Selected Solutions to Munkres’s Topology, 2nd Ed.

Takumi Murayama
December 20, 2014

These solutions are the result of taking MAT365 Topology in the Fall of 2012 at
Princeton
Princeton Universit y. This is not a  complete  set
University  complete  set of solutions; see the  List of Solved
Exercises at
Exercises  at the end. Please e-mail   [email protected] with any corrections.

Contents

I Gene
Genera
rall Topolo
opology
gy 3
1 Set
Set Theo
Theory
ry and
and LLog
ogic
ic 3
7 Coun
Counta
tabl
blee and
and Un
Unco
coun
unta
tabl
blee Sets
Sets   . . . . . . . . . . . . . . . . . . . . 3

2 Topological
opological Spaces and ContinContinuousuous Functi
Functions
ons 3
13 Basi
Basiss for
for a Topolo
opology
gy . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
16 Thee Su
Th Subs
bspa
pace
ce Topolo
opology
gy   . . . . . . . . . . . . . . . . . . . . . . . . . 5
17 Clos
Closed
ed Sets
Sets and
and Limi
Limitt Po
Poinints
ts   . . . . . . . . . . . . . . . . . . . . . . 6
18 Cont
Contininuo
uous
us Fun
unct
ctio
ions
ns   . . . . . . . . . . . . . . . . . . . . . . . . . . 8
19 Thee Produ
Th Product ct Topolo
opology
gy . . . . . . . . . . . . . . . . . . . . . . . . . 9
20 Thee Metr
Th Metricic Topol
opolog
ogyy . . . . . . . . . . . . . . . . . . . . . . . . . . 10
21 The Metri
Metricc Topology
opology (cont
(contininued
ued)) . . . . . . . . . . . . . . . . . . . . 15
22 Thee Quot
Th Quotieien
nt Topolog
opologyy . . . . . . . . . . . . . . . . . . . . . . . . . 18

3 Connec
Connected
tednes
nesss and Compact
Compactnes nesss 20
23 Conn
Co nnec
ecte
tedd Sp
Spac
aces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
24 Conn
Co nnec
ecte
tedd Su
Subs
bspa
pace
cess of the
the Real
Real Line
Line   . . . . . . . . . . . . . . . . . 21
25 Compone
Component ntss and Local
Local Con
Connec
nected
tednes
nesss   . . . . . . . . . . . . . . . . . 26
27 Comp
Co mpacactt Su
Subsbspac
paces
es of the
the Real
Real Line
Line   . . . . . . . . . . . . . . . . . . 27
29 Loca
Locall Co
Compmpac actn
tnes
esss . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

1
4 Countabil
Countabilit
ity
y and Separati
Separation on Axioms
Axioms 28
30 Thee Co
Th Coun
unta
tabi
bili
litty Axio
Axiomsms   . . . . . . . . . . . . . . . . . . . . . . . . 28
31 Thee Sepa
Th Separa
rati
tion
on Axio
Axiomsms   . . . . . . . . . . . . . . . . . . . . . . . . . 30
32 Norma
ormall Sp
Spac
acees   . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
33 The Urys
Urysoh
ohnn Lemm
Lemmaa . . . . . . . . . . . . . . . . . . . . . . . . . . 32
34 Thee Urys
Th Urysoh
ohnn Metr
Metriz izat
atio
ion
n Th
Theo
eore
rem
m. . . . . . . . . . . . . . . . . . . 32
36 Imbed
Imbeddiding
ngss of Mani
Manifo fold
ldss . . . . . . . . . . . . . . . . . . . . . . . . 33

II Alge
Algebr
brai
aicc Topolo
opology
gy 35
9 The Fund
Fundame
ament ntal
al Group
Group 35
51 Homo
Homototoppy of Paths
aths . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
52 Thee Fun
Th undadame
ment
ntal
al Grou
Groupp . . . . . . . . . . . . . . . . . . . . . . . . 36
53 Cov
Coverin
eringg Sp
Spac
aces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
54 Thee Fun
Th Fundadame
ment
ntal
al Grou
Groupp of
of the
the Circ
Circle
le   . . . . . . . . . . . . . . . . . 38
58 Deform
Deformati
ation
on Retrac
Retracts
ts and Homoto
Homotopy py Type
Type . . . . . . . . . . . . . . 39
59 Thee Fun
Th undadame
ment
ntal
al Grou
Groupp of  S  . . . . . . .
n
. . . . . . . . . . . . . . 43
60 Fund
undame
ament ntal
al Groups
Groups of Some
Some Surfac
Surfaces
es . . . . . . . . . . . . . . . . . 44

11 The Seifert-van Kampen Theorem 45

67 Dire
Direct
ct Su
Sums
ms of Abeli
Abelian
an Grou
Groups
ps . . . . . . . . . . . . . . . . . . . . . 45
68 Free
ree Produ
Product
ctss of Group
Groupss . . . . . . . . . . . . . . . . . . . . . . . . 45
71 The Fund
undame
amentntal
al Group
Group of a Wedg
Wedgee of
of Circ
Circles
les . . . . . . . . . . . . 46
73 The Fund
undame
amentntal
al Grou
Groups
ps of
of the
the Toru
Toruss and
and the
the Dunce
Dunce Cap . . . . . 48

12 Classification of Surfaces 49
74 Fund
undame
ament
ntal
al Groups
Groups of Surfac
Surfaces
es   . . . . . . . . . . . . . . . . . . . . 49
75 Homo
Homolology
gy of Su
Surf
rfac
aces
es . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2
4 Countabil
Countabilit
ity
y and Separati
Separation on Axioms
Axioms 28
30 Thee Co
Th Coun
unta
tabi
bili
litty Axio
Axiomsms   . . . . . . . . . . . . . . . . . . . . . . . . 28
31 Thee Sepa
Th Separa
rati
tion
on Axio
Axiomsms   . . . . . . . . . . . . . . . . . . . . . . . . . 30
32 Norma
ormall Sp
Spac
acees   . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
33 The Urys
Urysoh
ohnn Lemm
Lemmaa . . . . . . . . . . . . . . . . . . . . . . . . . . 32
34 Thee Urys
Th Urysoh
ohnn Metr
Metriz izat
atio
ion
n Th
Theo
eore
rem
m. . . . . . . . . . . . . . . . . . . 32
36 Imbed
Imbeddiding
ngss of Mani
Manifo fold
ldss . . . . . . . . . . . . . . . . . . . . . . . . 33

II Alge
Algebr
brai
aicc Topolo
opology
gy 35
9 The Fund
Fundame
ament ntal
al Group
Group 35
51 Homo
Homototoppy of Paths
aths . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
52 Thee Fun
Th undadame
ment
ntal
al Grou
Groupp . . . . . . . . . . . . . . . . . . . . . . . . 36
53 Cov
Coverin
eringg Sp
Spac
aces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
54 Thee Fun
Th Fundadame
ment
ntal
al Grou
Groupp of
of the
the Circ
Circle
le   . . . . . . . . . . . . . . . . . 38
58 Deform
Deformati
ation
on Retrac
Retracts
ts and Homoto
Homotopy py Type
Type . . . . . . . . . . . . . . 39
59 Thee Fun
Th undadame
ment
ntal
al Grou
Groupp of  S  . . . . . . .
n
. . . . . . . . . . . . . . 43
60 Fund
undame
ament ntal
al Groups
Groups of Some
Some Surfac
Surfaces
es . . . . . . . . . . . . . . . . . 44

11 The Seifert-van Kampen Theorem 45

67 Dire
Direct
ct Su
Sums
ms of Abeli
Abelian
an Grou
Groups
ps . . . . . . . . . . . . . . . . . . . . . 45
68 Free
ree Produ
Product
ctss of Group
Groupss . . . . . . . . . . . . . . . . . . . . . . . . 45
71 The Fund
undame
amentntal
al Group
Group of a Wedg
Wedgee of
of Circ
Circles
les . . . . . . . . . . . . 46
73 The Fund
undame
amentntal
al Grou
Groups
ps of
of the
the Toru
Toruss and
and the
the Dunce
Dunce Cap . . . . . 48

12 Classification of Surfaces 49
74 Fund
undame
ament
ntal
al Groups
Groups of Surfac
Surfaces
es   . . . . . . . . . . . . . . . . . . . . 49
75 Homo
Homolology
gy of Su
Surf
rfac
aces
es . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2
Part I
General Topology
1 Set
Set Theor
heory
y an
and
d Logi
Logicc
7 Coun
Countable
table and
and Uncou
Uncoun
ntable
table Sets
Sets
Exercise 7.5.  Determine, for each of the following sets, whether or not it is count-
able. Justify your answers.
J  of all finite subsets of  Z+ .
 j )   The set  J  of
( j)
Solution for  ( j)
 j ).   We claim J 
claim  J    is countable. Consider I 

Consider  I  = n∞=0 I n  where I 
 where  I n  is the set
of sequences with n
with  n elements.
 elements. Each I 
Each  I n  is countable by Theorem 7. 7 .6 so I 
so  I  is
 is countable
by Theorem 7. 7.5. Ident
Identify
ifying
ing each
each finite
finite subset
subset in J  
J   with the finite sequence with
the same elements in increasing order, we see that J   I ,  I , and so J  is
 ⊆
 ⊆ J  is countable by
Corollary 7.
7.3.

2 Topologic
opological
al Spaces
Spaces and
and Contin
Continuou
uouss Func
Functio
tions
ns
13 Basi
Basiss for
for a Top
Topol
olog
ogy
y
Exercise 13.3.  Show that the collection   T   given in Example  4 of  § §12 is
 T  c 12  is a topology 
on the set  X . Is the collection 
T   = {U  | |  X  \ \ U  is
∞ U  is infinite or empty or all of  X }
a topology on  X ? 
? 
Proof.   Recall
Recall Example 12.4: 12.4: Let X  be  T 
X  be a set; let c   be the collection of all subsets
U  of  X  such  \ \
X  such that X  U  eitherU  either is countable or is all of  X . We claim claim this forms
forms a
topology on X 
on X ;; we will follow the numbering for the definition of a topology on p. 76.
 
(1) X   \ \ ∅  \ \  ∅
= X  and X  X  =  is countable; (2) X   \ \
α∈A U α =  \ \
α∈A (X  U α ) is
countable since it is an intersection of countable sets, unless every U α  = , in which
 ∅
case X   \ \
α∈A U α = X ; (3) X   \ \ 
α∈A   finite U α =
  \ \
α∈A   finite (X   U α ) is countable
since it is the finite union of countable sets, unless every U α = , in which case ∅
X   \ \ 
α∈A   finite U α  =  X .
 X .
 T 
Now consider ∞ . It is not a topol topolog
ogy y, for
for if we let X  = [ 1, 1]−  ⊆ R, and let
−
U 1  = [ 1, 0) and U 
and  U 2  = (0,
(0, 1], we see that both  U 1 , U 2  ∈ T 
∞ , but X  \ ∪
but  X  (U 1 U 2 ) = 0 , {}
which is not infinite, and so U  ∪ ∈ T 
so  U 1 U 2 / ∞ .

3
Exercise
Exercise 13.5.
13.5.   Show that if    is a basis for a topology on  X , then the topology 
 A
 A
  AA
generated by   equals the intersection of all topologies on  X  that  A
 A
X  that contain  . Prove 
 A
 A
the same if   is a subbasis.
Proof. Let  T    be the topology generated by  A , and  T    be the intersection of all
A I 
I 
topologies that contain A.
T  ⊆ T  . This follows from the fact that  T   ⊇ A, and so is one of the topologies
I 
I  A A
that is intersected over in the construction of  T  . I 
I 

T   ⊆ T  . Let U  ∈
A I 
I   ∈ T  ; by Lemma 13.
 A 13.1, U  = A  for some collection  {A }  ⊆ α α α α
A. But U 
But  U  = A  ∈ T   since each A
α α each  A  ∈ T  .
I 
I  α I 
I 
Now let  A   be a subb
subbas
asis
is.. Th that T  ⊆ T    is identical; it remains to
Thee proof that I 
I  A
show T   ⊆ T  . Let U  ∈
A I 
I  ∈ T  ; by definition of the topology generated by  A, U  is
A U  is the
union of a finite intersection of elements  {A }  ⊆ A . Bu then U  ∈
Butt then  ∈ T    since each
α α I 
I 
A  ∈ T  .
α I 
I 

Exercise 13.6.  Show that the topologies of  R and  RK  are not comparable.
Proof. R ⊆ RK . For [a,
[a, b) R , there is no basis element U  RK    such that
∈ ∈
∈
a  U, U   [a, ⊆
 ⊆
 [ a, b).
RK  R . For ( 1, 1)  K  RK  which contains 0, there is no basis element
⊆ −  \ ∈
[a, b) R  such that 0 [a, b), [a, b)
 ∈  ∈ ( 1, 1)  K   ⊆ − \
 K    by the Archimedean property,
all   > 0, there exists N  N  such that 1/N
that is, for all   ∈
 ∈
1 /N < . .
Exercise 13.7.   Consider the following topologies on  R:
T   = the
1  =  the standard topology,
2  =  the topology of  R ,
T   = the K 
T   = the
3  =  the finite complement topology,
T   = the
4  =  the upper limit topology, having all sets  (a, b]  as basis,
T   = the
5  =  the topology having all sets  (−∞, a) = {x |  x < a}  as basis.
Determine, for each of these topologies, which of the others it contains.
Proof.  We claim we have the following Hasse diagram:
T 
4

T 
2

T 
1

T 
3 T  5

4
 c c
T   T  . Inclus
n
3 1 Inclusion
ion is true since since U  ∈ T  =⇒ U  finite, and so if we let U  =
 n
3
{x }  with x  in increasing order, U  = (x , x ), where x = −∞
i i=1 i i=0 i i+1  −∞, x = ∞. 0 n+1
Inequality follows since for (a, ( a, b) such that −∞ < a, a, b < ∞,  R \ (a, b) is not finite.
T   T . Inclus
5 1 Inclusion
ion is clear since since ( −∞, a) is of the form (b, ( b, c). Inequalit
Inequality y follows
since for (b,( b, c) ∈ T  and x ∈ (b,
1  ( b, c), there is no basis element ( −∞, a) ∈ T   such that 5
x ∈  (−∞, a), (−∞, a) ⊆  (b,  b > −∞).
 ( b, c) (if  b
T  and T   are not comparable.  T   ⊆ T   since  R \ {0} ∈ T , but if we take x > 0,
3 5 3 5 3
which is in this set, there is no basis element ( −∞, a) ∈ T    that contains x  but is 5
c
contained in  R \ {0}.  T    ⊆ T   since (−∞, 0) is not finite.
5 3
T    T   by Lemma 13.
1 2 13.4.
T    T  . For (a,
2 4 (a, b)  ∈ T  and x  ∈ (a, b), (a, x]  ∈ T    and (a,
2 (a, x]  ⊆ (a, b). For
4
(a, b) \ K  ∈ ∈ T  and  x ∈  (a,
and x2  (a, b) \ K , we note that  x ∈  (1/  (1 /(n + 1),
1), c] where x
where  x < c < 1
<  1/n
/n,,
x ∈ (a,
 ( a, 0], or x ∈  (1,
 (1 , d], where x
where  x < d < b; b ; in all three cases, these sets are subsets of 
(a, b) \ K  and
  and are members of  T   T . Inequality follows since for (a,
4 ( a, b] ∈ T  , there is no 4
basis element U  ∈  ∈ T   such that b
2that  b ∈  U, U  ⊂
 ⊂ (a,
 ( a, b].

16 The Su
Subsp
bspace
ace Topolog
opology
y
Exercise 16.8. If  L  is a straight line in the plane, describe the topology  L  inherits 
as a subspace of  R R  and as a subspace of  R R . In each case it is a familiar 
× ×
topology.
Solution.  Note that the basis for R R consists of elements of the form [a,
× [ a, b) (c, d). ×
If  L = { }
 L  = (x, y)  x =
 x  = x∩ |
 x 0 , then L × ∅ { }×
then  L [a, b) (c, d) = or x0 (c, d), and so defining defining the
map ϕ :  L (R R)
map ϕ ∩ × → { }× R, x0 (c, d)  (c, →
 (c, d), it is bijective, open, and continuous,
and so the topology L  inherits is homeomorphic to R with the standard topology. If 
topology  L inherits
L  has finite slope, we first note that L (R R) = (x,mx + x,mx + b  b)) R2 x R ,
∩  ×  {  ∈ |  ∈ }
and that the basis for our topology are the sets of the form , [(a,ma [(a,ma +
 + b
 b)), (c,mc +
c,mc +  ∅
b)),
)), ((a,ma for  a, c R  and  a < c,
((a,ma + b), (c,mc + b)) for a, ∈
c , by Lemma 16.16.1. We then define

ϕ: L ∩ (R × R) → R ,
  (a,ma +
a,ma + b
 b)) → a.
This implies

((a,ma
((a,ma +
 + b
 b)), (c,mc +
c,mc + b ))
 b)) → (a,
 ( a, c),
[(a,ma
[(a,ma +
 + b
 b)), (c,mc +
c,mc + b ))
 b)) → [a,
 [ a, c).

