components of vector fields, but instead
satisfies the transformation property for differential forms.
In fact, we can see that that
∂f ∂y j
i
∂f i ∂x j ∂f
i
dx = j i j
dy = j dy j ,
∂x ∂y ∂x ∂y ∂y
whereas 2
∂xi
∂f ∂ ∂f ∂ ∂f ∂
i i
= j j 6= .
∂x ∂x ∂y ∂y ∂y j ∂y j ∂y j
∂f
Therefore, ∂xi
dxi is a well-defined 1-form on M , but not a well-defined vector field.
Now, suppose we have the tensor g. Such a tensor g is a non-degenerate inner product on
each tangent space and is called a metric (if g is positive-definite, then it is a Riemannian metric).
(Remark: we don’t need the symmetric property here, but it makes indices a little easier to deal
with.)
If g ∈ C ∞ (M, T M ∗ ⊗ T M ∗ ), then we can equivalently consider
g ∈ C ∞ (M, Hom(T M, T M ∗ ))
g −1 ∈ C ∞ (M, Hom(T M ∗ , T M )) ∼
= C ∞ (M, T M ⊗ T M ).
d g −1
C ∞ (M ) −→ C ∞ (M, T M ∗ ) −→ C ∞ (M, T M )
f 7−→ df 7−→ g −1 (df ) = ∇f
Since g is non-degenerate, the matrix gij is invertible. Hence, we also have the tensor
∂ ∂
g −1 = g ji ⊗ .
∂xi ∂xj
We now see that we can use g −1 to transform the above df into a vector field. We define
−1 kj ∂ ∂ ∂f i
∇f = g (df ) = g ⊗ k dx
∂xj ∂x ∂xi
∂ ∂f
= g kj j δki i
∂x ∂x
∂f ∂
= g ij i j
∂x ∂x
It is simple to check that ∇f is now a well-defined vector field. Suppose that
∂ ∂
g −1 = g βα .
∂y α ∂y β
Then, we see that
α
∂y β ∂f ∂xk
l
αβ ∂f ∂ ij ∂y ∂x ∂
g = g
∂y α ∂y β ∂xi ∂xj ∂xk ∂y α ∂y β ∂xl
∂f ∂
= g ij δik δjl k l
∂x ∂x
ij ∂f ∂
=g
∂xi ∂xj
Therefore, we see that ∇f is a well-defined vector field.
The above demonstrates the general principle that if some quantity is written in local coordi-
nates, each index should appear “up” and “down” an equal number of times.
∂f
In summary, we see that we can use the tensor g to transform the functions ∂x i into components
of a vector field. In the situation where M = Rn with standard coordinate x, and gij = δij is the
usual Euclidean metric, we see that
∂f ∂ ∂f ∂ ∂f ∂
∇f = g ij i j
= δ ij i j = .
∂x ∂x ∂x ∂x ∂xi ∂xi
Letting u and p denote the Euclidean and polar coordinate systems, respectively, we know that
D(p ◦ u−1 ) = D((u ◦ p−1 )−1 ) = [D(u ◦ p−1 )]−1 ].
As easy calculation shows us that
∂x ∂x
−1 ∂r ∂θ cos θ −r sin θ
D(u ◦ p ) = ∂y ∂y = .
∂r ∂θ
sin θ r cos θ
Therefore,
∂r ∂r −1
−1 ∂x ∂y cos θ −r sin θ
D(p ◦ u ) = ∂θ ∂θ =
∂x ∂y
sin θ r cos θ
cos θ sin θ
= − sin θ cos θ
r r
Therefore, we see that
∂ ∂ 1 ∂
= cos θ − sin θ
∂x ∂r r ∂θ
∂ ∂ 1 ∂
= sin θ + cos θ
∂y ∂r r ∂θ
d2 x dx
2
= −kx − c ,
dt dt
where x(t) is the displacement of a mass at time t, and k, c are constants related to the strength of
the spring force and resistance force, respectively.
dx
a. Let y = , and rewrite the above equation in terms of a vector field on M = R2 .
dt
b. Show that flow induced by the vector field in (a) is a 1-parameter subgroup of Gl(2, R).
