3

Smooth maps

3.1 Smooth functions on manifolds

A real-valued function on an open subset U ✓ Rn is called smooth if it is infinitely

differentiable. The notion of smooth functions on open subsets of Euclidean spaces
carries over to manifolds: A function is smooth if its expression in local coordinates
is smooth.
Definition 3.1. A function f : M ! Rn on a manifold M is called smooth if for all
charts (U, j) the function
f j 1 : j(U) ! Rn
is smooth. The set of smooth functions into the real-line f : M ! R is denoted C• (M).
One does not have to actually check every single chart. Since transition maps
are diffeomorphisms, it suffices to check the condition for the charts from any given
atlas A = {(Ua , ja )}, which need not be the maximal atlas. Indeed, if the condition
on f holds for all charts from the atlas A , and if (U, j) is another chart compatible
with A , then the functions
1
f j j(U\Ua )
= ( f ja 1 ) (ja j 1
) : j(U \Ua ) ! Rn

are smooth, and since the open sets j(U \Ua ) cover j(U) this implies smoothness
of f j 1 .

Smoothness, like continuity, is a local condition: given an open subset U ✓ M,

we say that a function f is smooth on U if its restriction f |U is smooth. (Here we are
using that U itself is a manifold.) Given p 2 M, we say that f is smooth at p if it is
smooth on some open neighborhood of p.
Example 3.1. The ‘height function’

f : S2 ! R, (x, y, z) 7! z
38 3 Smooth maps

is smooth. In fact, we see that for any smooth function h 2 C• (R3 ) (for example the
coordinate functions), the restriction f = h|S2 is again smooth. This may be checked
using the 6-charts atlas given by projection onto the coordinate planes: E.g., in the
chart U = {(x, y, z)| z > 0} with j(x, y, z) = (x, y), we have
✓ q ◆
( f j 1 )(x, y) = h x, y, 1 (x2 + y2 )

which is smooth on j(U) = {(x, y)| x2 + y2 < 1}. (The argument for the other charts
in this atlas is similar.)
Of course, if h is not a smooth function on R3 , it may still happen that its restric-
tion to S2 is smooth.

Exercise 19. Show that the map

yz + xz + xy
f : RP2 ! R, (x : y : z) 7!
x 2 + y2 + z2
is well-defined and smooth.

Exercise 20. Prove that the map

p
f : S2 ! R, (x, y, z) 7! 1 z2

is smooth only on S2 \{(0, 0, 1), (0, 0, 1)}. To analyse the situation near the
north
p pole, use the coordinate
p p(U, j) as above. In these coordinates, z =
chart
1 (x2 + y2 ), hence 1 z2 = x2 + y2 which is not smooth near (x, y) =
(0, 0).

Example 3.2. Let

p : Rn+1 \{0} ! RPn
be the quotient map. Given f : RPn ! Rm , the function

fb = f p : Rn+1 \{0} ! Rm

satisfies fb(l x) = fb(x) for l 6= 0; conversely any fb with this property descends to
a function f on the projective space. We claim that f is smooth if and only if fb is
smooth. To see this, note that in the standard coordinate chart (Ui , ji ) for RPn , the
function ji 1 may be written as the composition of smooth maps p g, where

g : Rn ◆ Ui ! Rn+1 \{0}, (u1 , . . . , un ) 7! (u1 , . . . , ui , 1, ui+1 , . . . , un ).

Then, f j 1 = fb g. Hence, if fb is smooth then so is f . (The converse is similar.)

As a special case, we see that for all 0  j  k  n the functions

x j xk
f : RPn ! R, (x0 : . . . : xn ) 7! (3.1)
||x||2
 3.1 Smooth functions on manifolds 39

are well-defined and smooth. By a similar argument, the functions

z j zk
f : CPn ! C, (z0 : . . . : zn ) 7! (3.2)
||z||2

(where the bar denotes complex conjugation) are well-defined and smooth, in the
sense that both the real and imaginary parts are smooth.

Exercise 21. In the example above we have used the fact that the quotient map

p : Rn+1 \ {0} ! RPn , (x0 , . . . , xn ) 7! (x0 : . . . : xn )

is smooth. Prove this.

Exercise 22. Show once again that the map

yz + xy + xz
f : RP2 ! R, (x : y : z) 7!
x 2 + y2 + z2
is smooth, this time using the conclusion of Example ?? above.

