MATH534A, Exam 1, Soultions

Problem 2. Let S 2 be the standard unit sphere, U ⊂ S 2 be the northern hemisphere, and p be
a point in U . Let also (x1 , x2 ) be coordinates in U given by x1 (x, y, z) = x, x2 (x, y, z) = y, and let
∂ ∂ ∂ 2 3
∂x1 , ∂x2 be the associated basis of Tp M . Compute di( ∂x1 ), where i : S → R is the inclusion map,
and di is its differential (at p).
Denote by px , py , pz the coordinates of p in R3 .
Solution 1 (using smooth curves). By definition of the differential in terms of smooth curves,
we have 
∂ d 
di( ) =  i(γ(t)),
∂x1 dt t=0
where γ(t) is any smooth curve in S 2 such that γ(0) = p, γ 0 (0) = ∂x∂ 1 . In coordinates (x1 , x2 ), a
curve γ in S 2 near p can be represented by two functions x1 (t), x2 (t). In terms of these functions,
the condition γ(0) = p becomes x1 (0) = px , x2 (0) = py . Furthermore, the coordinates of the tangent
vector γ 0 (0) in the basis ∂x∂ 1 , ∂x∂ 2 are x01 (0), x02 (0). So, γ 0 (0) = ∂x∂ 1 is equivalent to x01 (0) = 1, x02 (0) = 0.
Thus, as γ we can take the curve given in coordinates (x1 , x2 ) by

x1 (t) = px + t, x2 (t) = py .

Further, i(γ(t)) = γ(t) where in the right-hand side we interpret γ as a curve in R3 . In R3 , the curve
γ is given by three functions
p q
x(t) = px + t, y(t) = py , z(t) = 1 − x(t) − y(t) = 1 − (px + t)2 − p2y .
2 2

Therefore,
 
∂ d  d  q px
di( ) =  i(γ(t)) =  (px + t, py , 1 − (px + t)2 − p2y ) = (1, 0, − ).
∂x1 dt t=0 dt t=0 pz

Since we represent the curve γ in R3 by its x, y, z coordinates, the latter expression gives coordinates
of di( ∂x∂ 1 ) in the basis ∂x
∂ ∂
, ∂y ∂
, ∂z . So, another possible form of the answer is:

∂ ∂ px ∂
di( )= − .
∂x1 ∂x pz ∂z
∂
Note that ∂z component can also be found using the fact that the resulting vector should be orthogonal
to (px , py , pz ), and there is in fact no need to differentiate the square root function.
Solution 2 (using Jacobian matrices). Take (x1 , x2 ) coordinates in S 2 and standard (x, y, z)
coordinates in R3 . In these coordinates, the mapping i can be written as
q
x = x1 , y = x2 , z = 1 − x21 − x22 . (1)

The differential of i is a linear mapping Tp S 2 → Tp R3 whose matrix in bases ∂x∂ 1 , ∂x∂ 2 for Tp S 2 and
∂ ∂ ∂ 3
∂x , ∂y , ∂z for Tp R is the Jacobian of (1). Furthermore, by definition of the matrix of a linear map,
the image of the first basis vector is the first column, so
   
∂x  1 
∂x
∂  ∂y1 

0
 ∂ px ∂
di( ) =  ∂x1  =   = − .
  
∂x1 ∂z  x
− √ 12 2  ∂x pz ∂z
∂x

1 p 1−x −x 1 2 p

1
 Solution 3 (using differential operators). By definition of the differential, for any function
f = f (x, y, z) in R3 , we have
∂ ∂ ∗ ∂ ∂
di( )f = (i f ) = (f ◦ i) = (f | 2 ).
∂x1 ∂x1 ∂x1 ∂x1 S
In coordinates (x1 , x2 ), we have
q
f |S 2 = f (x1 , x2 , 1 − x21 − x22 ),
so
∂ ∂ ∂
q
di( )f = (f | 2 ) = f (x1 , x2 , 1 − x21 − x22 )
∂x1 ∂x1 S ∂x1
 
∂f ∂f ∂ ∂ px ∂
q 
= + 2 2
1 − x1 − x2  = ( − )f,
∂x ∂z ∂x1 p ∂x pz ∂z
meaning that
∂ ∂ px ∂
di( )= − .
∂x1 ∂x pz ∂z