We claim this defines a homeomorphism with R . ClearlyClearly,, it is continuous,

continuous, for the
basis elements of  R   have preimages that are basis elements in the topology on L.
of  L  map to sets that are open in  R  by
Likewise, it is open since the basis elements of  L

5
Lemma 13.4. Finally this is a bijection since there exists an inverse just by reversing
the arrows above.
For  R  R , by following the same steps as above if  L = (x, y)  x = x 0 , then
×  {  | }
L (R R ) is homeomorphic to  R  . For L  with m  < , we must split it up into
∩ ×  | | ∞
 ≥
two cases. When m  0, we have a similar situation as above, except we only have
to consider basis elements of the form [ a, b); thus, L (R R ) is homeomorphic
∩  ×
to R . When m < 0, since for every point (x, y) L, we can find a basis element
 ∈
[x, a) [y, b)  ( R R ) such that L [x, a) [y, b) = (x, y) , and these form the
× ∈ × ∩ × { }
open sets of our new topology by Lemma 16.1. We see then that the topology on  L
is homeomorphic to the discrete topology on  R .

Exercise 16.9.  Show that the dictionary order topology on the set  R R is the same 
×
as the product topology  Rd R, where  Rd  denotes  R in the discrete topology. Compare 
×
this topology with the standard topology on  R2 .

Proof.  We see that the basis elements of ( R R)lex  consist of intervals of the form
×
× ×
(a b, c d) for a < c, and for a = c and b < d, as in Example 14.2. These basis
elements are open in  R d R  since×
× b, c × d) = (a, c) × R ∪ {a} × (c, ∞) ∪ {b} × (−∞, d) ∈ T R R.
(a d×

For the reverse situation, consider the basis elements for R  × R; these consist of 
d
all {a} × (b, c) since {a | a ∈ R}  forms a basis for R  by Example 13.3. But then,
d
{a}× (b, c) are open in R×R with the order topology since it is of the form (a×b, c×d)
for a = c.
2
We now compare this to the standard topology on R . Since (a, b)×(c, d) ∈ R ×R, d
2 2
we see that  R ⊆ R × R. Moreover, since  {a} × (b, c) ∈ ( R × R) \ R , we see that
d d
2
R  R × R. d

17 Closed Sets and Limit Points

Exercise 17.2.  Show that if  A  is closed in  Y  and  Y   is closed in  X , then  A  is closed 
in  X .

Proof. A is closed in Y  iff there exists  B  X  closed in X  such that A  = Y  B by
 ⊆  ∩
Theorem 17.2. But then, A is the intersection of closed sets, and so is closed.

Exercise 17.3.  Show that if  A is closed in  X  and  B  is closed in  Y   , then  A × B is 
closed in  X  Y  .  ×

6
 \  \
Proof.  We see that X  A, Y  B are open in X, Y  respectively by definition of a closed
\ × × \ ×
set. By definition of the product topology, (X  A) Y, X  (Y  B) are open in X  Y  .
 \ ×  × \ ×  ×  \  × \  ×
We see that (X  A) Y  = (X  Y  ) (A Y  ), X  (Y  B) = (X  Y  ) (X  B),
× × × × × ∩ ×
and so A Y, X  B are closed in X  Y   . Finally, A B = (A Y  ) (X  B), and
so is the intersection of closed sets, i.e., closed.

Exercise 17.5. Let  X   be an ordered set in the order topology. Show that  (a, b)  ⊂
[a, b]. Under what conditions does equality hold? 

Proof.  Since (a, b) ⊆ [a, b] closed, and by the definition of closure,

(a, b) =

K  ⊆  [a, b].
K ⊇(a,b) closed

(a, b) = [a, b] ⇐⇒  ∈
a, b (a, b) ⇐⇒
any basis elements A a, B    
b  intersect
(a, b) by Theorem 17.5(b). We claim that this is equivalent to the fact that there is
no immediate successor α of  a and no immediate predecessor β  of  b. If either are
the case, say for a, then choosing A with upper bound α would not intersect (a, b),
and so equality doesn’t hold since  a /  (a, b); in the other direction, if neither are the
∈
case, we see that, say for a, the upper bound of  A, α would be such that (a, α) is
non-empty, and so A (a, b) = , satisfying the condition for Theorem 17.5(b). The
∩ ∅
same argument applies when considering  b and β , and so our claim holds.

Exercise 17.13.  Show that  X  is Hausdorff if and only if the  diagonal  ∆ = x x {× |

∈ }  ×
x  X   is closed in  X  X .

Proof.   Suppose ∆ is closed in X   X , i.e., the complement ∆ c is open. This is

 ×
equivalent to for all (x, y) X  X  such that x = y, there exists a basis element
 ∈  ×  
 × × ∈  ×
U  V  of  X  X  for U, V   open in X  such that (x, y)  U  V   but (U  V  ) ∆ = .  × ∩ ∅
But then, by definition of ∆, this is equivalent to saying for all x, y  X   such that
 ∈
    
x =  y, there exist open neighborhoods U   x and V   y  such that U  V  = , and  ∩  ∅
so X  is Hausdorff.

Exercise 17.16.  Consider the five topologies on  R  given in   Exercise  7 of  13.  §
(a)   Determine the closure of the set  K  = 1/n n Z+   under each of these 
 {  |  ∈ }
topologies.
(b)  Which of these topologies satisfy the Hausdorff axiom? the  T 1  axiom? 

Solution for  (a). For 3 , A   closed

 T  A   finite or all of  R. Since no finite set
⇐⇒
contains all of  K , we see that  R  is the only closed set containing K , and so K  = R.

7
  T  ∞  ∈ ∞
For 5 , we claim K  = [0, ). For x [0, ), the basis elements that contain
x are of the form ( −∞, a) for a > x. Since ( −∞ ∩    ∅
, a) K  =   by the Archimedean
property, that is,  > 0 N  N  such that 1/N < , K  = [0, ) by Theorem 17.5.
 ∀ ∃  ∈ ∞
For 1 , K  = 0  by Example 17.8, and so K  = K 
 T  {} 0 .  ∪{ }
For 2 , K  is closed since R K  = (
T  \ −∞ ∞ \
, ) K  is a basis element, and so K  = K .
For 4 , K  = K  since 4  is finer than 2 , and so  R K   is still open.
 T   T   T  \
Solution for  (b). T  is T   since any finite point set is closed by definition of the finite
3 1
complement topology. It is not Hausdorff, for if we choose  U    x, V    y  both open,
c c  c
(U  ∩ V  ) = U  ∪ V  is finite, where the equality follows from De Morgan’s Laws, and
so U  ∩ V   is infinite.
T   is not Hausdorff and not even  T  , for R \{x }  is not a union of basis elements,
5 1 0
and so {x }  is not closed.
0
T   is Hausdorff, for if we have x, y ∈  R  and 0 <  < |x − y|/2, then (x − , x +
1
) ∩ (y − , y + ) = ∅. Since Hausdorff =⇒ T  , we see that  T   is also T  . 1 1 1
Since T  , T   are both finer than  T  , we see that the open sets constructed above
2 4 1
are still open and separate x, y, and so T  , T   are still both Hausdorff and thus T  .
2 4 1

18 Continuous Functions
Exercise 18.1.  Prove that for functions  f  : R → R, the  -δ   definition of continuity 
implies the open set definition.
Remark.  Recall that f   is continuous if for every  > 0 and x0 R, there exists a  ∈
δ > 0 such that f (x) f (x0 )  <  for all x R  such that x x0  < δ .
− |  | ∈  | − |
Proof.   Consider x ∈ R, and a corresponding neighborhood V  of  f (x ); we then
0 0
have V  ⊇  (f (x ) − , f (x ) + ) for some  > 0 since V  is open. Then, by hypothesis
0 0
there exists a δ > 0 such that f (x) ∈ (f (x ) − , f (x ) + ) for all x ∈  R  such that
0 0
x ∈  (x − δ, x  + δ ) = U , which is open. Thus, f (U ) ⊆  V   , and so f  is continuous by
0 0
Theorem 18.1.
Exercise 18.12. Let F  :  R × R → R  be defined by the equation 
F (x × y) =
 2
xy/(x + y ) if  x × y 2
= 0 × 0.
0 if  x × y = 0 × 0.

(a)  Show that  F  is continuous in each variable separately.

(b)   Compute the function  g :  R R  defined by  g(x) = F (x
→ × x).
(c)  Show that  F  is not continuous.

8
Proof of  (a).   Since F  is symmetric in interchanging x  ↔
y, we only have to prove
y0 Y  , h(x) = F (x y0 ) is continuous as a function R  R . For y0  = 0, this is
∀  ∈ ×  →
trivially true for the image of  h is (0, 0) with preimage R. Now suppose y 0 = 0; then  
we have h(x) = xy0 /(x2 + y 02 ). This is continuous since xy0 and x2 + y 02  are both
continuous, and so their quotient is also continuous (since also x2 + y02 = 0), using  
the -δ   definition of continuity (see Theorem 21.5).
2 2 2 2 2
Proof of  (b).   Since F (x × x) for x = 0 equals x /(x + x ) = x /2x = 1/2, we have

g(x) =
 1/2 if  x 
= 0.
0 if  x = 0.
×
Proof of  (c).  We claim F (x y) is not continuous along the line L = x = y at (0, 0), { }
|
i.e., F  L  is not continuous at (0, 0); this suffices by Theorem 18.2(d). Note that the
line L in the subspace topology is homeomorphic to R, where the homeomorphism
is given by either of the coordinate projection maps. Now the preimage of the closed
 { } ⊆
set 1/2  R is L (0, 0) , which is not closed since R 0  is not closed, hence
\{ } \{ }
|
F  L  is not continuous, and neither is F .

19 The Product Topology

Exercise 19.6. Let  x 1 , x2 , . . . be a sequence of the points of the product space  X α .

Show that this sequence converges to the point  x  if and only if the sequence  πα (x1 ),
πα (x2 ), . . . converges to π α (x) for each  α. Is this fact true if one uses the box topology 
instead of the product topology? 
Proof.   Suppose xi
 { } →
  x , and fix some index γ . Then, for any neighborhood U γ   
πγ (x), letting U  = U α  where U α  = X α  for all α =  γ , there exists  N  N  such that
  ∈
xi U   for all i N , and so πγ (xi ) πγ (U ) = U γ   for all i N , i.e., πγ (xi )
 ∈  ≥  ∈  ≥  { }→
πγ (x). Note that this direction does not depend on the topology being the product
or box topology.
In the other direction, suppose πα (xi )
  {  π α (x) for all α. We take an arbitrary
}→ 
neighborhood V  of  x  ∈ X α ; it then contains a basis element of  X α  containing

x, which is a product of open sets U α . In the case of the product topology, there
then exist only finite U α   X α , and for these open sets there exist  N α N such that  ∈
πα (xi )  U α  for all i  N α  for each α. N α  = 1 works for all other α. Thus, we can
 ∈  ≥
 ∈   ⊆
take N  = max(N α ); then,  x i U α  V   for all i  N . ≥
We construct a counterexample for this direction in the case of the box topology.
Let  R N be the box topology on the product of copies of  R  indexed by  N , and let
xi  := ( 1i , 1i , 1i , . . .).

9
Then, for each α N, πα(xi )
 ∈  { (0, 0, 0, . . .) =: x, but this sequence does not
}→
converge in the box topology, for the open set
− ( 1 1
i, i
) −
= ( 1, 1) × (− 1  1
2, 2
) × (− 1 1
3, 3
) ×···
i∈N

in the box topology contains x = (0, 0, 0, . . .), but does not contain any xi .

20 The Metric Topology

Exercise 20.4.  Consider the product, uniform, and box topologies on  Rω .
(a)   In which topologies are the following functions from  R to  R ω continuous? 

f (t) = (t, 2t, 3t , . . .),

g(t) = (t , t , t , . . .),
h(t) = (t,  12 t, 13 t , . . .).

(b)  In which topologies do the following sequences converge? 

w1  = (1, 1, 1, 1, . . .), x1  = (1, 1, 1, 1, . . .),

w2  = (0, 2, 2, 2, . . .), x2  = (0, 12 ,  12 , 12 , . . .),
w3  = (0, 0, 3, 3, . . .), x3  = (0, 0, 13 , 13 , . . .),
... ...
y1  = (1, 0, 0, 0, . . .), z1  = (1, 1, 0, 0, . . .),
y2  = ( 12 , 12 , 0, 0, . . .), z2  = ( 12 ,  12 , 0, 0, . . .),
y3  = ( 13 , 13 , 13 , 0, . . .), z3  = ( 13 ,  13 , 0, 0, . . .),
... ...

Solution for  (a).  For the product topology, by Theorem 19.6, f , g, h are all continu-
ous since each coordinate function is continuous. This is because if an open set in
the image of a coordinate function is (a, b), its preimage would still be in the form
(a , b ) R where a  , b are determined by the linear equations defining  f , g, h above.
⊆
Now consider the uniform topology. Note by Theorem 21.1 we can use the familiar
-δ  definition for continuity since our spaces both are metric spaces. We claim f   is not
continuous. For, suppose it is continuous. Then, given   > 0 and x R, there exists ∈
|−|
δ > 0 such that x y  < δ  = ⇒| − | |−|
f (x) f (y)  = supn [min(n x y , 1)] < . But, this is
|−|
a contradiction since for n large, min(n x y , 1) = 1, and so is always greater than .
Now consider g. g is continuous since given  > 0 and x R, we let δ < min(, 1) and
∈
10
 |−| ⇒|
therefore have x y  < δ  = − | |−| |−|
f (x) f (y)  = supn [min( x y , 1)] = min( x y , 1) <
min(, 1)  . h is also continuous since given   > 0 and x R, we let δ < min(, 1)
≤ ∈
 |  − |
and therefore have x  y < δ  = ⇒ |  − | |  − |
f (x)  f (y)   = supn [min( x  y /n, 1)] ≤
| − | ≤
min( x y , 1) < min(, 1)  .
For the box topology, since the box topology is finer than the uniform topology
by Theorem 20.4, we see that f   is not continuous. For, if  V   open in the uniform
topology has preimage that is not open in R, V  is still open in the box topology and
still has the same non-open preimage. Next, by Example 19.2, we see that g is not
continuous. Last, for h, we choose
1 1 1 1
−
B = ( 1, 1) × (− , )
22 22
× (−
× · · ·  ,, )
32 32

−1
and suppose its preimage h (B) is open. This implies h((−δ, δ )) ⊆ B, and so
applying π  gives
n
δ δ 1 1
h ((−δ, δ )) = (− , ) ⊆  ( − , )
n n n n2 n2

for all n, a contradiction.