Lemma 3.1. Smooth functions f : M ! Rn are continuous: For every open subset
J ✓ Rn , the pre-image f 1 (J) ✓ M is open.

Proof. We have to show that for every (U, j), the set j(U \ f 1 (J)) ✓ Rm is open.
But this subset coincides with the pre-image of J under the map f j 1 : j(U) !
Rn , which is a smooth function on an open subset of Rm , and these are (by definition)
continuous.

Exercise 23. We have characterised smooth functions as functions that are

smooth “in charts.” There is a similar characterisation for continuous functions:
Show that f : M ! Rn continuous (i.e., the pre-image of any open subset J ✓ Rn
under f is open) if and only if for all charts (U, j) the function f j 1 is con-
tinuous.
From the properties of smooth functions on Rm , one immediately gets the fol-
lowing properties of smooth functions on manifolds M:
• If f , g 2 C• (M) and l , µ 2 R, then l f + µg 2 C• (M).
• If f , g 2 C• (M), then f g 2 C• (M).
• 1 2 C• (M) (where 1 denotes the constant function p 7! 1).
These properties say that C• (M) is an algebra with unit 1. (See the appendix to this
chapter for some background information on algebras.) Below, we will develop many
of the concepts of manifolds in terms of this algebra of smooth functions.
40 3 Smooth maps

Exercise 24. Prove the assertion that f , g 2 C• (M) =) f g 2 C• (M).

Suppose M is any set with a maximal atlas {(Ua , ja )}. The definition of C• (M)
does not use the Hausdorff or countability conditions; hence it makes sense in this
more general context. This means that we may use functions to check the Hausdorff
property:
Proposition 3.1. Suppose M is any set with a maximal atlas, and p 6= q are two
points in M. Then the following are equivalent:
(i) There are open subsets U,V ✓ M with p 2 U, q 2 V, U \V = 0,
/
(ii) There exists a continuous f : M ! R with f (p) 6= f (q).
(iii) There exists f 2 C• (M) with f (p) 6= f (q).
Proof. Lemma 3.1 shows that (iii) ) (ii). Thus, it suffices to show (i) ) (iii) and
(ii) ) (i).

“(i) ) (iii)”. Suppose (i) holds. As explained in Section 2.5, we may take U,V
to be the domains of coordinate charts (U, j) and (V, y) around p, q. Choose e > 0
such that the closed e-ball

Be (j(p)) = x 2 Rm ||x j(p)||  e

is contained in j(U); let A ✓ U be its pre-image under j. Let c 2 C• (Rm ) be a ‘bump

function’ centered at j(p), with c(j(p)) = 1 and c(x) = 0 for ||x j(p)|| e. (For
the existence of such a function see Munkres, Lemma 16.1, or Lemma A.2 in the
appendix.)

The function f : M ! R such that f = c j on U and f = 0 on M\A is smooth, and

satisfies f (p) = 1, f (q) = 0.

“(ii) ) (i)”. Suppose (ii) holds. Let d = | f (q) f (p)|/2, and put

U = {x 2 M| | f (x) f (p)| < d }, (3.3)

V = {x 2 M| | f (x) f (q)| < d } (3.4)

Thus, U, V are the f -preimages of d -balls centred at f (p) and f (q). Since f is
continuous, U, V are open, and clearly p 2 U, q 2 V , U \V = 0.
/
 3.1 Smooth functions on manifolds 41

A consequence of this result is:

Corollary 3.1 (Criterion for Haudorff condition). A set M with an atlas satisfies
the Hausdorff condition if and only if for any two distinct points p, q 2 M, there exists
a continuous function f : M ! R with f (p) 6= f (q). In particular, if there exists a
continuous injective map F : M ! RN , then M is Hausdorff.

Exercise 25. Prove the last assertion: if there exists a smooth injective map
F : M ! RN , then M is Hausdorff.