Solution:
dx
First, let y = dt
. The 2nd-order differential equation now becomes a system of first-order equations
(
dx
dt
=y
dy
dt
= −kx − cy
Remembering that we are solving for functions (x(t), y(t)), we think about solutions to the above
equation as paths γ = (γ x , γ y ) : (−, ) → R2 . The equation is then equivalent to
( x
dγ ∂ ∂
dt ∂x
= y ∂x
y
dγ ∂ ∂
dt ∂y
= (−kx − cy) ∂y
Therefore, the integral curves, with γ(0) = (x0 , y0 ), will be of the form
x(t) 0 1 x0
γ(t) = = exp t .
y(t) −k −c y0
φ0 = e0A = I, φt ◦ φs = eA(s+t) .
Solve 1 of the following 2 problems:
4. Implicit differentiation is a useful tool in calculus. Essentially, if we have
f (x, y) = c,
dy
then we solve for by implicitly differentiating and obtaining
dx
dy ∂f /∂x
=− .
dx ∂f /∂y
f (x, g(x)) = c.
Hence, we calculate
d d
|x0 (f (x, g(x))) = |x (c)
dx dx 0
dx dg
D1 (f (x0 , g(x0 ))) |x + D2 (f (x0 , g(x0 ))) |x0 = 0
dx 0 dx
Then, using the shorthand of y(x) = g(x), we obtain the standard expression
∂f ∂f dy
|(x0 ,y0 ) + |(x0 ,y0 ) |(x0 ,y0 ) = 0
∂x ∂y dx
∂f /∂x dy dg
− |(x0 ,g(x0 ) = |(x0 ,y0 ) = |x
∂f /∂y dx dx 0
dy ∂f /∂x
=− .
dx ∂f /∂y
∂f ∂f
Notice that since wish to divide by ∂y
, we would have wanted to require ∂y
to be non-zero.
f∗ : Or(M ) → Or(N ).
Because of continuity, f∗ cannot map a connected component to two different connected components.
Therefore, f∗ will preserve orientation at all points.
Alternatively, we can let x0 ∈ M be an arbitrary point. Assume f∗ : Tx0 M → Tf (x0 ) N preserves
orientation. Then, let x ∈ M be any other point. Since M is connected (and connectedness is
equivalent to path connectedness for manifolds), we can choose some path
γ : [0, 1] → M
0 7→ x0
1 7→ x
connecting x0 and x. The tangent spaces of M and N are continuously oriented, and f is a
smooth map. This implies that if f∗p preserves orientation, then f∗ preserves orientation in some
neighborhood of p. We then take open cover containing the image of γ and see that fx must preserve
orientation.
HW Due Thursday 4/24
L∗g ω = ω ∀g ∈ G.
H 1 (g) ∼
= (g/[g, g])∗ .
What is H 1 (gl(n, R), where gl(n, R) denotes the Lie algebra of all n × n real
matrices?
1
Math 531 Final
May 15, 2008
Solve any 2 of the following 3 problems:
1. Show directly that RP n is a smooth manifold by giving local coordinate patches and calculating
the transition functions (Hint: use homogeneous coordinates).
2. Let M = R3 , and consider the 2-dimensional sub-bundle of T R3 spanned by the vector fields
∂ ∂ ∂ ∂ ∂ ∂
+ + f (x, y) and + + g(x, y) .
∂x ∂y ∂z ∂x ∂y ∂z
For what functions f, g will this be an integrable distribution?
3. Let M = R2 . Apply Stokes Theorem to integrals of 1-forms (in local coordinates) to obtain
two standard results from multi-variable calculus: The Fundamental Theorem of Line Integrals and
Green’s Theorem. (You don’t need to remember what these theorems are to answer this question.)
Solve any 3 of the following 4 problems: (In the following, H k (M ) denotes de Rham coho-
mology.)
5. Let M n be a compact, connected, oriented n-manifold. Let Dn be a closed disc around a point
x0 ∈ M . Calculate the de Rham cohomology of M n − Dn .
7. Let M be a compact, connected, oriented, n-manifold. What is the Euler class of the vector
bundle Λn T M ∗ . Specifically, in what cohomology group does it live, and is it non-zero?