Solution for  (b).  We note that since the product topology is Hausdorff by Theorem
19.4 and both the uniform and box topologies are finer than the product topology by
Theorems 19.1 and 20.4, if a sequence converges to a point  p  in one topology, it must
converge to the same point in the finer topologies. For, if the sequence converges to
q  in the finer topology, then it also converges to q  in the coarser topology, and by
the Hausdorff property  p  =  q .
 Consider wn . For the product topology, we recall that any basic open set U  =
U α   0 is the product of finitely many open subsets of  R  with infinitely many

copies of  R. Letting N   be the largest α  such that U α  R, we see that wn  ∈ U 
for all n > N   since the first N  components are zero, and the rest are trivially in
the remaining copies of  R of  U . Thus, wn  →
 0 in the product topology. Now we
only have to check if the sequence converges to zero in the other topologies by the
above. In the uniform topology, ρ(wn , 0) = 1 for all n, and so the sequence does not
converge. For, if we choose any ball U  = B(0, r)  R ω for r < 1, wn /  U   for all n.
 ⊆ ∈
Finally, since the box topology is finer than the uniform topology by Theorem 20.4,
we see that this same open set  U  is such that  w n /  U  for all n, and so wn  does not
∈
converge in the box topology, either.
Consider xn  and yn . We claim they both converge to zero in the uniform topology.
For any open set 0  U  Rω in the uniform topology, we can find B (0, ) such that
∈  ⊆
B(0, ) ⊆ U ; then, we can find N  such that 1/N < . We then see that xn , yn  ∈
B(0, ) U   for all n N , and so xn , yn   0 in the uniform topology. Moreover,
 ⊆  ≥  →
since the uniform topology is finer than the product topology, we see that this implies

11
xn , yn  0 in the product topology as well. For the box topology, though, we see that
 → 
neither sequence converges. For, we can construct the set 0  U  = ∞ ∈
n=1 ( 1/n, 1/n) −
(where we only consider sets containing zero by the above), which does not contain
xn , yn  for any n. 
For zn , we see that for any open set 0 U  = U α ∈ Rω in the box topology,
 ⊆
 ∈  ≥  ∈
for N  large enough 1/n  U 1 , U 2 for n  N , and so zn  U   for all n  N , since by  ≥
hypothesis 0  U , the third component onwards of  zn  are always in their respective
∈
 →
U α . Thus,  z n  0 in the box topology; since the box topology is finer than both the
uniform and product topologies, we see that this implies zn  converges in the other
two topologies as well.
Exercise 20.5. Let  R∞ be the subset of  Rω consisting of all sequences that are 
eventually zero. What is the closure of  R∞ in  Rω in the uniform topology? Justify 
your answer.
Solution.  We claim that A = R∞ is the set of all sequences that converge to zero; we
denote this latter set by X . It suffices to show by Theorem 17.5 that x  X  if and  ∈
 
only if every basis element U   x   intersects A. First suppose x  X  and let U   x ∈  
be a basis element in the uniform topology; we then see that we can find an open
ball B(x, )  U . We know we can find N   N  such that xn  <  for all n  N  by
 ⊆  ∈  | |  ≥
the definition of convergence. Then, define y such that yn = x n  for all n < N , and
zero otherwise; this means y  A. Then, ρ(x, y) < , and so y  B(x, ) A.
 ∈  ∈ ∩
∈
Now suppose x /  X ; it suffices to find a basis element containing x that does not
intersect A. Since x /  X , there exists a ball B(0, ) R such that xn n≥N   B(0, )
∈ ⊆ { }  ⊆
for any N . The ball B(x,/2), then, does not intersect A, since for any y  B(x,/2),  ∈
≥
it is not the case that  y n  = 0 for all n  N  for some N .
Exercise 20.6. Let  ρ  be the uniform metric on  Rω . Given  x = (x1 , x2 , . . .)  ∈ R ω

and given  0 <  < 1, let 

U (x, ) = (x1 − , x  + ) × · · · × (x − , x  + ) × · · · .
1 n n

(a)  Show that  U (x, ) is not equal to the  -ball  Bρ (x, ).
(b)  Show that  U (x, ) is not even open in the uniform topology.
(c)  Show that  
Bρ (x, ) = U (x, δ ).
δ<

Proof of  (a).   Consider the point

y  = (yn )n∈N , yn  = x n + 

n
1
= x n  +  1
−  1
.
k=1
2k 2n

12
yn  ∈ (x − , x  + ) for all n  implies  y ∈ U (x, ), while  y ∈/  B (x, ) since
n n ρ

ρ(x, y) = sup d(xn , yn ) = .

n

Proof of  (b). U (x, ) is not open since the point y in (a) has no neighborhood con-
⊆
tained in B ρ (x, ). For, suppose Bρ (y, δ )  U (x, ). We can find N  such that

 ∞
δ  1
> k
,
2 k=N +1
2

since
 1/2k converges, and so its tail becomes infinitesimally small. We see that
then, defining y such that yn = y n  for all n =  N  and y N 
  
= y N  + δ/2, y  B ρ (y, δ )
∈
but y  /  U (x, ) since yn  = y n +δ/2 > yn + ∞
∈ 1
k=N +1 2k = x n +, a contradiction.

 ⊇ ⊆
Proof of  (c). The  direction is clear, since each  U (x, δ )  B ρ (x, ) by the fact that
δ < . Now suppose z Bρ (x, ); if  ρ(x, z) = ξ , then we can find δ   (ξ, ) so that
 ∈  ∈
z  U (x, ξ ), i.e., B ρ (x, )
∈ ⊆
δ< U (x, δ ).

Exercise 20.8. Let  X   be the subset of  Rω consisting of all sequences  x   such that 
 x2i   converges. Then the formula 

d(x, y) =
∞
(xi −y) 2

1/2

i
i=1

defines a metric on  X . On  X  we have the three topologies it inherits from the box,
uniform, and product topologies on  Rω . We have also the topology given by the metric 
d, which we call the  2 -topology.
(a)   Show that on  X , we have the inclusions 
box topology  2 -topology   uniform topology.
 ⊃  ⊃
(b)   The set  R∞ of all sequences that are eventually zero is contained in  X . Show 
that the four topologies that  R∞ inherits as a subspace of  X  are all distinct.
(c)   The set  
H  = [0, 1/n]
n∈Z+

is contained in  X ; it is called the  Hilbert cube. Compare the four topologies 
that  H   inherits as a subspace of  X .

13
Proof of  (a).   box topology 2 -topology. Let
 ⊃ x  be a basis element in the 2 -
 U 
topology. Then, there exists a basis element U  = Bd (x, )   of the 2 -topology.
 ⊆ U 
We claim that V  = X   ∩   −
(xi  /2i/2 , xi  + /2i/2 ) x  basis element of the box
 
 ∈
topology is contained in U . Suppose  y  V   ; then
−  ∞ ∞
2 2 2
d(x, y) = (xi yi ) < i
=  2 = ⇒ y ∈ U,
i=1 i=1
2

⊆  ⊆ U 
i.e., V   U  , and box topology   2 -topology by Lemma 13.3.
 ⊃
2 -topology   uniform topology. Let
 ⊃ x  be a basis element in the uniform
 U 
topology. Then, there exists a basis element U  = X  Bρ (x, )  ∩
  of the uniform  ⊆ U 
2
topology. We claim V  = B d (x, ) basis element of the   -topology is contained in U .
If   > 1, then trivially  V   U  = X , so we assume   1. Suppose  y  V   ; then
⊆ ≤  ∈
ρ(x, y) = sup xi
| − y | ≤ d(x, y) <  i =⇒ y ∈ U,
2
i.e., V  ⊆  U  ⊆ U , and  -topology ⊃  uniform topology by Lemma 13.3.
Proof of  (b). By (a) and Theorem 20.4, we have the inclusions
2
box topology  ⊃  -topology ⊃ uniform topology ⊃ product topology,
  
since if  T   ⊃ T    in the ambient space, we would also have  T  ⊃ T  , where  T  are
1 2 1 2 i
the subspace topologies induced by T  on X , by the fact that the open sets in the
i
subspace X   are the open sets of the ambient space intersected with X . We claim
these are strict inclusions. 
box topology   2 -topology. Consider the open set U  = R∞ ( 1/i, 1/i)  0 ∩ −  
in the box topology. Consider any neighborhood
∞
 V 
  0; then, there exists some
V  = R ∩
 B d (0, ) for  >   0 contained in . We can find N  V N   such that ∈
1/N < , and so x  such that xi   = 0 for all i  except xN  = 1/N   is contained in V 
 V 
since d(0, x) = 1/N <  and therefore , but not in U . Hence, U  is open in the box
topology but not in the  2 -topology by p. 78, and so box topology     2 -topology.
2 -topology    uniform topology. Consider the open set U  = R∞  Bd (0, 1) in ∩
the 2 -topology. Consider any neighborhood  V 
 0; then, there exists some V  =
R∞ Bρ (0, ) for  > 0 contained in . We can find N  N  such that N2 > 4, and
∩  V   ∈
so x  such that xi = /2 for all 1 i N   and 0 otherwise is contained in V  and
 ≤  ≤
 V 
therefore , yet
 ∞
2 2
d(x, 0) = x2i = N  > 1 = ⇒ d(x, 0) > 1,
i=1
4

14
and so  x /  U . Hence, U  is open in the  2 -topology but not in the uniform topology
∈
by p. 78, and so  2 -topology    uniform topology.
uniform topology   product topology. Consider the open set  U  = R∞ Bρ (0, 1) ∩
in the uniform topology. Consider any V  = R∞ ∩
U α  0 where U α  = R for all but
 
finitely many α. Let N  be such that U N  = R; then,  x  such that xi  = 0 for all i except
| |≥
xN   1 is in V  but not in U . Hence, U  is open in the uniform topology but not in
the product topology by p. 78, and so uniform topology    product topology.
Solution for  (c).  We claim that
box topology    (2 -topology = uniform topology = product topology),
i.e., the box topology is strictly finer than the other topologies, which are equal.
To show the equality, it suffices to show product topology 2 -topology, for ⊃
then we have  2 -topology  uniform topology  product topology   2 -topology by
 ⊃  ⊃  ⊃
the same argument as in (b), and so we must have equality throughout. So, consider
 U   x open in the 2 -topology; then, we can find a basis element U  = H  Bd (x, ) ∩ ⊆ U 
of the 2 -topology for some  >  0. Let δ  = /[ζ (2) + 1]1/2 , and choose N   such that
 ∞ 2 2
i=N  1/i < δ  , which exists since ζ (2)  <  | | ∞
. We claim that

V  = H 
 ∩   −N −1
xi
 δ 
, xi  +
 δ 
×  
∞
R  x ,
i=1
i i i=N 

basis element of the product topology is contained in  U . Suppose  y  ∈ V  ; then
∞
 N 
  ∞
2 2 2 1 1
d(x, y) = (xi −y) i < δ  2
+ 2
< δ 2 [ζ (2) + 1] =  2 = ⇒ y ∈ U,
i=1 i=1
i i=N 
i

i.e., V  ⊆ ⊆ U 
U  , and product topology 2 -topology by Lemma 13.3. Thus, we
 ⊃
have the equality desired.
It remains to show the box topology is strictly finer than the other topologies;
since the other topologies are equal it suffices to show box topology  2 -topology.

But the open set  U  = H  ∩ − 
( 1/i, 1/i)  0 is open in the box topology but not the
 -topology by the same argument as the proof that box topology     2 -topology in
2

(b), and so we are done.

21 The Metric Topology (continued)

Exercise 21.1. Let A  X . If  d is a metric for the topology of  X , show that  d A×A
⊆ |
is a metric for the subspace topology on  A.

15
 |
Proof.   Clearly d A×A  is a metric since it inherits all the properties for a metric from
the metric d for X ; it therefore suffices to show that every basis element for the
subspace topology on A contains some open ball defined by d A×A , and vice versa, |
by Lemma 13.3.
So, suppose B  is a basis element for the metric topology on A; B  = Bd|A A (x, r) ×

for some x A and r R. But then, B = Bd (x, r) A  by definition, and so the

 ∈  ∈ ∩
subspace topology is finer than the metric topology.
Conversely, suppose B is a basis element for the subspace topology on A; it equals
Bd (x, r) A for some basis element  B d (x, r) of  X . Let y  Bd (x, r) A; we see that
∩  ∈ ∩
− ⊆ ∩
the set Bd|A A (y, r d(x, y))  A Bd (x, r) is a basis element for the metric topology
×

contained in A Bd (x, r), since z   Bd|A A (y, r d(x, y)) is such that
∩  ∈ ×
−
d(z, x) ≤  d(z, y) + d(x, y) = d | A×A (z, y) + d(x, y) < r − d(x, y) + d(x, y) = r.
Thus, the metric topology is finer than the subspace topology. Combining the two
inclusions, we see the topologies are equal.
Exercise 21.2. Let  X  and  Y  be metric spaces with metrics  d X  and  d Y  , respectively.
 →
Let f  : X   Y  have the property that for every pair of points  x1 , x2 of  X ,

dY  (f (x1 ), f (x2 )) = dX (x1 , x2 ).

Show that  f  is an imbedding. It is called an  isometric imbedding of  X  in  Y  .
Proof.  We first show f  is injective:

f (x1 ) = f (x2 ) = ⇒ dY  (f (x1 ), f (x2 )) = 0 = ⇒ dX (x1 , x2 ) = 0 = ⇒ x1  = x 2

by properties of metrics, and so we have an injective map.

Now we consider the map f  : X   im(X )  Y  , which is a bijection; it suffices to
 → ⊆
show that f  , f −1 are continuous to show f  is an imbedding. Let x  X  and  > 0  ∈
be given; then, letting  δ  = , we have

dX (x, y) < δ  = ⇒ dY  (f  (x), f  (y)) < ,

and so f   is continuous. Given y  ∈ Y  and  > 0, letting δ  =   gives

−1 −1
dY  (x, y) = dY  (f (a), f (b)) < δ  =⇒ d (f  (x), f  (y)) =  d
X  X (a, b) < ,

where a, b exist by the bijectivity of  f , and so f −1 is continuous.

Exercise 21.3. Let X n  be a metric space with metric  dn , for  n ∈Z+.

16
 (a)  Show that 
{
ρ(x, y) = max d1 (x1 , y1 ), . . . , dn (xn , yn ) }
is a metric for the product space  X 1 × · · · × X  . n
{ }
(b) Let di  = min di , 1 . Show that 
{
D(x, y) = sup di (xi , yi )/i }
is a metric for the product space 
 X i .
Proof of  (a). ρ satisfies properties (1) and (2) on p. 119 since the components do;
it then suffices to show the triangle inequality. We first have di (xi , z i )  d(xi , yi ) +  ≤
d(yi , z i ) for all i. Then, by definition of  ρ, di (xi , z i )  ρ(x, y) + ρ(y, z ) for all i. But
 ≤
≤
since this is true for all  i, we have that ρ(x, z )  ρ(x, y) + ρ(y, z ). 
We now show that this defines a metric for the product space. First let  B = U i

be a basis element of  X i , and let x  ∈
B. For each i, there is an i   such that
 ⊆
Bdi (xi , i ) { }
U i . Choosing   = min 1 , . . . , n , we see that Bρ (x, ) B, since if   ⊆
y Bρ (x, ), di (xi , yi )
 ∈  ≤ ρ(x, y) <  i , and so y
 ≤  ∈ 
U i   as desired. Thus the
metric topology is finer than the product topology.
Conversely, let B ρ (x, ) be a basis element in the metric topology; since it is the
product Bdi (xi , ), we see that the product topology is finer than the metric topology.
These two facts imply the two topologies are equal.
Proof of  (b). D satisfies properties (1) and (2) on p. 119 since the components do; it
then suffices to show the triangle inequality. We first have
di (xi , z i ) di (xi , yi )  d i (yi , z i )
i
≤ i
+
i
≤
 D(x, y) + D(y, z )

for all i. But since this is true for all i, we have that

D(x, z ) = sup

 di (xi , z i )
≤  D(x, y) + D(y, z ).
i
We now show that this defines a metric for the product space. Let  U  be open in
 ∈
the metric topology and let x U ; choose an open ball BD (x, ) U . Choose N   ⊆
such that 1/N < , and let
V  = Bd1 (x1 , ) ×···×B × X  × X  × · · · .
dN  (xN , ) N +1 N +2

We claim V  ⊆B D (x, ) ⊆ U . Given y ∈


X  , d (x , y )/i ≤ 1/N  for i ≥ N .
i i i i
Therefore,  
d1 (x1 , z 1 )  d N (xN , z N ) 1
D(x, y) ≤ max , ···  , , .
1 N  N 

17
If  y  V   , this expression is less than  , so V   B D (x, ) as desired, and the product
 ∈ ⊆
topology is finer than the metric topology.

Conversely, let U  = U i   where U i   is open in X i for α1 , . . . , αn and U i = X i
otherwise. Let x ∈ U   be given, and choose Bdi (xi , i ) ⊆
X i for i = α1 , . . . , αn ,
 ≤ {  |
where each i   1. Then, defining    = min i /i i = α1 , . . . , αn , we claim that }
∈ ⊆
x  B D (x, )  U . Let  y  be a point of  B D (x, ). Then, for all i,

di (xi , yi )
i
≤ D(x, y) < .
Now if  i = α1 , . . . , αn , then   ≤ ≤
i /i, so that di (xi , yi ) < i   1; it follows that
| − |
xi yi  < i , and so y  ∈  U i  as desired. We thus have that the metric topology is
finer than the product topology; combined with the above this implies the topologies
are equal.

22 The Quotient Topology

Exercise 22.2.
 →
(a) Let p : X   Y  be a continuous map. If there is a continuous map f  : Y   X  →
◦
such that  p f  equals the identity map of  Y  , then  p is a quotient map.
(b) If  A  ⊂X , a  retraction of  X   onto A is a continuous map r : X  A  such   →
∈
that  r(a) = a  for each  a  A. Show that a retraction is a quotient map.

Proof of  (a). If  V  ⊆ Y   with U  = p−1 (V  ) X   open, f −1 (U ) = f −1 ( p−1 (V  )) =
⊆
( p f )−1 (V  ) = V   is open. Thus, p  is a quotient map.
◦
Proof of  (b). Let ι : A X   be the inclusion map; then, r
 → ◦ ι is the identity on A,
hence r is a quotient map by (a).

Exercise 22.4.
(a)   Define an equivalence relation on the plane  X  = R2 as follows:

x0 × y  ∼ x × y
0 1 1 if  x0  + y02 = x 1  + y12 .