Example 3.3 (Projective spaces). Write vectors x 2 Rn+1 as column vectors, hence
x> is the corresponding row vector. The matrix product xx> is a square matrix with
entries x j xk . The map
x x>
RPn ! MatR (n + 1), (x0 : . . . : xn ) 7! (3.5)
||x||2
is a smooth; indeed, its matrix components are the functions (3.1). For any given (x0 :
. . . : xn ) 2 RPn , at least one of these components is non-zero. Identifying MatR (n +
1) ⇠ = RN , where N = (n + 1)2 , this gives the desired smooth injective map from
projective space into RN ; hence the criterion applies, and the Hausdorff condition
follows. For the complex projective space, one similarly has a smooth and injective
map
z z†
CPn ! MatC (n + 1), (z0 : . . . : zn ) 7! (3.6)
||z||2
(where z† = z> is the conjugate transpose of the complex column vector z) into
MatC (n + 1) = RN with N = 2(n + 1)2 .

Exercise 26. Verify that the map

Gr(k, n) ! MatR (n), E 7! PE , (3.7)

taking a subspace E to the matrix of the orthogonal projection onto E, is smooth

and injective, hence Gr(k, n) is Hausdorff. Discuss a similar map for the com-
plex Grassmannian GrC (k, n).

In the opposite direction, the criterion tells us that for a set M with an atlas, if the
Hausdorff condition does not hold then no smooth injective map into RN exists.
Example 3.4. Consider the non-Hausdorff manifold M from Example 2.6. Here,
there are two points p, q that do not admit disjoint open neighborhoods, and we
see directly that any smooth function on M must take on the same values at p and q:
With the coordinate charts (U, j), (V, y) in that example,
1 1 1 1
f (p) = f (j (0)) = lim f (j (t)) = lim f (y (t)) = f (y (0)) = f (q),
t!0 t!0

since j 1 (t) = y 1 (t) for t < 0.

42 3 Smooth maps

3.2 Smooth maps between manifolds

The notion of smooth maps from M to Rn generalizes to smooth maps between
manifolds.
Definition 3.2. A map F : M ! N between manifolds is smooth at p 2 M if there are
coordinate charts (U, j) around p and (V, y) around F(p) such that F(U) ✓ V and
such that the composition
1
y F j : j(U) ! y(V )

is smooth. The function F is called a smooth map from M to N if it is smooth at all

p 2 M.

As before, to check smoothness of F, it suffices to take any atlas {(Ua , ja )} of

M with the property that F(Ua ) ✓ Va for some chart (Va , ya ) of N, and then check
smoothness of the maps

ya F ja 1 : ja (Ua ) ! ya (Va ).

This is because the condition for smoothness at p does not depend on the choice of
charts: Given a different choice of charts (U 0 , j 0 ) and (V 0 , y 0 ) with F(U 0 ) ✓ V 0 , we
have
y 0 F (j 0 ) 1 = (y 0 y 1 ) (y F (j) 1 ) (j (j 0 ) 1 )
on j 0 (U \U 0 ). This demonstrates once again that the requirement that the transition
maps be diffeomorphisms is the correct idea.

The collection of smooth maps f : M ! N is denoted C• (M, N). Note that since
R itself is a manifold, we now have two definition for a smooth map f : M ! R. You
should convince yourself that the two definitions coincide:
 3.2 Smooth maps between manifolds 43

C• (M, R) = C• (M).

Smooth functions g : J ! M from an open interval J ✓ R to M are called

(smooth) curves in M. Note that the image of a smooth curve need not look smooth.
For instance, the image of g : R ! R2 , t 7! (t 2 ,t 3 ) has a ‘cusp singularity’ at (0, 0).

Exercise 27. Suppose F 2 C• (M, N).

1. Let (U, j) be a coordinate chart for M and (V, y) a coordinate chart for N,
with F(U) ✓ V . Show that for all open subsets W ✓ N the set U \ F 1 (W )
is open.
2. Given p 2 M and any chart (V, y) around F(p), show that there exists a
chart (U, j) around p such that F(U) ✓ V .

Proposition 3.2. Smooth maps between manifolds are continuous.

Exercise 28. Use the previous exercise to prove the proposition above.

Proposition 3.3. Suppose F1 : M1 ! M2 and F2 : M2 ! M3 are smooth maps. Then

the composition
F2 F1 : M1 ! M3
is smooth.

Exercise 29. Prove the proposition.

Hint: Use the previous exercise.