Problem 3. Let S 2 be the standard unit sphere, n ∈ S 2 be its north pole, and let σ : S 2 \{n} → C
be the stereographic projection from n (here we identify R2 and C). Define a map φ : S 2 → CP1 by
(
[σ(p) : 1] if p 6= n,
φ(p) =
[1 : 0] if p = n.
Prove that φ is a diffeomorphism.
Solution. Let s ∈ S 2 be the south pole. Then we have a smooth altas {(U, σ), (Ũ , σ̃)} on S 2 ,
where U = S 2 \ {n}, Ũ = S 2 \ {s}, and σ̃ : Ũ → C is the stereographic projection from s.
˜ on CP1 , where V = CP1 \ {[1 : 0]}, Ṽ = CP1 \ {[0 : 1]},
Similarly, we have an altas {(V, ξ), (Ṽ , ξ)}
˜
ξ([x, y]) = x/y, ξ([x, y]) = y/x.
Note that φ maps U bijectively to V . Indeed, φ(U ) ⊂ V by definition of φ. Furthermore, φ|U = i◦σ,
where i : C → V is given by i(z) = [z : 1]. Both i and σ are bijections, so φ|U : U → V is a
bijection. And since we also have φ(n) = [1 : 0], it follows that φ maps S 2 = U t {n} bijectively to
CP1 = V t {[1 : 0]}.
Now we prove that φ is smooth. First take p ∈ U . Then φ(U ) ∈ V , so to show that φ is smooth at
p we can use the coordinate representation of φ in charts (U, σ), (V, ξ). This coordinate representation
is
ξ ◦ φ ◦ σ −1 (z) = ξ([σ ◦ σ −1 (z) : 1]) = ξ([z : 1]) = z,
i.e. it is the identity map and hence smooth. (In this computation we used that φ(p) = [σ(p) : 1]
in U .) So it remains to verify smoothness of φ at n. To do that, it suffices to take any chart whose
domain contains n and any chart whose domain contains φ(n) = [1 : 0], and then verify smoothness
˜ The
of the corresponding coordinate representation of φ. As such charts, we take (Ũ , σ̃) and (Ṽ , ξ).
corresponding coordinate representation of φ is ξ˜ ◦ φ ◦ σ̃ (z). For z 6= 0, we have σ̃ (z) 6= n, so
−1 −1

ξ˜ ◦ φ ◦ σ̃ −1 (z) = ξ([σ
˜ ◦ σ̃ −1 (z) : 1]) = ξ([1/z̄
˜ : 1]) = z̄.
But for z = 0 we have ξ˜ ◦ φ ◦ σ̃ −1 (z) = ξ˜ ◦ φ(n) = ξ([1˜ : 0]) = 0 = z̄, so the formula
ξ˜ ◦ φ ◦ σ̃ −1 (z) = z̄
is actually valid for any z ∈ C, which proves that ξ˜ ◦ φ ◦ σ̃ −1 (z) is smooth, and thus φ is smooth at n
(and hence everywhere).
These computations also show that, first, φ maps U diffeomorphically to V , and second, φ maps
some open neighborhood of n diffeomorphically to an open neighborhood of [1 : 0] (it is not hard to
see that in fact φ maps Ũ diffeomorphically to Ṽ ). Since U and an open neighborhood of n cover
S 2 , it follows that φ is a local diffeomorphism. And since it is also bijective, it is actually a global
diffeomorphism, as desired.

2
 Problem 4. Prove that the set {(x, y, z) ∈ RP2 | xy = z 2 }1 is a smooth submanifold of RP2 .
Solution 1. First note that even though the coordinates x, y, z for a point in RP2 are only defined
up to a common non-zero factor, the equation xy = z 2 is invariant under such rescaling of variables
and hence the set of its solutions in RP2 is well-defined. Now we prove that this set is a submanifold.
We have the following charts in RP2 : Ux = {(x : y : z) ∈ RP2 | x 6= 0}, Uy = {(x : y : z) ∈ RP2 | y 6=
0}, Uz = {(x : y : z) ∈ RP2 | z 6= 0}. To prove that S = {(x : y : z) ∈ RP2 | xy = z 2 } is a submanifold
of RP2 , it suffices to show that S ∩ Ux is a submanifold of Ux , S ∩ Uy is a submanifold of Uy , and
S ∩ Uz is a submanifold of Uz . Indeed, let p ∈ S. Then p ∈ Ux , or p ∈ Uy , or p ∈ Uz . Assume, for
example, that p ∈ Ux . Then, if we know that S ∩ Ux is a submanifold of Ux , it follows that there is an
open subset U 3 p of Ux such that S ∩ U is, in appropriate coordinates, a vector subspace. But U is
also an open subset of RP2 , which means that S satisfies the definition of a submanifold of RP2 near
p. Applying this argument for every p ∈ RP2 , we get that S is a submanifold.
Now we show that S ∩ Ux , S ∩ Uy , S ∩ Uz are sumbanifolds of the corresponding charts. In Ux ,
we have coordinates x1 = y/x, x2 = z/x. The equation xy = z 2 in this chart can be rewritten as
y/x = (z/x)2 , so S ∩ Ux = {(x1 , x2 ) ∈ Ux | x1 = x22 }. It is a graph of a smooth function and hence a
submanifold. Similarly, in Uy we have x1 = x/y, x2 = z/y, and the equation xy = z 2 is equivalent to
x1 = x22 , which is also a submanifold. Finally, in Uz we have x1 = x/z, x2 = y/z, and the equation
xy = z 2 is equivalent to x1 x2 = 1, which is the graph of x2 = 1/x1 and hence a submanifold.
Also note that in fact S ⊂ Ux ∪ Uy . Indeed, the only point of RP2 not belonging to Ux ∪ Uy is
(0 : 0 : 1) ∈
/ S. So, since Ux ∪ Uy cover S, we do not really need to consider Uz .