Let X ∗ be the corresponding quotient space. It is homeomorphic to a familiar 

space; what is it? 
(b)   Repeat  (a) for the equivalence relation 

x0 × y  ∼ x × y
0 1 1 if  x20  + y02 = x 21  + y12 .

18
Solution for  (a). Set g(x  y) = x +  y 2
 × R. We see it is a surjection onto R
∈
since R × { } →
0 R. It is continuous since for x0  y 0 X , given  >   0, letting
 ×  ∈
| | × ×
δ   = min(1, /2( y0 + 1)), ρ(x0 y0 , x y) < δ   implies
2 2
|g(x × y ) − g(x × y)| = |(x  + y ) − (x + y )|
0 0 0 0
2 2
≤ |x − x| + |y  − y | 0 0
≤ |x − x| + |y  + y||y − y|
0 0 0
2
≤ |x − x| + |y − y | + 2|y ||y − y|
0 0 0 0
< 2( |y | + 1)δ < . 0

If we define  f  :  R →  X  by x →  x × 0, which is continuous since (a, b) × (c, d) maps
back to (a, b) which is open in R, we see g ◦ f  is the identity on R, and so g is a quotient
map by the lemma above. Since x × y  ∼ x  × y ⇐⇒ g(x  × y ) = g(x  × y ),
0 0 1 1 0 0 1 1
 ∗
by Corollary 22.3, this induces a bijective continuous map g : X  → R, which is a
homeomorphism since g  was a quotient map.
Solution for  (b). Set g(x y) = x 2 + y2 R. We see it is a surjection onto R≥0 , since
× ∈
R 0
×{ } → R≥0 , and it does not map to anything else since  x 2 + y2  0 for all x, y. It
≥
is continuous since for x0 y0  X , given  > 0, letting δ  = min(1, /2( y0 + x0 +1)),
×  ∈ | | | |
× ×
ρ(x0 y0 , x y) < δ   implies
2 2 2 2
|g(x × y ) − g(x × y)| = |(x  + y ) − (x + y )|
0 0 0 0
2 2 2 2
≤ |x − x | + |y  − y |
0 0
≤ |x − x||x − x  + 2x | + |y − y||y − y  + 2y |
0 0 0 0 0 0
≤ |x − x|(1 + 2|x |) + |y − y|(1 + 2|y |)
0 0 0 0
≤ 2δ (|x | + |y | + 1) < .
0 0
 √ 
We define f  : R  → X  by x → x × 0, which is continuous since the preimage of 
≥0
   2
(a, b) ×  (c, d), if (c, d)     0, is the open set R  ∩ (a , b ), where a = a if  a  ≥ 0,
≥0

and −1 otherwise, and similarly for b (we chose  −1 out of convenience; we really

only have to make sure the preimage is a half-open set [0 , b ) or the empty set in
these cases); if (c, d)     0, then the preimage would be empty. We then see g ◦ f 
is the identity on R , and so g   is a quotient map by the lemma above. Since
≥0
x  × y  ∼ x × y ⇐⇒ g(x  × y ) = g(x × y ), by Corollary 22.3, this induces a
0 0 1 1 0 0 1 1
 ∗
bijective continuous map g : X  → R , which is a homeomorphism since g   was a
≥0
quotient map.
Exercise 22.6.  Recall that  R K  denotes the real line in the  K -topology. Let  Y   be the 
quotient space obtained by  RK  by collapsing the set  K  to a point; let  p :  R K   Y  be   →
the quotient map.

19
 (a)  Show that  Y   satisfies the  T 1  axiom, but is not Hausdorff.
(b)  Show that  p  p : RK  RK   Y  Y  is not a quotient map.
×  ×  →  ×
Proof of  (a). Recall by p. 141 that it suffices to show every element in the partition,
i.e., one-point sets x for x / K  and K   itself, are closed in RK . The former are
 { } ∈
closed since R K  is T 1  since it is Hausdorff by Example 31.1, and the latter is closed
since it is the complement of  R K . Thus, Y  is T 1 .
\
We now show Y  is not Hausdorff. We claim that p(0), p(K ) are not separable; note
 p(0) =  p(K ) since they are in different equivalence classes. Suppose  Y  is Hausdorff,

and let V 1
−1
 p(0), V 2  p(K ) be a separation in Y ; they have open preimages
−1
   ⊇
U 1 = p (V 1 ) 0, U 2 = p (V 2 ) K  by definition of a quotient map. There then
exists (a, b) K   0 contained in U 1 , and choosing n  N  such that 1/n < b, there
\    ∈
 ≤
exists (c, d)  1/n contained in U 2 , where we can assume 1/(n + 1)  c, since if not,
 ⊆  ∩
we can take the intersection with (1 /(n + 1), d). Then, (c, 1/n) U 1 U 2 , and so
⊆ ∩
 p((c, 1/n))  V 1 V 2 , which is a contradiction, and so  Y  is not Hausdorff.
Proof of  (b).   By Exercise   17.13,   we see that since Y   is not Hausdorff by (a), the
diagonal ∆Y  ⊆ × Y   Y   is not closed. ( p−1  p −1 )(∆Y  ) = ∆K   (K   K ), where
×  ∪  ×
∆K  RK  RK   is the diagonal in RK . However, ∆K   is closed by Exercise   17.13
⊆  ×
since RK  is Hausdorff by Example 31.1, and so ∆K  (K  K ) is closed since is is
 ∪  ×
the finite union of closed sets. Thus, the inverse image of the non-closed set ∆ Y  is
×
closed, and so p  p is not a quotient map.

3 Connectedness and Compactness

23 Connected Spaces
Exercise 23.8.  Determine whether or not  R ω is connected in the uniform topology.
Solution. Let Rω = A B, where A is the set of bounded sequences and  B  is the set
∪
of unbounded sequences of reals. A, B are disjoint, and so it remains to show they
 ∈  ∈  | |
are open. Suppose a = (a1 , a2 , . . .)  A and b = (b1 , b2 , . . .)  B. Since ai  < N  for
 | |
all i for some N , and since bi > N   + 1 for all i larger than some I , we have that
 ≥  ∈  ∈
d(ai , bi ) = 1 for all i  I . Thus, ρ(a, b) = 1 for any a A, b B, and so the open
balls with radius 1/2 around a, b are fully contained in  A, B respectively.

Exercise 23.11. Let  p : X   Y   be a quotient map. Show that if each set  p−1 ( y )
 → {}
is connected, and if  Y  is connected, then  X  is connected.

20
  ∪
Proof.   Suppose not. Then, X  = A B for A, B   open, disjoint sets. Consider
C  = y  Y  p−1 ( y )  A , D = y  Y   p −1 ( y ) B ; we see that these sets
 {  ∈ | { }  ⊆ }  {  ∈ | { }  ⊆ }
are such that C  D = Y   since p−1 ( y ) connected implies it is in either  A or B by
 ∪ {}
Lemma 23.2. C, D are then disjoint by definition and p−1 (C ) = A, p−1 (D) = B by
the fact that p is surjective. p quotient map implies that C, D are then open, and so
 ∪
Y  = C  D is a separation, a contradiction.

24 Connected Subspaces of the Real Line

Exercise 24.7.
 →
(a) Let X  and  Y   be ordered sets in the order topology. Show that if  f  : X   Y  is 
order preserving and surjective, then  f   is a homeomorphism.
(b) Let  X  = Y  = R+ . Given a positive integer  n, show that the function  f (x) = x n
is order preserving and surjective. Conclude that its inverse, the  nth root
function, is continuous.
(c) Let  X  be the subspace  ( , 1) [0, ) of  R. Show that the function  f  : X 
−∞ − ∪ ∞  →
R  defined by setting  f (x) = x + 1 if  x < 1, and  f (x) = x if  x  0, is order 
 −  ≥
preserving and surjective. Is  f   a homeomorphism? Compare with  (a).

Proof of  (a). f   is injective since if  f (a) = f (b) but a = b, then (with possible

swapping) a < b, and so f (a) < f (b), a contradiction. We thus must show f 
and f −1 are continuous. But f   is continuous since f −1 ((a, b)) = (f −1 (a), f −1 (b)) is
open (apply the same argument to the intervals of the form [ a0 , b), (a, b0 ] for a0 , b0
minimal and maximal respectively); the same argument applies for  f −1 as well.
Proof of  (b). f (x) = x n is order preserving since a < b =   a/b < 1 = an /bn <
⇒ ⇒
1 = ⇒ an < bn =
⇒ f (a) < f (b). f   is continuous since it is the product of  n
copies of the identity function, which is continuous. We want to show  f  is surjective.
Letting N  = xn x Z≥0 , we see that every real number y Y   is between two
 { |  ∈ }  ∈
consecutive members of  N , or it is already an nth power of an integer, in which case
 ∈
it is trivially mapped to by its nth root. In the case y Y   is not an nth power,
we have f (n) < y < f (n + 1), and so by the Intermediate value theorem (Theorem
∈
24.3), we see that there exists a point  c  X  such that f (c) = r, i.e., f  is surjective.
Since f   is order preserving and surjective, by (a) it is then a homeomorphism,
and so f −1 , the nth root function, is also continuous.
Proof of  (c). f  is order-preserving on ( , 1) since a < b =
−∞ − f (a) = a + 1 <
⇒
∞
b + 1 = f (b), and on [0, ) since it is the identity. We check that it is order
 −  ≥
preserving around the boundary. So, suppose a < 1 and b   0. Then, a < b but

21
also a +1 < 0  b, and so f  is order-preserving. f   is surjective since if  x R, if  x  0
≤ ∈ ≥
−
its preimage is itself, and if  x < 0, its preimage is x 1. f  is not a homeomorphism
by Theorem 23.6 since  R  is connected but X  is not, by considering f −1 (R).
This does not contradict (a) since X  is not in the order topology. Even if  R is in
the order topology, the subspace topology induced on X  is not the order topology.
For, ( 1/2, 1) is open in R, and so ( 1/2, 1) X  = [0, 1) is open in X , but not open
− − ∩
in the order topology on  X .

Exercise 24.8.
(a)  Is a product of path-connected spaces necessarily path connected? 
⊂
(b) If  A  X  and  A is path connected, is  A necessarily path connected? 
 →
(c) If  f  : X   Y   is continuous and  X  is path connected, is  f (X )  necessarily path 
connected?  
 { }
(d) If  Aα  is a collection of path-connected subspaces of  X   and if  Aα = , is 
    ∅
Aα  necessarily path connected? 
  ∈
Solution for  (a).   Yes. Let X  = X α , x, y  X . Since each X α is path connected, we
have f α : [0, 1]  X α  continuous such that f α (0) = x α , f α (1) = y α , where we assume
→
the closed interval is [0, 1] after composition with multiplication and addition, which
are continuous operations. Thus we have the function  f  = (f α ), which is continuous
by Theorem 19.6, with f (0) =  x , f (1) =  y , and so X  is path-connected.
Solution for  (b).  No, since S  in Example 24.7 is not path-connected while  S  is: it is
the image of the continuous map  x  (x, sin(1/x)) from  R >0 to  R 2 .
→
−1 −1
Solution for  (c).   Yes. For, let x, y  ∈ f (X ), and choose x  ∈ f  (x), y  ∈ f  (y).
0 0
Then, there exists continuous g : [a, b] → X  such that g(a) = x , g(b) = y , and so
0 0
its composition f  ◦ g : [a, b] →  Y  is continuous with (f  ◦ g)(a) = x, (f  ◦ g)(b) = y.

  α

Solution for  (d).   Yes. Let x, y ∈ A and p ∈ A . Then, there exists a contin-
α 
uous map f  : [a, b] → A  with f (a) = x, f (b) = p, and similarly g : [b, c] → A
α α
with f (b) = p, f (c) = y, since a, p ∈  A  for some α and similarly for y (we are free
α
to have Dom g = [b, c] by composition with multiplication and addition, which are
continuous). Then, by the pasting lemma (Theorem 18.3) since f, g are continuous
and f (b) = g(b), we see that h = f  on [a, b] and h  = g on [b, c] is a continuous map
such that h(a) = x, h(c) = y.

Exercise 24.12.  Recall that  S Ω   denotes the minimal uncountable well-ordered set.
 ×
Let  L   denote the ordered set  S Ω  [0, 1)   in the dictionary order, with its smallest 
element deleted. The set  L is a classical example in topology called the  long line.

22
Theorem. The long line is path connected and locally homeomorphic to R, but it
cannot be imbedded in  R .
(a) Let  X   be an ordered set; let  a < b < c  be points of  X . Show that  [a, c) has 
the order type of  [0, 1) if and only if both  [a, b) and  [b, c) have the order type 
of  [0, 1).
(b) Let  X  be an ordered set. Let  x 0  < x1  < ···
 be an increasing sequence of points 
{ }
of  X ; suppose  b = sup xi . Show that  [x0 , b) has the order type of  [0, 1) if and 
only if each interval  [xi , xi+1 ) has the order type of  [0, 1).
(c) Let  a0  denote the smallest element of  S Ω . For each element  a of  S Ω  different 
× × ×
 from a0 , show that the interval  [a0 0, a 1) of  S Ω [0, 1) has the order type 
of  [0, 1).
(d)  Show that  L is path connected.
(e)  Show that every point of  L  has a neighborhood homeomorphic with an open 
interval in  R.
(f )  Show that  L cannot be imbedded in  R , or indeed in  Rn  for any  n.

Proof of  (a). We first note order-preserving maps are injective. Letting f  : A B  →
 ∈
be such a map, if  a1 , a2  A, then one is larger than the other by the comparability
property of order relations, so one of  f (a1 ), f (a2 ) is larger than the other, hence
unequal.
Now suppose [a, c) has order type [0, 1), and let f  : [0, 1)  →
[a, c) be the order
isomorphism. We claim

g(x) = f  [f −1 (b)]x ,

{ } h(x) = f  f −1 (b) + [1
{ − f  (b)]x} −1

define order isomorphisms g : [0, 1) → [a, b) and h : [0, 1) → [b, c). They are order-
preserving since if  x, y ∈ [0, 1),
−1 −1 −1 −1
x < y =⇒ [f  (b)]x < [f  (b)]y =⇒ g(x) = f {[f  (b)]x}  < f {[f  (b)]y}  = g(y)
−1 −1 −1 −1
x < y =⇒ f  (b) + [1 − f  (b)]x < f  (b) + [1 − f  (b)]y
−1 −1 −1 −1
=⇒ h(x) = f {f  (b) + [1 − f  (b)]x}  < f {f  (b) + [1 − f  (b)]y}  = h(y)

where the first implications are due to our linear transformations being strictly mono-
tonic increasing, and the second since f  is order-preserving. This also implies injec-
tivity by the above. It remains to show surjectivity. Let z   [a, b), z   [b, c). Then,
 ∈ ∈
0  f −1 (z )  < f −1 (b) and f −1 (b)  f −1 (z  ) < 1, and so
≤ ≤
 
f −1 (z )
g −1 = z, h

f −1 (z  ) f −1 (b)
−1
− 
= z  .
f  (b) −
1 f  (b)

23
 Conversely, suppose [a, b) and [b, c) have order type [0, 1), and let g : [0, 1) → [a, b),
→
h: [0, 1)  [b, c) be the order isomorphisms. We claim

f (x) =
 g(2x) if 0  x < 1/2 ≤
−
h(2x 1) if 1/2  x < 1 ≤
is an order isomorphism. It preserves orders since g, h   preserve orders on their
≤
respective domains, and since if  x < 1/2  y, applying f   gives

f (x) = g(2x) < b ≤  h(2y − 1) = f (y).

This also shows injectivity by the above. f  is surjective since if  z   [a, c),  ∈
z<b = ⇒ f [g−1 (z )/2] = z,  ≥
z   b = ⇒ f {[h −1
}
(z ) + 1]/2  = z.