3.2.1 Diffeomorphisms of manifolds

Definition 3.3. A smooth map F : M ! N is called a diffeomorphism if it is invert-

ible, with a smooth inverse F 1 : N ! M. Manifolds M, N are called diffeomorphic
if there exists a diffeomorphism from M to N.
44 3 Smooth maps

In other words, a diffeomorphism of manifolds is a bijection of the underlying sets

that identifies the maximal atlases of the manifolds. Manifolds that are diffeomorphic
are therefore considered ‘the same manifold’.
Similarly, a continuous map F : M ! N is called a homeomorphism if it is invert-
ible, with a continuous inverse. Manifolds that are homeomorphic are considered ‘the
same topologically’. Since every smooth map is continuous, every diffeomorphism
is a homeomorphism.

Example 3.5. By definition, every coordinate chart (U, j) on a manifold M gives a

diffeomorphism j : U ! j(U) onto an open subset of Rm .

Example 3.6. The standard example of a homeomorphism of smooth manifolds that

is not a diffeomorphism is the map

R ! R, x 7! x3 .
1
Indeed, this map is smooth and invertible, but the inverse map y 7! y 3 is not smooth.

Example 3.7. Give a manifold M with maximal atlas A , any homeomorphism F :

M ! M can be used to define a new atlas A 0 on M, with charts (U 0 , j 0 ) 2 A 0 ob-
tained from charts (U, j) 2 A as U 0 = F(U), j 0 = j F 1 . One can verify (please
do) that A 0 = A if and only if F is a diffeomorphism. Thus, if F is a homeomor-
phism of M which is not a diffeomorphism, then F defines a new atlas A 0 6= A .
However, the new manifold structure on M is not genuinely different from the
old one. Indeed, while F : M ! M is not a diffeomorphism relative to the atlas A
on the domain M and target M, it does define a diffeomorphism if we use the atlas
A on the domain and the atlas A 0 on the target. Hence, even though A and A 0 are
different atlases, the resulting manifold structures are still diffeomorphic.

Exercise 30. Consider R with the trivial atlas A := (R, id). The homeomor-
phism from Example 3.6 defines a new atlas A 0 := (R,t 7! t 3 ).
1. Show that R equipped with the atlas A 0 is a 1-dimensional manifold.
2. Show that the maximal atlases generated by A and A 0 are different. Thus,
we get two distinct manifolds M = (R, A ) and M 0 = (R, A 0 ).
an atlas.
Hint: Recall that this mean that the union of these two atlases does not form
3. Show that the map f : M ! M 0 given by f (x) = x1/3 is a diffeomorphism.
Reread Example 3.7 with this concrete example in mind.

Remark 3.1. In the introduction, we explained (without proof) the classification of 1-

dimensional and 2-dimensional connected compact manifolds up to diffeomorphism.
This classification coincides with their classification up to homeomorphism. This
means, for example, that for any maximal atlas A 0 on S2 which induces the same
system of open subsets as the standard maximal atlas A , there exists a homeomor-
phism F : S2 ! S2 taking A to A 0 , in the sense that (U, j) 2 A if and only if
 3.3 Examples of smooth maps 45

(U 0 , j 0 ) 2 A 0 , where U 0 = F(U) and j 0 F = j. In higher dimensions, it becomes

much more complicated: .
It is quite possible for two manifolds to be homeomorphic but not diffeomor-
phic (unlike example 3.7). The first example of ‘exotic’ manifold structures was
discovered by John Milnor in 1956, who found that the 7-sphere S7 admits manifold
structures that are not diffeomorphic to the standard manifold structure, but induce
the standard topology. Kervaire and Milnor in 1963, proved that there are exactly 28
distinct manifold structures on S7 , and in fact classified all manifold structures on
all spheres Sn with the exception of the case n = 4. For example, they showed that
S3 , S5 , S6 do not admit exotic (i.e., non-standard) manifold structures, while S15 has
16256 different manifold structures. For S4 the existence of exotic manifold struc-
tures is an open problem; this is known as the smooth Poincare conjecture.
Around 1982, Michael Freedman (using results of Simon Donaldson) discovered
the existence of exotic manifold structures on R4 ; later Clifford Taubes showed that
there are uncountably many such. For Rn with n 6= 4, it is known that there are no
exotic manifold structures on Rn .