Solution 2. xy−z 2 is a non-degenerate indefinite quadratic form, so the set of points in R3 satisfy-
ing the equation xy − z 2 = 0 is a cone (see https://en.wikipedia.org/wiki/Quadric#Euclidean_
space). Therefore, there exists a non-zero linear function l(x, y, z) = ax + by + cz such that the
plane l(x, y, z) = 0 intersects the cone xy = z 2 only its vertex (0, 0, 0). For instance, one can take
l(x, y, z) = x + y. Indeed,
( 
(
2
(
2 2
 y = 0, x = 0,

xy = z , −y = z , 
⇔ ⇔ z = 0, ⇔ y = 0,
y = −x y = −x  
y = −x z = 0.
 

Therefore, the set S = {(x : y : z) ∈ RP2 | xy = z 2 } is completely contained in the open set
U = {(x : y : z) ∈ RP2 | l(x, y, z) 6= 0}, and it suffices to show that S is a submanifold of U . To that
end, consider the function on U given by

xy − z 2
f ([x : y : z]) = .
l(x, y, z)2
It is well-defined because the numerator and denominator are of the same degree, and the denominator
does not vanish. Since S is the zero set of f , it remains to show that the differential of f does not
vanish at points of S. To that end, we rewrite f in local coordinates, which is achieved by dividing
both the numerator and denominator by x2 , y 2 , or z 2 (depending on the chart). So, the coordinate
representation of f will be of the form f = a/b, where a = xy − z 2 , divided by the square of one of
the variables, and b = l(x, y, z)2 , divided by the square of the same variable. Therefore,
a b · grad a − a · grad b
grad f = grad = .
b b2
At points of S, we have a = 0, so
a 1
grad f = grad = · grad a.
b b
1
The formula should in fact read {(x : y : z) ∈ RP2 | xy = z 2 }.

3
Possible forms of the function a are x1 − x22 and x1 x2 − 1. Both have non-vanishing gradients on their
zero sets, so S is a submanifold.

Solution 3. The quadratic form xy − z 2 has signature (+, −, −), therefore there is an invertible
linear transformation mapping it to the form −x2 − y 2 + z 2 . This linear transformation can be written
as (x, y, z) 7→ (x, y, z)A, where A is a suitable 3 × 3 matrix. It induces a mapping from RP2 to itself
given by the formula
[x : y : z] 7→ [(x, y, z)A] ,
where [v] on the right-hand side means the point in RP2 corresponding to the line spanned by v ∈ R3 .
(This formula indeed defines a map, because [(λx, λy, λz)A] = [λ(x, y, z)A] = [(x, y, z)A], so the right-
hand side does not depend on the choice of homogeneous coordinates x, y, z of a point in RP2 .) Such
mappings RP2 → RP2 are known as projective transformations. Invertible projective transformations
(which correspond to invertible matrices A) are diffeomorphisms, because in coordinates they are given
by rational functions, and rational functions are smooth wherever they are defined. Furthermore, the
inverse of a projective transformation is also projective and hence smooth. So, there is a diffeomor-
phism of RP2 to itself taking our set to x2 + y 2 = z 2 . The latter is completely contained in the chart
z 6= 0 and is given in the corresponding coordinates by x21 + x22 = 1, which is a submanifold. But
diffeomorphisms take submanifolds to submanifolds, so our initial set is also a submanifold.
This argument also shows that the submanifold in question is diffeomorphic to S 1 .