Proof of  (b).  Suppose [x0 , b) has order type [0, 1). For any i Z+ , by (a), [xi , b) has ∈
order type [0, 1); applying (a) again gives that [xi , xi+1 ) has order type [0, 1).
Now suppose every [xi , xi+1 ) has order type [0, 1). If f i : [0, 1) [xi , xi+1 ) are  →
order isomorphisms, first define

∞) → [x , b), x → f  (x − i) if  x ∈ [i, i + 1),

f  : [0, 0 i

which is well-defined since any x ∈ [0, ∞) is in some set of the form [i, i + 1). We
claim f  is an order isomorphism. If  x, y ∈ [0, ∞), then x ∈  [i, i + 1), y ∈ [ j, j + 1) for
some i, j. Suppose x < y. Then,


i = j = ⇒ f (x) = f i (x − i) < x  ≤ x  ≤ f  (y − j) = f (y),
i+1  j  j
i = j = ⇒ f (x) = f i (x − i) < f  (y − i) = f (y),
i

since the f i  are order-preserving; this also implies injectivity by the above. To show

surjectivity, we first know f  maps onto i [xi , xi+1 ) by definition. So let z   [x0 , b).  ∈
{ }
Since b is the least upper bound of the xi , z  is not an upper bound, so there exists i
such that z   [xi , xi+1 ). But then, since the f i  are bijective as well, f (f i−1 (z ) +i) = z .
 ∈
 → ∞  →
Now let g : [0, 1)  [0, ) be defined as x  x/(1 x); this is an order isomor- −
phism since it has inverse x/(1 + x), and since it is strictly monotonic increasing.
◦ →
Thus, f  g : [0, 1)  [x0 , b) is a bijection, and preserves orientation since  f, g do.
Proof of  (c). Let a > a0 . We proceed by transfinite induction. S Ω  is a well-ordered
 ∈
set, and so if we let J  be the set of  a  S Ω  such that the claim holds, it suffices to
∈  ⊆
show that for every  a  J , S a  J  = a  J . ⇒ ∈
24
 We first show that either a has an immediate predecessor or there exists a se-
 { } ⊆
quence ai { }
S a  such that a = sup ai . Suppose a  does not have an immediate
 { }
predecessor. Then, we have the section S a = bi , which is countable by definition
   ∅
of  S Ω . (b1 , a] =  since a has no immediate predecessor, and so let a1  (b1 , a]. We  ∈
construct the ai  inductively as follows: if we have an , let an+1  (sup an , bn+1 , a],  ∈ { }
which is nonempty as above. We then get a sequence of elements  a 1  < a2  <  < a. ···
But since an > bn  for all n by construction, we see that a sup ai . Moreover, if   ≥ { }
{ } { }
a > sup ai , then sup ai  = b k  for some k, for S a  contains all elements less than a,
and hence sup ai  < ak , contradicting that sup ai  is an upper bound.
{ } { }
 ⊆ − × ×
Now suppose S a  J . If  a has an immediate predecessor a 1, then [a0 0, a 1) =
[a0 0, (a 1) 1) [a 0, a 1) has order type [0, 1) by (a), for we have the order
× − × ∪ × ×
× × → × →
isomorphism [a 0, a 1)  [0, 1) defined by a x  x, which is trivially bijective
×
and order-preserving since S Ω [0, 1) was constructed with the dictionary order. On
the other hand, if  a  does not have an immediate predecessor, then there exists a
 { } ⊆
sequence ai { }
 S a  such that a = sup ai , and so the claim follows by (b).
 
Proof of  (d). Let a × b, a × b be two points in L; suppose without loss of generality
  
that a × b < a × b . By (c), [a × 0, a × 1) and [a × 0, a × 1) have order type [0, 1); by
0 0
 
9a), this implies [a × 0, a × b) and [a × 0, a × b ) have order type [0, 1). Hence, by (a),
0 0
 
Y  = [a × b, a × b ) has order type [0, 1). Let f  : [0, 1) →  Y  be the order isomorphism.
We claim f   is continuous. First, since Y  is an interval, it is convex, and so by Theorem
16.4 the order topology on Y  is the same as the subspace topology on Y  inherited from
L. Then, for any basis set A = (c d, c d )  Y  , f −1 (A) = (f −1 (c d), f −1 (c d ))
× × ⊆ × ×
since f  is an order isomorphism, and moreover this preimage is open. Also, for any
basis set B  = [a b, c d )  Y  , f −1 (B) = [f −1 (a b), f −1 (c d )), which is again
× × ⊆ × ×
open. Thus, f   is continuous. Finally, if we define

F (x) =
 f (x) if  x  [0, 1)
∈
× a b if  x = 1
 
we have a continuous path F  : [0, 1] →  [a × b, a × b ] by the pasting lemma (Theorem
18.3), and so L  is path connected.
×
Proof of  (e). Let a b be a point in L. Since S Ω [0, 1) does not have a maximal×
element, there is some a  b > a b. Now by (c), there exists an order isomorphism
× ×
f  : [0, 1)  [a0 0, a b ). Restricting f  to (0, 1), we get another order isomorphism
→ × ×
 →  × ×
f  : (0, 1) (a0 0, a b ). The set [a0 0, a b ) is open in S Ω [0, 1), and so
 × ×  ×
(a0 0, a b ) is open in L, and is a neighborhood of  a b.
× × ×
We claim (a0 0, a b ) is homeomorphic to (0, 1). We already have a bijection
× ×
that is continuous by the same argument as in (d), and so it suffices to show f  is

25
  ⊆
open as well. But if (x, y)  (0, 1) is a basis set, then f (x, y) = (f (x), f (y)) since f 
is an order isomorphism, and moreover open since the topology on (a0 0, a b ) is
× ×
the order topology.
Proof of  (f ).   Suppose L  could be imbedded in Rn ; then, every subspace of  Rn has a
countable basis by Example 30.1, and since L is homeomorphic with such a subspace,
it also has a countable basis. Now, since X  = (S Ω ×{ } \{ × }
0 ) a0 0  is a convex subset
of  L, the subspace topology on X   is the same as the order topology by Theorem
16.4. Thus, the intersection of the countable basis for L with X  forms a countable
basis by Theorem 30.2. This implies that there is a countable subset  Y  of  X  that is
dense in X  by Theorem 30.3. By Theorem 10.3, though, this subset Y  has an upper
bound x in X . Thus, = (x, Ω)  X  Y   , and so the closure of  Y  is not all of  X , a
 ∅   ⊆  \
contradiction.

25 Components and Local Connectedness

Exercise 25.2.
(a) What are the components and path components of  Rω (in the product topology)? 
(b)   Consider  Rω in the uniform topology. Show that  x and  y   lie in the same 
component of  Rω if and only if the sequence 

x − y = (x − y , x − y , . . .)
1 1 2 2

is bounded.
(c)   Give  R ω the box topology. Show that  x  and  y  lie in the same component of  R ω
−
if and only if the sequence  x y  is “eventually zero.”

Solution for  (a).  By Exercise 24.8(a), Rω is path connected, for Theorem 19.6 is not
limited to finite product topologies. Thus, Rω is the only path component, and so
Rω is the only component as well since path connected =   connected.
⇒
Proof of  (b).   We first define ϕ : x x  y . We recall that since ρ(ϕ(x), ϕ(z)) =
 → −
ρ(x, z), by Exercise 21.2 ϕ is an isometric imbedding that is moreover surjective (the
−
preimage of any z is z  +  y ), ϕ   is a homeomorphism. Thus, x  y  is in the same
component as 0 if and only if  x  is in the same component as y, for ϕ, ϕ−1 do not
modify the topology of  R ω .
It therefore suffices to check the case y   = 0. Suppose x  is bounded; then, we
 →
define f  : [0, 1] Rω where f (t) = (x1 t, x2 t , . . .). This is continuous since given
 ⊇
 > 0, B(f (t), ) {| |}
f (B(t,/ sup xn )), where sup xn < {| |}  ∞   by boundedness

26
of  x. Thus, f   connects 0 and x, i.e., they are in the same path component, and
therefore the same component by Theorem 25.5.
Conversely, recall by Exercise   23.8   that we have the separation Rω = A  B,  ∪
where A is the set of bounded sequences and B  is the set of unbounded sequences of 
reals. If  x  is unbounded it is in B and so is not in the same component as 0.
 ≥
Proof of  (c). x  “eventually zero” here means that xi  = 0 for all i  N  for some N .
Note by the same argument as in (b), it suffices to consider the case  y  = 0.
Suppose first that x  is not eventually zero. Define the function f  = (f n ), where
f n (a) = na/ xn if  xn = 0, and a otherwise. f  is continuous since each f n  is continuous
| |    
since it is linear, and so if  f n−1 (U n ) = V n , we have f −1 ( U n ) = V n . Note that
−1
this is a bijection since each component has an inverse f n (a) = xn a/n if  xn = 0,
 | |  
and a  otherwise, and moreover since the inverse is continuous since each component
is linear, we have a homeomorphism f :  R ω → Rω . Since there are infinitely many  n
such that x n = 0, and so infinitely many n such that f n (xn ) = n, we have that  f (x) is
 
unbounded, and thus, by the separation of  Rω in the box topology in Example 23.6,
we have that f (x) and 0 are in different components. Since f  is a homeomorphism,
this implies  x  and 0 are in different components as well.
Conversely, suppose x   is eventually zero. Then, xn  = 0 for all n N   for some
 ≥
N , and so x ∈ RN  × { } × { } × ··· ⊆
0 0 Rω ; this subspace is homeomorphic to
RN . Since RN  is connected by Theorem 23.6, we see that x  and 0 are in the same
component.

27 Compact Subspaces of the Real Line

Exercise 27.4.  Show that a connected metric space having more than one point is 
uncountable.

Proof. Let X  be a connected metric space with the metric d, and let x0 , x1  ∈
X 
be distinct. Let d(x0 , x1 ) = r, and define f (x) = d(x0 , x). f   is continuous by the
discussion on p. 175. We see that f (x0 ) = 0, f (x) = r, and so by the intermediate
value theorem (Theorem 24.3), f (X )  [0, r], i.e., f  maps onto [0, r].
⊇
Now suppose X   is countable. Then, by Theorem 7.1 there exists a surjective func-
tion g :  Z +  X , and so f  g : Z+  f (X ) maps onto [0, r], which is a contradiction
 → ◦  →
since [0, r] is uncountable by Corollary 27.8.

29 Local Compactness
Exercise 29.4.  Show that  [0, 1]ω is not locally compact in the uniform topology.

27
Proof.   Suppose X  = [0, 1]ω is locally compact, and in particular at 0. Then, there
  ∩ ⊆
exists C   compact that contains a neighborhood U   0. There exists X  Bρ (0, )  U ;
we see that 0, /3 ω  X  Bρ (0, ). 0, /3 ω is closed since
 { } ⊆  ∩  { }
ω
  
{0, /3} = {0, /3} = {0, /3} = {0, /3} = {0, /3} ω

in the product topology by Theorem 19.5, which is finer than the uniform topology,
and so it is compact by Theorem 26.2 since it is a closed subset of  C   compact, i.e.,
limit point compact by Theorem 28.2.
We claim this is a contradiction. Consider x  X , and the ball X  Bρ (x, /9).
 ∈  ∩
Note that the distance between any two distinct points of  0, /3 ω is /3, and so
 { }
∩ ∩
since the diameter of  X  Bρ (x, /9) is at most 2/9, X  Bρ (x, /9) contains at most
one point of  0, /3 ω . Thus, 0, /3 ω contains no limit points, and so is not limit
 { }  { }
point compact, a contradiction.

Exercise 29.8.  Show that the one-point compactification of  Z +  is homeomorphic to
the subspace  0 1/n  n Z+ of  R.
 { } ∪ { | ∈ }
Proof. Let K  = {1/n | n ∈  Z }. Let f  :  R  →  R  such that f (x) = 1/x; this is a
+ + +
homeomorphism since it is continuous and is its own inverse. By Theorem 18.2(d)
and 18.2(e), f  : Z → f (Z ) = K   is continuous, and again is a homeomorphism
+ +
since it is its own inverse. Now consider  Y  = {0} ∪ K , which is closed and bounded
and therefore compact by Theorem 27.3, and Hausdorff by Theorem 17.11. Since

K  = {0}  by Example 17.8, we know Y  is the one-point compactification of  K . If 
X  = { p} ∪ Z  is the one-point compactification of  Z , and letting g : p → 0 ∈ Y  ,
+ +
which is clearly continuous, the function h : X  → Y   defined by the pasting lemma
(Theorem 18.3) applied to f, g is also continuous, and has continuous inverse defined
−1 −1
by the pasting lemma applied to  f  , g , and so is a homeomorphism X  ↔  Y  .

4 Countability and Separation Axioms

30 The Countability Axioms
Exercise 30.4.  Every compact metrizable space  X  has a countable basis.
Proof.   For given n Z+ , we have an open cover of  X  by B (x, 1/n) for each x  X ;
∈ A ∈
 A
since X  is compact, let n  be the finite subcover. n n  is countable since it is the
⊆
countable union of finite sets; we claim it is a basis for X . Let U  X   open and
 ∈  ⊆
x  U . Since X   is metrizable, there exists  B (x, δ )  U  for some δ > 0. Let N   such

28
 A  ⊆
that 2/N < δ . Since N  covers X , there exists B(y, 1/N )  x. B(y, 1/N )  B(x, δ ),
 ∈  ≤  ≤
for if we choose z   B(y, 1/N ), d(x, z )  d(x, y) + d(y, z )  1/N + 1/N < δ . Thus,
A
∈ ⊆
x  B(y, 1/N )  U , and so n n  is a countable basis by Lemma 13.2.

Exercise 30.5.
(a)  Show that every metrizable space with a countable dense subset has a countable 
basis.
(b)  Show that every metrizable Lindel¨ 
of space has a countable basis.

Proof of  (a). Let X  be a metrizable space and A a countable dense subset. We claim
that the set of open balls in  X  below is a basis for X :

B  := {B(a, 1/n) ⊆ X  | a ∈ A, n ∈ N}.

Note B   is countable since is in bijection with  A × N. So let x be contained in an open
subset U  ⊆ X ; since X   is metrizable, x ∈ B(x, ) ⊆ U   for some small . Let n be
such that 1/n < /2. Then, since A is dense, some a ∈  A  is contained in B(x, 1/n),
and conversely x ∈ B(a, 1/n). By the triangle inequality, x ∈ B(a, 1/n) ⊆ U , so by
Lemma 13.2 we are done.
Proof of  (b). Let X   be a metrizable space. Then, the set of open balls


B   := {B(x, 1/n) ⊆ X  | x ∈ X }
n

is an open cover of  X   for each n ∈ N; since X   is Lindelöf, it has a countable

subcover  B  . We claim B  :=
n N B   is a basis for X ; note it is countable since it is
n∈ n
a countable union of countable sets. So let x ∈  B(x, ) ⊆  U  as before, and let n such
 
that 1/n < /2. Then, there is some x ∈  X  such that B(x , 1/n) ∈ B   contains x.
n

By the triangle inequality,  x ∈  B(x , 1/n) ⊆  U , so by Lemma 13.2 we are done.

Exercise 30.8.  Which of our four countability axioms does  Rω in the uniform topol-
ogy satisfy? 

Solution. Rω is first countable since it is metrizable (see p. 130 and Example 30.2),
but is not second countable by Example 30.2. By Exercise 30.5, we then see that Rω
does not have a countable dense subset, and is also not Lindelöf.

Exercise 30.9. Let  A be a closed subspace of  X . Show that if  X  is Lindel¨ 
of, then 
A is Lindel¨ 
of. Show by example that if  X  has a countable dense subset, A need not 
have a countable dense subset.

29
Proof. X   is Lindelöf if and only if a collection of closed subsets of  X  with empty
intersection has a countable subcollection with empty intersection by taking comple-
C
ments in Theorem 30.3(a). Now suppose we have a collection  of closed subsets of  A
with empty intersection; it is then also a collection of closed subsets of  X  with empty
intersection by Theorem 17.3 since A  is closed, and so has a countable subcollection
with empty intersection since  X  is Lindelöf. Thus, A  is also Lindelöf.
Now let X  = R2 . We see Q2 is countable, and is dense in X  since if we take
 ∈ 
x X   and a neighborhood U  x, there exists a basis element [a, b)  [c, d)
2 2
×  ⊆ U 
containing x, which intersects  Q by the fact that (a, b) (c, d) Q =  since  Q is
× ∩ ∅
dense in R. Thus, X  has a countable dense subset; we claim that L = x ( x) { × −  |
x ∈ } R   is a closed subspace of  X  that does not have a countable dense subset.
 ∈  \
L  is closed since if (x1 , x2 ) X   L, then letting d = x1  + x 2 , the basis element
− −
[x1 d/3, x1  + d/3), [x2 d/3, x2  + d/3) does not intersect L. But then, L has the
 { − } ∩ ×−
discrete topology since (x, x)  = L [x, b) [ x, d) is open in L. Thus, if  A  L, ⊆
A =  A  by discreteness, and so A =  L  is true if and only if  A =  L. Thus, L has no
countable dense subset.

Exercise 30.17.   Give  Rω the box topology. Let  Q∞ denote the subspace consisting 
of sequences of rationals that end in an infinite string of  0’s. Which of our four 
countability axioms does this space satisfy? 