3.3 Examples of smooth maps

3.3.1 Products, diagonal maps

a) If M, N are manifolds, then the projection maps

prM : M ⇥ N ! M, prN : M ⇥ N ! N

are smooth. (This follows immediately by taking product charts Ua ⇥Vb .)

b) The diagonal inclusion
DM : M ! M ⇥ M
is smooth. (In a coordinate chart (U, j) around p and the chart (U ⇥ U, j ⇥ j)
around (p, p), the map is the restriction to j(U) ✓ Rn of the diagonal inclusion
Rn ! Rn ⇥ Rn .)
c) Suppose F : M ! N and F 0 : M 0 ! N 0 are smooth maps. Then the direct product

F ⇥ F 0 : M ⇥ M0 ! N ⇥ N 0

is smooth. This follows from the analogous statement for smooth maps on open
subsets of Euclidean spaces.

3.3.2 The diffeomorphism RP1 ⇠

= S1

We have stated before that RP1 ⇠ = S1 . To obtain an explicit diffeomorphism, we con-

struct a bijection identifying the standard atlas for RP1 with (essentially) the standard
atlas for S1 . Recall that the atlas for RP1 is given by
46 3 Smooth maps

U1 = {(u : 1)| u 2 R}, j1 (u : 1) = u,

U0 = {(1 : u)| u 2 R}, j0 (1 : u) = u

with ji (Ui ) = R and j0 (U0 \ U1 ) = j1 (U0 \ U1 ) = R\{0}, with the transition map
j1 j0 1 : u 7! u 1 . Similarly, the atlas for S1 is
x
U+ = {(x, y) 2 S1 | y 6= 1} j+ (x, y) = ,
1+y
x
U = {(x, y) 2 S1 | y 6= +1} j (x, y) = .
1 y

again with j± (U± ) = R, j± (U+ \U ) = R\{0}, and transition map u 7! u 1 .

Hence, there is a well-defined diffeomorphism F : RP1 ! S1 which identifies
the chart (U , j ) with (U1 , j1 ) and (U+ , j+ ) with (U0 , j0 ), in the sense that both

j F j1 1 : R ! R, j+ F j0 1 : R ! R

are the identity idR . Namely, the restriction of F to U1 is FU1 = j 1 j1 : U1 ! U ,

the restriction to U0 is F|U0 = j+ 1 j0 : U0 ! U+ . The inverse map G = F 1 : S1 !
RP1 is similarly given by j0 1 j+ over U+ and by j1 1 j over U .

Exercise 31. Calculate the diffeomorphism F : RP1 ! S1 and its inverse G :

S1 ! RP1 (give explicit formulas).

3.3.3 The diffeomorphism CP1 ⇠

= S2

By a similar reasoning, we find CP1 ⇠

= S2 . For S2 we use the atlas given by stereo-
graphic projection.
1
U+ = {(x, y, z) 2 S2 | z 6= 1} j+ (x, y, z) = (x, y),
1+z
1
U = {(x, y, z) 2 S2 | z 6= +1} j (x, y, z) = (x, y).
1 z
u
The transition map is u 7! ||u||2
, for u = (u1 , u2 ). Regarding u as a complex number
u = u1 + iu2 , the norm ||u|| is just the absolute value of u, and the transition map
becomes
u 1
u 7! 2 = .
|u| u
Note that it is not quite the same as the transition map for the standard atlas of CP1 ,
which is given by u 7! u 1 . We obtain a unique diffeomorphism F : CP1 ! S2 such
that j+ F j0 1 is the identity, while j F j1 1 is complex conjugation.
 3.3 Examples of smooth maps 47

Exercise 32. Calculate the diffeomorphism F : CP1 ! S2 and its inverse G :

S2 ! CP1 .

3.3.4 Maps to and from projective space

We can now generalize Example 3.2. In Exercise 21 you verified that the quotient
map
p : Rn+1 \{0} ! RPn , x = (x0 , . . . , xn ) 7! (x0 : . . . : xn )
is smooth. Given a map F : RPn ! N to a manifold N, let Fe = F p : Rn+1 \{0} ! N
be its composition with the projection map p : Rn+1 \{0} ! RPn . That is,
e 0 , . . . , xn ) = F(x0 : . . . : xn ).
F(x
e x0 : . . . : l xn ) = F(x
Note that F(l e 0 , . . . , xn ) for all non-zero l ; conversely, every map
Fe with this property descends to a map F on projective space. We claim that the map
F is smooth if and only the corresponding map Fe is smooth. One direction is clear: If
F is smooth, then Fe = F p is a composition of smooth maps. For the other direction,
assuming that Fe is smooth, note that for the standard chart (U j , j j ), the maps

(F j j 1 )(u1 , . . . , un ) = F(u
e 1 , . . . , ui , 1, ui+1 , . . . , un ),

are smooth.
An analogous argument applies to the complex projective space CPn , taking the
x to be complex numbers zi . That is, the quotient map p : Cn+1 \{0} ! CPn is
i

smooth, and a map F : CPn ! N is smooth if and only if the corresponding map
Fe : Cn+1 \{0} ! N is smooth.