Proof.  We claim Q∞ is not first countable, and therefore not second countable. Sup-
pose we have a countable basis U i  at  0  = (0, 0, 0, . . .) Qω . Let V  j  π j (U  j ) open
 { } ∈ 
in  Q  with the subspace topology induced by  R . Then, the neighborhood  j V  j  0  
does not contain any U i , so U i is not a basis and Q∞ is not first or second countable.
{ }
We now show Q∞ has a countable dense subset. For, Qn =  Q n ×{ }×{ }
0 0 are
countable since they are finite products of countable sets, and so their countable

union Q∞ = Qn is also countable. Thus, Q∞ is countable and so is a countable
dense subset of itself.
We now show Q∞ is Lindelöf. Suppose   is an open covering of  Q∞ . Then, since
 V 
Q∞ is countable, choosing for every x Q∞ one element V 
 ∈   such that x V  ,
∈ V   ∈
∞
we get a countable subcover of  Q .

31 The Separation Axioms

Exercise 31.3.  Show that every order topology is regular.

Proof. Let X  be an ordered set with the order topology. X  is Hausdorff and therefore
T 1  by Theorem 17.11. It therefore suffices to show the condition in Lemma 31 .1(a).

30
  ∈
So let x X   and let U   be an open neighborhood of  x; we will construct a basis
element V  of the order topology such that  x  V  and V   U . ∈ ⊆
Suppose x is neither the smallest nor largest element of  X . Then, x  (a, b)  U  ∈ ⊆
for some basis element (a, b) of the order topology. If (a, x), (x, b) are nonempty,
then let V  = (u, v) where u  ∈
(a, x) and v  ∈
(x, b). If (a, x) = , let u = a, so  ∅
 ∅
that V  = (u, v) = [x, v); if (x, b) = , let v = b, so that V  = (u, v) = (u, x]. Then,
∈ ⊆
x  (u, v) and (u, v)  (a, b)  U . ⊆
Now suppose x is the smallest (resp. largest) element of  X . Then, x  [x, b)  U  ∈ ⊆
(resp. x  (a, x]  U ) for some basis element [x, b) (resp. (a, x]) of the order topology.
∈ ⊆
Now if (x, b) (resp. (a, x)) is nonempty, then let V  = [x, v) (resp. (u, x]) where
v  (x, b) (resp. u  (a, x)). On the other hand, if (x, b) (resp. (a, x)) is empty, then
 ∈ ∈
{}
let v = b  (resp. u = a), so that V  = x . We then have  x  V  and V   U . ∈ ⊆
{}
If  x is the smallest and largest element of  X , then X  = x is trivially regular.

32 Normal Spaces
Exercise 32.1.  Show that a closed subspace of a normal space is normal.
Proof.   Suppose Y   is our closed subspace of our normal space X , and A, B Y  ⊆
disjoint and closed. By Theorem 17.3, A, B are closed in X . Let U, V  be a separation
 ∩  ∩
of  A, B in X . Then, Y  U, Y  V   separate A, B in Y   , and so Y   is normal.
Exercise 32.3.  Show that every locally compact Hausdorff space is regular.
Proof. Let X  be a locally compact Hausdorff space; in particular it is  T 1 . Let x  X  ∈

with neighborhood U  x. By Theorem 29.2, there exists a neighborhood V  x 
⊆
such that V   U . By Lemma 31.1(a), X  is then regular.
Exercise 32.4.  Show that every regular Lindel¨ 
of space is normal.
Proof. Let A, B be disjoint closed subsets of  X  regular and Lindelöf. For all x  A, ∈
there exists a neighborhood  U 
x  disjoint from B. By regularity, there exists a
 ⊆  ⊆ U 
neighborhood U   U    containing x; since these U   cover A, and A  is Lindelöf by
 §  { }
Exercise 30.9, there exists a countable subcover U i  such that U i B =  for all i. ∩ ∅
Similarly, we can construct a countable subcover V i  of  B  such that V i A =  for
 { } ∩ ∅
all i. By the exact same argument as in the proof of Theorem 32 .1, then, the sets

U  =
  \    \ 
U n
n

V i , V  = V n
n
U i
n∈Z+ i=1 n∈Z+ i=1

are open and U    

⊇ A, V  ⊇ B, U  ∩ V  = ∅, and so X  is normal.
31
Exercise 32.5. Is  Rω normal in the product topology? In the uniform topology? 
Proof.   Since Rω is metrizable in the product topology by Theorem 20.5 and in the
uniform topology by definition on p. 124, both are normal by Theorem 32.2.

33 The Urysohn Lemma

Exercise 33.1.  Examine the proof of the Urysohn lemma, and show that for given 
r,  −
f −1 (r) = U  p U q ,
 p>r q<r

 p, q   rational.

Proof. . Suppose x  f −1 (r), i.e., x  U r  by definition, and x /  U q  for all q < r by
⊆ ∈ ∈ ∈
∈  ⊆
definition in Step 3. Then,  x  U r  U  p  for all p > r by construction in Steps 1, 2.
⊇. Suppose x  ∈  −
 p>r U  p
 ∈  ⊆
q<r  U q . This implies x  U  p  U  p  for all p > r, and
 ≤ ∈  ≥
so f (x) r  by Step 4(1), and also x / U q  for all q < r, and so f (x) r  by Step
4(2). Thus, f (x) = r.

34 The Urysohn Metrization Theorem

Exercise 34.3. Let  X  be a compact Hausdorff space. Show that  X  is metrizable if 
and only if  X  has a countable basis.
⇒
Proof. . X  is compact and metrizable, hence second countable by Exercise  30.4.
⇐. X   is compact and Hausdorff, and so X   is regular by Theorem 32.3. X 
is second countable as well, and so by the Urysohn metrization theorem (Theorem
34.1), X   is metrizable.

Exercise 34.5. Let X  be a locally compact Hausdorff space. Let  Y  be the one-point 
compactification of  X . Is it true that if  X  has a countable basis, then  Y  is metrizable? 
Is it true that if  Y   is metrizable, then  X  has a countable basis? 

Solution. If  Y  is metrizable, then it is second countable by Exercise 30.4. X   is then

second countable by Theorem 30.2.
Now suppose X  has a countable basis = Bi . We claim that   with sets of 
 B   { }  B 
the form Y   \  Bi , where the union is finite and the closure is taken in Y   , form a
basis of  Y   ; call this larger basis + . The Bi
 B  ∈ B    are open by Lemma 16.2, and
the Y   \ Bi  by Theorem 26.3. B  +
 B 
is countable since   is countable and there are
countably many Y   \  Bi  by Exercise 7.5( j).

32
 Now recall that we have two types of open sets of  Y   , sets U   open in X , and
 \
sets Y  C  for C  compact in X , by construction in Theorem 29.1. By Lemma 13.2,
it suffices to show that for each x   in an open set, we can find an element of  +  B 
containing x properly contained in the open set. So first consider the U . For every
 ∈
x U , there exists Bi  ∈ B   ∈  ⊆
  such that x Bi  B 
U   since   is a basis of  X . Now
\  ∈ \  ∈  \  ∩ 
consider the Y   C  and x Y   C . If  x X , then Y   C   X  x  is open in X 
by Theorem 29.1 (which says X  is a subspace), and so the previous argument for
 ⊆  ∞  {∞}  \
U   X  applies. It remains to show the case x = , where  = Y  X . For each
y C , we can find a neighborhood V  y in X   such that V 
 ∈  ⊆ \{ } X  = Y  x by
 ⊆
local compactness. Since V   is open in X , there exists some Bi V   containing y.

These Bi  cover C  and so there is a finite subcover by compactness C 
   ⊆ Bi ; also,
+
∈  ⊇
x / Bi  C . Thus, we have x  Y   ∈  \  ⊆  \ B 
Bi  Y  C . is therefore a basis for Y 
by Lemma 13.2.
Since Y  is compact and Hausdorff, it is normal by Theorem 32.3, and in particular,
Y  is regular. Finally, since Y   also has a countable basis, it is metrizable by the
Urysohn metrization theorem (Theorem 34.1).

36 Imbeddings of Manifolds
Exercise 36.1.  Prove that every manifold is regular and hence metrizable. Where 
do you use the Hausdorff condition? 

Proof. Let X   be our m-manifold. We first show X   is locally compact. Let x  X  and ∈

 
a neighborhood U   x be given. Since X  is a manifold, there exists a homeomorphism
 →
f  : U  f (U ) Rm . Since Rm is locally compact by Example 29.2, there exists a
 ⊆
⊆ ⊆
neighborhood V   f (U ) of  f (x) such that V  is compact and V   f (U ) by Theorem
29.2. Then, x  f −1 (V  )  f −1 (V  )  U . But then, f −1 (V   ) is compact and therefore
∈ ⊆ ⊆
closed by Theorems 26.3 and 26.5, and so f −1 (V  )  f −1 (V  ) since the closure of a set
⊆
is the intersection of all closed sets containing it, and moreover  f −1 (V  ) = f −1 (V  ) by
Theorem 18.1. Finally, we have x  f −1 (V  )  f −1 (V  )  U , with f −1 (V  ) compact,
 ∈  ⊆  ⊆
and so X   is locally compact by Theorem 29.2.
Now since X   is locally compact and Hausdorff, X   is regular by Exercise   32.3.
Since X   is regular and has a countable basis, it is metrizable by the Urysohn metriza-
tion theorem (Theorem 34.1). Note we used that X   is Hausdorff in showing X  is
regular, for the characterization of local compactness, and the assertion that compact
implies closed.

Exercise 36.5.   The Hausdorff condition is an essential part of the definition of a 

manifold; it is not implied by the other parts of the definition. Consider the following 

33
space: Let  X  be the union of the set  R −{ }  { }
0  and the two-point set   p, q  . Topologize 
X   by taking as basis the collection of all open intervals in  R   that do not contain 
−  ∪ { } ∪
0, along with all sets of the form  ( a, 0)  p  (0, a)   and all sets of the form 
−
( a, 0) ∪{ }∪
q  (0, a), for  a > 0. The space  X  is called the  line with two origins .
(a)  Check that this is a basis for a topology.
 −{ }
(b)  Show that each of the spaces  X   p and  X   −{ } q   is homeomorphic to R.
(c)  Show that  X   satisfies the  T 1  axiom, but is not Hausdorff.
(d)  Show that  X  satisfies all the conditions for a  1-manifold except for the Haus-
dorff condition.

Proof of  (a).   For any x ∈ −

X , for a   large enough either ( a, 0)  ∪ { } ∪
 p  (0, a) or
−
( a, 0) ∪{ }∪
q   (0, a) contains x. Moreover, if we have two basis elements B1 , B2 ,
their intersection is either empty, already another basis element, or a set of the form
− ∪ ∈ ∩ ∩
( a, 0) (0, a), we have a basis for X , for if  x  B 1 B2 , B 1 B2 , or in the last case
−
choosing ( a, 0) or (0, a), would be a basis element containing x that is contained in
the intersection.
 \ { } →
Proof of  (b). Let f  : X  q   R  be defined such that x  →  →
0 if  x =  p, and x  x
otherwise. Clearly f   is a bijection; it suffices to show it is continuous and open.
A basis element not containing p, q   maps to a basis element not containing 0 by
definition and vice versa. A basis element of the form ( a, 0)  p  (0, a) maps
− ∪{ }∪
− −
to ( a, a), and in the other direction, an open interval ( a, b) for a, b > 0 maps to
−
( a, 0)  p∪{ }∪ ∪
(0, a) (0, b) if  a < b, and similarly if  a > b. Thus, f  is a homeomor-
 \{ } →
phism. Note g : X   p R  is also a homemorphism by the same argument.
{} ∈  { } { }
Proof of  (c).  To show X  is T 1 , it suffices to show x is closed for all x  X .  p , q 
\ { } −∞ ∪{ } ∪ ∞
are closed since X   p  = ( \{ } −∞ ∪{ }∪ ∞
, 0) q  (0, ), X  q   = ( , 0)  p (0, )
{}    \ { } −∞ ∪
are open. x  is closed for x = p, q   since if  x < 0, X  x = ( ∪
, x)  (x, 0)
{ }∪ ∞
 p, q  (0, ) is open, and likewise for if  x > 0. Note X  is not Hausdorff since any
{} {}
neighborhood U  of   p intersects any neighborhood V  of  q  , since U, V  must contain
 ∩ − ∪
basis elements B p , Bq   that contain p, q  respectively; however, B p Bq = (( a, 0)
{ }∪
 p ∩ − ∪{ }∪ − ∪
(0, a)) (( b, 0)  p ∅ { }
(0, b)) = ( c, 0) (0, c) =  for  c = min a, b .
Proof of  (d).  We claim the basis elements with rational end points form a countable
basis. For any (a, b) basis element containing x, we can find a < r < s < b  such that
x  (r, s)  (a, b); for any ( a, 0)  p
 ∈  ⊆ − ∪{ }∪ (0, a) basis element containing x, we can
∈ − ∪{ }∪ ⊆− ∪{ }∪
find 0 < r < a such that x  ( r, 0)  p (0, r)  ( a, 0)  p (0, a). Thus they
form a basis by Lemma 13.2.
 
Now for any x = p, q , we see that there is a neighborhood U  of  x  not contain-
ing p, q , and U   is homeomorphic to a neighborhood in R by (b). For x = p, q ,

34
any neighborhood of  x will contain a basis element ( a, 0)− ∪ {x} ∪ (0, a), which is
−
homeomorphic to ( a, a) by (b).

Part II
Algebraic Topology
9 The Fundamental Group
51 Homotopy of Paths
Exercise 51.1.   Show that if  h, h : X  Y   are homotopic and  k, k : Y 
→ → Z  are 
homotopic, then  k h and  k  h are homotopic.
◦ ◦
Proof. Let F   be the homotopy between h, h , and G the homotopy between k, k .
Let H : X  1 ×  → Z  where H (x, t) = G(F (x, t), t). Then H (x, 0) = G(F (x, 0), 0) =
G(f (x), 0) = (k h)(x), H (x, 1) = G(F (x, 1), 1) = G(f (x), 1) = (k  h )(x).
◦ ◦
It remains to show H   is continuous. H   is the map (x, t) (F (x, t), t)
→ →
G(F (x, t), t); since G  is already continuous and the composition of continuous func-
→
tions is continuous, it suffices to show (x, t)  (F (x, t), t) is continuous. But this is
clear since this map is continuous in each coordinate in the codomain.
Exercise 51.2.  Given spaces  X  and  Y   , let  [X, Y  ] denote the set of homotopy classes 
of maps of  X   into Y  .
(a) Let I  = [0, 1]. Show that for any  X , the set  [X, I ]  has a single element.
(b)   Show that if  Y  is path connected, the set  [I, Y  ] has a single element.
Proof of  (a). Fix f 0 : X  →
I . For arbitrary f  : X  → I , f  
f 0   by straight-line
homotopy since any straight line in I  is contained in I . Thus, since f   was arbitrary,
{ }
[X, I ] = [f 0 ] .
 →  →
Proof of  (b). Let f  : I   Y  and y = f (0). Let g : I   Y  be the constant map with
 ∈
value y Y . Define F  : I  I   ×  → −
Y  by F (s, t) = f (s(1 t)). Since F (s, 0) = f (s)
   ∈
and F (s, 1) = g(s), we see F  is a homotopy f   g. Now fix y0  Y   , and let ρ be a
 ×  →
path connecting y, y0 . Define H : X  I   Y   such that H (s, t) = ρ(t). Then, if  f 0
 ∈  
is the constant map with value y0  Y  , H  is a homotopy g  f 0 . By transitivity of 
  { }
homotopy, we see f   f 0 , and so since f  was arbitrary, [X, Y  ] = [f 0 ] .
Exercise 51.3.  A space  X   is said to be   contractible if the identity map iX  : X 
 → X 
is nulhomotopic.

35
 (a)  Show that  I  and  R  are contractible.
(b)  Show that a contractible space is path connected.
(c)   Show that if  Y  is contractible, then for any  X , the set  [X, Y  ]   has a single 
element.
(d)  Show that if  X  is contractible and  Y  is path connected, then  [X, Y  ] has a single 
element.

Proof of  (a).  Since both I  and R are convex, we see that any two maps are homotopic
by straight-line homotopy as on pp. 324–325 since then the straight lines are fully
 
contained in convex set. In particular,  i X   f  for any constant map  f .
Proof of  (b). If  X  is our contractible space, and  F  is our homotopy between i X  and
 ∈  →
f 0  for  f 0  a constant map with value x 0  X , the map ρ : I   X  where ρ(t) = F (x, t)
∈
is a path connecting x, x0  for any x  X .
→
Proof of  (c). Let g0 : Y  Y   be the constant map with value y0 ∈Y ; we have
iY   →
g0 . Now define f 0 : X  Y   such that f 0 (x) = y0 . Then, for any f  : X   →
Y  ,
 ◦   ◦
we have f  = iY   f  g0  f  = f 0   by Exercise  51.1, and so since f  was arbitrary,
{ }
[X, Y  ] = [f 0 ] .
Proof of  (d). Let g0 : X  → X   be the constant map with value x0 ∈X ; we have
   → ◦   ◦
iX   g 0 . Now define y = f (x0 ). Then, for any f  : X   Y  , we have f  = f  iX   f  g0 ,
 ∈  →
which is the constant map at y. Fix y 0  Y   with f 0 : X   Y  the constant map with
 ∈ ×  →
value y 0  Y  , and let ρ  be a path connecting y, y0 . Define H : X  I   Y   such that
 ◦
H (x, t) = ρ(t). Then, H  is a homotopy between f   g0 and f 0 . By transitivity of 
  { }
homotopy, we see f   f 0 , and so since f  was arbitrary, [X, Y  ] = [f 0 ] .