Exercise 33. The argument above demonstrates how to lift maps from projec-
tive spaces; show how we can use a similar technique to lift maps whose target
is projective space.

As an application, we can see that the map

CP1 ! CP2 , (z0 : z1 ) 7! ((z0 )2 : (z1 )2 : z0 z1 )

is smooth, starting with the (obvious) fact that the lifted map

C2 \{0} ! C3 \{0}, (z0 , z1 ) 7! ((z0 )2 , (z1 )2 , z0 z1 )

is smooth.

3.3.5 The quotient map S2n+1 ! CPn

As we explained above, the quotient map q : Cn+1 \{0} ! CPn is smooth. Since any
class [z] = (z0 : . . . : zn ) has a representative with |z0 |2 + · · · + |zn |2 = 1, and |zi |2 =
48 3 Smooth maps
p
(xi )2 + (yi )2 for zi = xi + 1yi , we may also regard CPn as a set of equivalence
classes in the unit sphere S 2n+1 ✓ R2n+2 = Cn+1 . The resulting quotient map

p : S2n+1 ! CPn

is again smooth, because it can be written as a composition of two smooth maps

p = q i where i : S2n+1 ! R2n+2 \{0} = Cn+1 \{0} is the inclusion map.
For any p 2 CPn , the corresponding fiber p 1 (p) ✓ S2n+1 is diffeomorphic to a
circle S1 (which we may regard as complex numbers of absolute value 1). Indeed,
given any point (z0 , . . . , zn ) 2 p 1 (p) in the fiber, the other points are obtained as
(l z0 , . . . , l zn ) where |l | = 1.
In other words, we can think of
[
S2n+1 = p 1
(p)
p2CPn

as a union of circles, parametrized by the points of CPn . This is an example of what

differential geometers call a fiber bundle or fibration. We won’t give a formal defini-
tion here, but let us try to ‘visualize’ the fibration for the important case n = 1.
Identifying CP1 ⇠ = S2 as above, the map p becomes a smooth map
p : S3 ! S2

with fibers diffeomorphic to S1 . This map appears in many contexts; it is called the
Hopf fibration (after Heinz Hopf (1894-1971)).

Let S 2 S3 be the ‘south pole’, and N 2 S3 the ‘north pole’. We have that S3 {S} ⇠ =
R3 by stereographic projection. The set p 1 (p(S)) {S} projects to a straight line
(think of it as a circle with ‘infinite radius’). The fiber p 1 (N) is a circle that goes
around the straight line. If Z ✓ S2 is a circle at a given ‘latitude’, then p 1 (Z) is
is a 2-torus. For Z close to N this 2-torus is very thin, while for Z approaching the
south pole S the radius goes to infinity. Each such 2-torus is itself a union of circles
p 1 (p), p 2 Z. Those circles are neither the usual ‘vertical’ or ‘horizontal’ circles of
a 2-torus in R3 , but instead are ‘tilted’. In fact, each such circle is a ‘perfect geometric
circle’ obtained as the intersection of its 2-torus with a carefully positioned affine 2-
plane.
Moreover, any two of the circles p 1 (p) are linked:
 3.3 Examples of smooth maps 49

The full picture looks as follows:

A calculation shows that over the charts U+ ,U (from stereographic projection), the
Hopf fibration is just a product. That is, one has
1
p = U+ ⇥ S1 , p
(U+ ) ⇠ 1
= U ⇥ S1 .
(U ) ⇠

In particular, the pre-image of the closed upper hemisphere is a solid 2-torus D2 ⇥ S1

(with D2 = {z 2 C| |z|  1} the unit disk), geometrically depicted as a 2-torus in R3
together with its interior.⇤ We hence see that the S3 may be obtained by gluing two
solid 2-tori along their boundaries S1 ⇥ S1 .

⇤ A solid torus is an example of a ”manifold with boundary”, a concept we haven’t properly

discussed yet.