52 The Fundamental Group

Exercise 52.4. Let  A ⊂
X ; suppose  r : X  →
A   is a continuous map such that 
r(a) = a   for each  a  ∈
A. (The map r   is called a   retraction of  X   onto A.) If 
 ∈
a0  A, show that 
r∗ : π1 (X, a0 ) −→
 π 1 (A, a0 )
is surjective.

→ ◦
Proof.   Letting ι : A   X   be the inclusion map, we see r ι = idA , and so by Theorem
◦ ◦
52.4, r ∗ ι∗  = (r ι)∗  = (idA )∗  = idπ1 (A,a0) . This implies r ∗  is surjective.

36
53 Covering Spaces
 →
Exercise 53.3. Let  p : E   B  be a covering map; let  B   be connected. Show that if 
 p−1 (b0 ) has  k  elements for some  b 0  B, then  p −1 (b) has  k  elements for every  b  B.
 ∈ ∈
In such a case, E   is called a  k-fold covering of  B.

Proof.   Since p is a covering map, for any b  with  p−1 (b) = j   there exists a neigh-
 | |
borhood U  b such that p−1 (U ) is the disjoint union of  j  open neighborhoods V α
 
homeomorphic to U , for each V α  contains a unique preimage of  b, which must be one
point since U, V α  are homeomorphic. By the same argument, all x  U  are such that ∈
 p−1 (x)  = j. Thus, we can partition B  into disjoint open sets A j  where each x  A j
| | ∈
are such that p −1 (x) = j  (note that j  can be any cardinal since the argument above
 
does not depend on j  being finite), and A k  b 0  is one of the A j ; however, since B is
connected, B = A k , for otherwise the  A j  will be a separation of  B.

Exercise 53.5.  Show that the map of Example  3 is a covering map. Generalize to
the map p(z ) = z n .

Proof.  Note that p(z ) when considered as a map on S 1 as a subset of  R 2 is given by
 p(cos 2πx, sin2πx) = (cos2πnx, sin2πnx). Letting q (x) = (2cos πx, sin2πx) be the
covering R  → S 1 in Theorem 53.1, and r : R R  be the multiplication by n, we
 →
then have the commutative diagram
r
R2 R2
q −1 q

 p
S 1 S 1
Now let x ∈ S 1 , and consider U  = S 1
 \{ }
x . Taking s q −1 (x), we see that
∈
we have a partition q −1 (U ) = i∈Z W i   where W i = (s + i, s + i + 1), and we also
have r −1 (W i ) = ((s + i)/n, (s + i + 1)/n) =: Z i . Letting V i = q (Z i ), we see that

 p−1 (U ) = 0≤i<n V i . Thus, for all i,
r|W i
W i Z i
(q|W i )−1 q|Z i
 p|V i
V i U 
commutes, where the top and vertical arrows are homeomorphisms, and so we have
a homeomorphism between V i and U  for all i.

37
54 The Fundamental Group of the Circle
Exercise 54.1.  What goes wrong with the “path-lifting lemma” (Lemma  54.1) for 
the local homeomorphism of Example  2 of  53?   §
Solution.   Since b0 = (1, 0) cannot be covered by an open set that is evenly covered
by p by Example 53.2, we see that the first step in the proof of Lemma 54.1 fails.
2
Exercise 54.4.  Consider the covering map p : R × R  → R − 0 of Example  6 of 
+
§53. Find liftings of the paths 

−
f (t) = (2 t, 0),
g(t) = ((1 + t)cos2πt, (1 + t)sin2πt),
 ∗
h(t) = f  g.

Sketch these paths and their liftings.

Solution.  We see first that the covering map  p  is the mapping

(x, s) → ((cos 2πx, sin2πx), s) → s(cos 2πx, sin2πx).

Thus, we have the family of liftings

f ˜n (t) = (n, 2 t) −

g̃n (t) = (t + n, 1 + t)

h̃n (t) =
 −
(n, 2 2t) if  t ∈ [0, 1/2]
−
(2t + n 1, 2t) if  t ∈ [1/2, 1]
where n ∈ Z. We omit the sketches.
1 1
Exercise 54.5.  Consider the covering map p × p : R × R → S  × S  of Example  4
 §
of  53. Consider the path 

f (t) = (cos 2πt, sin2πt) × (cos 4πt, sin4πt)

1 1 1 1
in  S  × S  . Sketch what  f   looks like when  S  × S  is identified with the doughnut 
surface  D. Find a lifting  f ˜ of  f  to  R × R, and sketch it.

Solution.  We have the family of liftings f ˜ (t) = (n + t/2, n + t), where n ∈ Z. We
n
omit the sketches.

38
58 Deformation Retracts and Homotopy Type
Exercise 58.1.  Show that if  A is a deformation retract of  X , and  B is a deformation 
retract of  A, then  B  is a deformation retract of  X .
×  → ×  →
Proof. Let F  : X  I   X , G : A I   A be the deformation retractions of  X  onto
A and A  onto B respectively. We claim that

H (x, t) =
 F (x, 2t) if 0  t  1/2
≤ ≤
◦ −
ι G(F (x, 1), 2t ≤ ≤
1) if 1/2  t  1

is a deformation retraction of  X   onto B, where ι : A → X   is the inclusion map.
We see H (x, 0) = F (x, 0) = x, and that if  x  ∈ B, H (x, 1) = ι ◦  G(F (x, 1), 1) =
ι ◦ F (x, 1) = ι ◦ x = x; moreover, H   is continuous by the pasting lemma, since
H (x, 1/2) = F (x, 1) = ι ◦ G(F (x, 1), 0) = ι ◦ F (x, 1) = F (x, 1), and since ι ◦
G(F (x, 1), 2t − 1) is a composition of continuous functions.
Exercise 58.2.   For each of the following spaces, the fundamental group is either 
trivial, infinite cyclic, or isomorphic to the fundamental group of the figure-eight.
Determine for each space which of the three alternatives holds.
(a)  The “solid torus,” B 2 S 1 .
×
(b)  The torus  T  with a point removed.
(c)  The cylinder  S 1 R.
×
(d)   The infinite cylinder  S 1 R. ×
(e) R3 with the nonnegative  x, y, and  z   axes deleted.
The following subsets of  R2 :
(f ) x x  > 1
{ |  }
{ | ≥ }
(g) x x  1
(h) x x  < 1
{ |  }
(i) S 1 (R+ 0)
∪ ×
( j) S 1 (R+ R)
∪ ×
(k) S 1 (R 0)
∪ ×
(l) R2 (R+ 0)
− ×
Remark.  We use Theorem 60.1 to say that π 1 ( ) and −  × commute.
Solution. (a). π1 (B 2 S 1 ) = π 1 (B 2 ) π1 (S 1 ) = π 1 (S 1 ), for we have a deformation
× ×
retraction from B 2 onto 0 .
 { }
 \ { }
(b). π1 (T   p ) is isomorphic to the fundamental group of the figure-eight, for
 \ { }
we have a deformation retraction onto the figure-eight by deforming T   p  in the
following manner:

39
 b b b

a a a a a a a b

b b b

where the last arrow comes from the construction of the torus as the quotient R2 /Z2 .
(c). π1 (S 1 I ) = π 1 (S 1 ) π1 (I ) = π 1 (S 1 ), for we have the deformation retraction
× ×
 { }
from I   onto 0 .
(d). π1 (S 1 R) = π 1 (S 1 ) π1 (R), for we have the deformation retraction from
× ×
R  onto 0 .
 { }
(e). This space retracts onto  S 2  p, q, r
\{ R2 s, t  (where the isomorphism
}  \{ }
is from Theorem 59.3), which retracts onto the figure-eight, and so the fundamental
group is isomorphic to the fundamental group of the figure-eight.
(f ). π1 ( x x > 1 ) = π1 (S 1 ), by a deformation retraction onto a circle of 
{  |   }
radius > 1, whose fundamental group is π 1 (S 1 ) since it is homeomorphic to  S 1 .
(g). π1 ( x x
{ | ≥ }  1 ) = π s (S 1 ), for we have the deformation retraction onto  S 1 .
{ |  } {}
(h). π1 ( x x  < 1 ) = π 1 ( 0 ) = 1, by a deformation retraction onto 0 .  { }
(i). π1 (S 1 (R+ 0)) = π 1 (S 1 ), since we can retract R+  to one point (1, 0)  S 1 .
∪ × ∈
( j). π1 (S 1 (R+ R)) = π 1 (S 1 ), since we can retract R+ R  to the half-circle
∪ × ×
S 1 R+ R.
∩ ×
(k). π1 (S 1 (R 0)) is isomorphic to the fundamental group of the figure-eight,
∪ ×
for we can retract ( R 0) onto the line segment from ( 1, 0) to (1, 0), which gives
× −
a topological space homotopy equivalent to the figure-eight.
(l). π1 (R2  ( R+  0)) = 1, since we can retract onto any arbitrary point p
\  ×  ∈
R2 (R+ 0).
\ ×
Exercise 58.9.  We define the   degree  of a continuous map h : S 1  S 1 as follows:
→
Let b0  be the point  (1, 0) of  S 1 ; choose a generator  γ   for the infinite cyclic group
π1 (S 1 , b0 ). If  x0  is any point of  S 1 , choose a path  α in  S 1  from  b0 to x0 , and define 
γ (x0 ) = α̂(γ ). Then  γ (x0 ) generates  π(S 1 , x0 ). The element  γ (x0 ) is independent of 
the choice of the path  α, since the fundamental group of  S 1 is abelian.
Now given  h : S 1 → S 1 , choose  x0 ∈ S 1 and let  h(x0 ) = x1 . Consider the 
homomorphism 
h∗ : π(S 1 , x0 )
−→  π 1 (S 1 , x1 ).
Since both groups are infinite cyclic, we have 
·
h∗ (γ (x0 )) = d γ (x1 ) ∗
( )

40
 for some integer  d, if the group is written additively. The integer  d   is called the 
degree of  h and is denoted by  deg h.
The degree of  h is independent of the choice of the generator  γ ; choosing the other 
generator would merely change the sign of both sides of  ( ). ∗
(a)  Show that  d is independent of the choice of  x0 .
(b)   Show that if  h, k : S 1  S 1 are homotopic, they have the same degree.
→
◦
(c)  Show that   deg(h k) = (deg h) (deg k). ·
(d)  Compute the degrees of the constant map, the identity map, the reflection map
ρ(x1 , x2 ) = (x1 , x2 ), and the map h(z ) = z n , where  z   is a complex number.
−
(e)   Show that if  h, k : S 1  S 1 have the same degree, they are homotopic.
→
Proof of  (a). Let x0 =  y 0  S 1 , and let h(y0 ) = y 1 . Let α be a path from x0 to y0 ;
   ∈ 
then, β  = h β   is a path from x 1 to y1 . Then, γ (y1 ) = [β  ] γ (x1 ) [β  ], and so
◦ ∗ ∗

h∗ (γ (y0 )) =  h ∗ ([β ] γ (x0 ) [β ]) = [β  ] h∗ (γ (x0 )) [β  ]
∗ ∗ ∗ ∗
   
= [β  ] ∗ d · γ (x ) ∗ [β  ] = d · ([β  ] ∗ γ (x ) ∗ [β  ]) = d · γ (y ).
1 1 1
1
Proof of  (b).   Choose x  ∈  S  and let h(x ) = y , k(x ) = y . By Lemma 58.4, there
0 0 0 0 1
exists a path α  from y to y  such that k  = α̂ ◦ h . Then, we have
0 1 ∗ ∗
∗
deg k · γ (y ) = k (γ (x )) = α̂(h (γ (x )))
0 ∗ 0 0
= α̂(deg h · γ (y )) = deg h ·  α̂(γ (y )) = deg h · γ (y ),
1 1 0

and so deg k = deg h.

1
 ∈ S  and let k(x ) = y , h(y ) = y . Then,
Proof of  (c).   Choose x 0 0 0 0 1

deg(h ◦ k) · γ (y ) = (h ◦ k) (γ (x )) = (h ◦ k )(γ (x ))

1 ∗ 0 ∗ ∗ 0
= h (deg k · γ (y )) = deg k · h (γ (y )) = (deg h) · (deg k)γ (y ),
∗ 0 ∗ 0 1

and so deg(h ◦ k) = (deg h) · (deg k).

1 1
Solution for  (d). Let k : S  → S  be the constant map. Then, we have k (γ (x )) = 0, ∗ 0
and so deg k = 0.
1 1
Let id: S  → S  be the constant map. Then, we have id (γ (x )) =  γ (x ), and so ∗ 0 0
deg id = 1.
1 1 n 1
Let h : S  → S  such that z  → z  . Let p : R  → S  be the standard covering.
 p−1 (b0 ) = 0, which is fixed by h. Let γ  be the loop γ (s) = e2πis , which generates
π1 (S 1 , b0 ); note we can choose γ, b0  in this way by the fact that degree is independent
of choice of generator and x0 . Then, (h γ )(s) = e2πins = p(ns), and so the lift of 
◦
◦ →
h γ  is s  ns, i.e., h ∗ (γ (b0 )) = nγ (b0 ). Thus, deg h = n.
The case for ρ  follows from that of  h, where n = 1. −
41
 Proof of  (e). By (a), we use b0   as our base point, and let h(b0 ) = x0 , k(b0 ) = y0 .
Consider h∗ : π1 (S 1 , b0 )
 → π1 (S 1 , x0 ), k∗ : π1 (S 1 , b0 ) π1 (S 1 , y0 ); by assumption,
 →
· ·
there exists n, such that h ∗ (γ (b0 )) = n γ (x0 ), k∗ (γ (b0 )) = n γ (y0 ).
Now let γ (s) = e 2πis , which generates π1 (S 1 , b0 ) as above. Then, let   h γ, kγ  be
◦ ◦
the lifts of  h γ, k γ   through the standard covering p :  R  S 1 , at h0  p −1 (x0 ), k0
◦ ◦ →  ∈  ∈
 p−1 (y0 ) respectively. Let h1  = hγ (1), k1  = kγ (1).
◦ ◦
 −  −
Now we have h∗ (γ (b0 ) ) = (h1  h 0 )γ (x0 ), k∗ (γ (b0 ) ) = (k1  k 0 )γ (y0 ), and so
h1 h0 = k 1 k0 = n. We want to find a path homotopy from hγ  to kγ . We
− − ◦ ◦
  
first define h
◦ γ  = h ◦ γ − h that both hγ, hγ   have the same end points;
0 + k0  so ◦ ◦
we then see there exists a path homotopy F̃  between them since  R  is contractible.

Now consider h̃(x) = e 2πi(k0−h0) h(x). Then, hγ   is the lift of  h̃ γ  at k 0 . Then,
◦ ◦

 p  F̃  is a path homotopy from hγ  to  
◦ ◦ ◦
k γ . Since this is a path homotopy, it factors
through ( p, id) to get a homotopy  F   from  h̃  to  k.
We now see that there exists a path homotopy G   between h̃ and h   given by
G(x, t) = h(x)e2πis(k0 −h0 ) , and so there exists a path homotopy combining F, G be-
tween h  and k.
Exercise 58.10.  Suppose that to every map h : S n  S n we have assigned an integer,
→
denoted by  deg h and called the   degree of  h, such that 
(i)   Homotopic maps have the same degree.
◦ ·
(ii) deg(h k) = (deg h) (deg k).
(iii)  The identity has degree  1, any constant map has degree  0, and the reflection 
−  −
map ρ(x1 , . . . , xn+1 ) = (x1 , . . . , xn , xn+1 ) has degree  1.
Prove the following:
(a)  There is no retraction  r : B n+1  S n .→
(b) If  h : S n  S n has degree different from  ( 1)n+1 , then  h has a fixed point.
→ −
(c) If  h : S n → S n has degree different from  1, then  h  maps some point  x   to is 
 −
antipode  x.
(d) If  S n has a nonvanishing tangent vector field  v, then  n is odd.
Proof of  (a).   We use polar coordinates to denote B n+1 and S n , i.e., the points in
B n+1 are given by (θ, φ2 , . . . , φn , ρ) and the points in S n by (θ, φ2 , . . . , φn ), where
0  θ < 2π, 0  φ i  < π, and 0  ρ  1. Thus, the inclusion S n   B n+1 is given by
≤ ≤ ≤ ≤ →
(θ, φ2 , . . . , φn )  (θ, φ2 , . . . , φn , 1).
→
Now, suppose a retraction r : B n+1 →S n exists. Then, define the homotopy
H : S n [0, 1]  S n by H (x1 , . . . , xn , t) = r(x1 , . . . , xn , t), which is continuous since
× →
r is. Then, H (x1 , . . . , xn , 0) is constant, hence has degree 0, whereas H (x1 , . . . , xn , 1)
is the identity, hence has degree 1, which contradicts ( i).

42
Proof of  (b).   We return to Cartesian coordinates. Suppose h  has no fixed points.
We first construct a homotopy:

H (x, t) =
− t)h(x) + ta(x) .
(1
 − t)h(x) + ta(x)
(1

Since H   is continuous since it is a composition of continuous functions in both

   − 
variables, it suffices to show (1  t)h(x) + ta(x) = 0 for all x, t. So, suppose
− −
(1  t)h(x) + ta(x) = 0. Then, (1  t)h(x) = tx. Comparing norms, this is only
possible when t = 1/2, and so h(x) = x, which contradicts that h has no fixed points.
Thus, we see that h is homotopic to the antipodal map. Now, since the antipodal
map is the composition of  n +1 reflections, each one reflecting each coordinate of  S n ,
we have that deg(h) = deg(a) = ( 1)n+1 by (i), (ii), and (iii), a contradiction.
−
Proof of  (c).   Suppose h maps no point x  to its antipode  −x. We first construct a
homotopy:
G(x, t) =
−
(1 t)h(x) + tx
.
 −
(1 t)h(x) + tx 
Since G   is continuous since it is a composition of continuous functions in both
  − 
variables, it suffices to show (1 t)h(x) + tx = 0 for all x, t. So, suppose
− −  −
(1  t)h(x) + tx   = 0. Then, (1  t)h(x) = tx. Comparing norms, this is only
−
possible when t = 1/2, and so h(x) = x, which contradicts that h maps no point x
 −
to its antipode x.
Thus, we see that h is homotopic to the identity. Since the identity has degree 1,
we have deg(h) = 1 by (i), a contradiction.
Proof of  (d).  Suppose such a v(x) exists. Then, let h(x) = v(x)/ v(x) ; h(x) then is
 
a map S n  S n . Since v(x) is a tangent vector field, h(x), x  = v(x), x / v(x)  =
→      
0, where ,   is the standard inner product in Rn+1 . Then, there are no points
 · ·
 ±       ±
h(x) = x  since h(x), x = 0 = 1. By the proofs of (b) and (c), h(x) is then
homotopic to the identity and the antipodal map; hence, the identity is homotopic
to the antipodal map. By (i), (iii), and the proofs of (b) and (c), this then implies
that deg 1 = deg( 1)n+1 , and so n  is odd.
−
59 The Fundamental Group of  S n
Exercise 59.1. Let  X   be the union of two copies of  S 2 having a single point in 
common. What is the fundamental group of  X ? Prove that your answer is correct.

43
Solution. Let U  = X  a , V  = X  b  such that a, b are in different copies of  S 2 ;
 \ { }  \ { }
U, V  are open since X  is Hausdorff. Then, since S 2 a , S 2 b  are homeomorphic
\{ } \{ }
to R2 by the proof in Theorem 59.3, we see that there exists a deformation retraction
from X  a  onto S 2 by taking the copy of  S 2 containing a and retracting it into
 \ { }
x0   the intersection of the two copies of  S 2 ; likewise, there exists a deformation
{ }
retraction from X  b  onto S 2 . Thus, both U, V  are simply connected by Theorem
\{ }
59.3. U  V  is path connected since it is homeomorphic to two copies of  R2 adjoined
∩
at x 0 , and so X  is simply connected by Corollary 59.2.

60 Fundamental Groups of Some Surfaces

Exercise 60.2. Let  X  be the quotient space obtained from  B 2 by identifying each 
point  x of  S 1 with its antipode  x. Show that  X   is homeomorphic to the projective 
 −
plane  P 2 .
Proof.   Consider X   as constructed from B 2 identified with the closed upper hemi-
sphere of  S 2 . Let p : S 2→ P 2 , q : B 2 X  be the quotient maps, and π : S 2
→ →B2
the map sending x  to either itself or x if  x /  B 2 ; note this is a quotient map since
 − ∈
U   S 2 open implies π(U ) open, and V   B 2 open implies π −1 (V  ) = V 
 ⊆ ⊆  ∪ −
V   open.
We then have the commutative diagram
π
S 2 B2
 p q
f 
P 2 X 
◦
and since q  π is a quotient map by p. 141, we see  f  is a quotient map by Theorem
22.2. But since for x  X , (q  π)−1 (x) = x, x  P 2 , and moreover any equivalence
∈ ◦ { − }∈
 { − }
class x, x  can be realized as the inverse image of this type, f  is a bijection and
P 2 = (q  π)−1 (x)  x  X  , and so f  is a homeomorphism by Corollary 22.3(a).
{ ◦ | ∈ }
Exercise 60.3. Let p : E  →  X  be the map constructed in the proof of Lemma  60.5.

Let  E  be the subspace of  E   that is the union of the  x-axis and the  y-axis. Show that 
 p|  is not a covering map.
E 

Proof.  Consider the base point x 0 , the center of the figure-eight  X . A neighborhood
 
U   x 0  contains the union V  of open intervals in A and B  which intersect in exactly
x0 . ( p E  )−1 (V  ) is then equal to the union of open intervals around integers on the
{ } | 

x-axis and open intervals around integers on the  y-axis. But none of these intervals
are homeomorphic to V  , since removing an integer in an interval gives two connected
components, while removing its image x 0  in  V  gives four connected components.

44
11 The Seifert-van Kampen Theorem
67 Direct Sums of Abelian Groups
Exercise 67.2.  Show that if  G 1  is a subgroup of  G, there may be no subgroup G2 of 
⊕
G such that  G = G 1 G2 .

Proof. Let G = Z and G1 = 2Z, and suppose that such a  G 2  exists. Then, we have
G = 2Z Z/2Z by Corollary 67.3, and so (0, 1)  G 1 G2 . But since 2(0, 1) = (0, 0),
⊕ ∈ ⊕
⊕
G1 G2  contains an element of order 2 while  G  does not, a contradiction.

Exercise 67.4. The   order  of an element  a of an abelian group G is the smallest 
positive integer  m such that  ma = 0, if such exists; otherwise, the order of  a is said 
to be infinite. The order of  a thus equals the order of the subgroup generated by  a.
(a)  Show the elements of finite order in  G form a subgroup of  G, called its   torsion
subgroup.
(b)   Show that if  G is free abelian, it has no elements of finite order.
(c)  Show the additive group of rationals has no elements of finite order, but is not 
 free abelian.

Proof of  (a).  It suffices to show the torsion elements T (G) of  G is closed under sums
∈
and inverses. If  a, b  T (G) with orders m, n respectively, then  mn(a + b) = n(ma) +
⇒ ∈ − −
m(nb) = 0 = a + b  T (G); likewise, m( a) = (ma) = 0 = ⇒− ∈ a  T (G).
   ∈
Proof of  (b).   Any 0 =  a G is of the form a =

kα aα  where {a }  is our basis of 
α
G and kα  Z . If  ma = 0, then mkα aα  = 0 for all α, which is a contradiction since
 ∈
each a α  also has infinite order.
Proof of  (c).   Suppose Q = aα , aα  a basis. Then for any a α , a α /2 must be of the
    { } 
form aα /2 = kα aα for kα  Z . But then, aα = 2 kα aα , and since aα  is a basis
 ∈
element, we have aα  = 2kα aα . This implies kα  = 1/2, which is a contradiction.

68 Free Products of Groups

∗
Exercise 68.2. Let G = G1 G2 , where  G1 and  G2  are nontrivial groups.
(a)   Show  G is not abelian.
∈
(b) If  x  G, define the   length  of  x to be the length of the unique reduced word in 
the elements of  G1 and  G2  that represents  x. Show that if  x has even length 
(at least  2), then  x  does not have finite order. Show that if  x has odd length 
(at least  3), then  x is conjugate to an element of shorter length.

45
 (c)  Show that the only elements of  G  that have finite order are the elements of  G 1
and  G2  that have finite order, and their conjugates.
   ∈    ∈  
Proof of  (a).   For 1 = g G1 , 1 = h G2 , gh = hg since otherwise we would have
distinct reduced word representations of the same element of  G.
Proof of  (b). If  x  G has even length, then without loss of generality,  x  starts with
∈
an element in G1  and ends with one in G2 , and so we cannot reduce xn to the identity,
and x  has infinite order. If  x G has odd length, then without loss of generality,
 ∈
x = ghg  for g, g  ∈ G1 , h G. Then, g−1 xg = hg g   has shorter length, for g  g
 ∈
reduces to one element in  G 1 .
Proof of  (c).   Suppose x  G  has finite order. Then, by (b) it must have odd length
 ∈
2k + 1. We proceed by induction on k. For k  = 0, we see the length is 1, and so
 ∈
x  G i  for some i, and has finite order in Gi . Now suppose k > 0. Since x has odd
length, x = g −1 yg for g, y  G, y  of shorter length. y has odd length for if not, y  has
 ∈
infinite order by (b), and so x also has infinite order since xn = g−1 yn g, which is a
contradiction. Since y  is of finite order by the fact  gxn g −1 = y n , y  is either equal to
an element of  G i  with finite order or conjugate to one by inductive hypothesis. If the
latter is true, x = g−1 yg = g −1 h−1 zhg = (hg)−1 zhg for h G, and z  Gi   having
 ∈  ∈
finite order for some i. x is therefore conjugate to a finite order element of  G i .

71 The Fundamental Group of a Wedge of Circles

Exercise 71.2.   Suppose  X  is a space that is the union of the closed subspaces  X 1 ,
 ∩  { }  
. . . , X n ; assume there is a point  p of  X   such that  X i X  j =  p for  i = j. Then 
we call  X  the   wedge   of the spaces  X 1 , . . . , X n , and write  X  = X 1  ∨ ··· ∨  X n .
Show that if for each  i, the point  p   is a deformation retract of an open set  W i of 
X i , then  π1 (X, p) is the external free product of the groups  π1 (X i , p) relative to the 
monomorphisms induced by inclusion.
Proof.  By induction, it suffices to show when  X  = X 1 X 2 ; moreover, it suffices to
∨
consider when the X i  are both path-connected since if  C i  are the path components
 ∪
containing p in X i , then π1 (C i , p) = π1 (X i , p) as on p. 332. So, let U  = X 1  W 2
∪
and V  = X 2 W 1 . Then, both U  and V  are path connected since they deformation
 ∩ ∪
retract to X 1 , X 2 , respectively, and  U  V  = W 1 W 2  is moreover simply connected
{}
since it deformation retracts to  p . Thus, by Corollary 70.3, there is an isomorphism
∗ 
π1 (X 1 , p) π1 (X 2 , p)  π 1 (X, p).
Exercise 71.4.  Show that if  X  is an infinite wedge of circles, then  X  does not satisfy 
the first countability axiom.

46
  ∈
Proof. Let p X  be the common point of the circles. Suppose X  has a countable
basis U i at p; we can assume without loss of generality that  U i   U  j if  i < j. For
 { }
∩
each U i , we know that  V ij = U i S  j  is open for any  i, j by the coherence condition,
and is nonempty since p  V ij .
 ∈
Now take V  = V ii ; we claim it is a neighborhood of  p  that does not contain
 ∩
any U i . It is open by coherence since  V  S i  = U i  is open for all i, and contains p  by
 ⊆
construction above. Now suppose U i  V   . Then, this implies U i S  j  V  S  j = U  j ∩  ⊆  ∩
for all j, and in particular when i < j, which contradicts that U i   U  j if  i < j.
Exercise 71.5. Let  S n  be the circle of radius  n in  R2 whose center is at the point 
(n, 0). Let Y  be the subspace of  R2 that is the union of these circles; let  p  be their 
common point.
(a)   Show that  Y  is not homeomorphic to a countably infinite wedge  X  of circles,
nor to the space of Example  1.
 { }
(b)   Show, however, that  π1 (Y, p)   is a free group with  [f n ]   as a system of free 
generators, where  f n  is a loop representing a generator of  π1 (S n , p).
Proof of  (a). Y  is a subspace of  R2 and so it is first countable by Theorem 30.2, and
so Y  is not homeomorphic to X  by Exercise 71.4, which says X  is not first countable.
Denote Z  as the space of Example 71.1; we claim Z  is compact while Y   is not.
First, denoting Dm  as the closed disc of radius 1/m with center at the point (1 /m, 0),
 ∪
we see that Z   D m   is closed since it is a finite union of closed sets Z   D m =  ∪
Dm  ∪  n<m C n . Then, Z   ⊆ 
m (Z   ∪
Dm ) trivially; moreover, since if  x / Z , then
 ∈
∈
x / Dm   for all m   large enough, we have that Z  = m (Z   D m ). Thus, Z  is  ∪
closed. Since Z   is bounded, it is then compact by Theorem 27 .3. On the other
hand, Y   is unbounded hence not compact by Theorem 27.3, and so Y, Z   are not
homeomorphic.
Proof of  (b). Let in : π1 (S n , p)  →
π1 (Y, p) be the homomorphism induced by inclu-
sion, and let their respective images be Gn ; we want to show that the homomorphism
J  J 

∗
∞

n=1
π1 (S n , p) → π (Y, p),
1

 j=1
j
[f mj ] 
→

 j=1
in ([f mj ])j (1)

is an isomorphism.
We first show that the homomorphism (1) is surjective. So, suppose f  : I  Y   →
is a loop in Y . Letting Y N  = Y  ∞
\ { }
n=N +1 (2n, 0) , since I   is compact, the image
f (I ) is also compact, thus bounded by Theorem 27.3, and so f (I )  Y N  for some N . ⊆
Letting rN  : Y  → N 
n=1 S n , we form the deformation retraction H : Y N   I  Y N ,  ×  →
|
where H (x, 0) = idY N  and H (x, 1) = rN  Y N  , by retracting the upper and lower

47
  ◦  ×
semicircles of  S n to p  for all n > N . Then, H   (f   idI ) : I   I   × →
Y N   is a path
 ◦  ◦
homotopy from f  to rN  f . Thus, [f ] = [rN  f ], and Theorem 71.1 implies that
◦ ≤
[rN  f ] is a product of elements of the groups  G n for n  N . Therefore, [f ] is in the
image of the homomorphism (1).
We now show that the homomorphism (1) is injective. Now suppose there is some
non-identity element
w  ∈ ∗
∞

n=1
π1 (S n , p),

whose image through the homomorphism (1) is the identity in π1 (Y, p). Let f  be a
loop in X  whose path-homotopy class is the image of  w. Then, f  is path homotopic to
a constant in X , so by the argument above, it is path homotopic to a constant in some

Y N , and therefore path homotopic to a constant in N  n=1 S n . But this contradicts
Theorem 71.1, which says that the map
  N 

∗
N 

n=1
π1 (S n , p) → π 1
n=1
S n , p

is injective.

73 The Fundamental Groups of the Torus and the Dunce

Cap
Exercise 73.1.  Find spaces whose fundamental groups are isomorphic to the follow-
ing groups. (Here  Z/n denotes the additive group of integers modulo n.)
(a) Z/n Z/m.
×
(b) Z/n1 Z/n2
× ×···× Z/nk .
(c) Z/n Z/m. (See   Exercise  2 of  71).
∗  §
(d) Z/n1 Z/n2
∗ ∗···∗Z/nk .
 ×  ×
Solution.  We use Theorem 60.1 to say π1 (X  Y, x0 y0 ) = π1 (X, x0 ) π1 (Y, y0 ), ×
∨ ∗  { }
and use Exercise 71.2 to say π 1 (X  Y, x0 ) = π 1 (X, x0 ) π1 (Y, x0 ) for x0  = X  Y  . ∩
We denote the n-fold dunce cap by Dn . Note since the Dn  are path connected,
we do not need to specify base points.
Through repeated applications of the above, and by using induction to get (a) ⇒

48
(b), (c) ⇒ (d), we get
π1 (Dn ×D = Z/n × Z/m =⇒
  
k k
Z/ni ,
m) π1 Dni =
i=1 i=1
  k
π1 (Dn ∨D m) = Z/n Z/m =
∗ ⇒ π1 ∗
i=1
Dni =
i=1
k
Z/ni .

12 Classification of Surfaces
74 Fundamental Groups of Surfaces
Exercise 74.3. The   Klein bottle  K  is the space obtained form a square by means 
of the labeling scheme  aba−1 b. Figure  74.11 indicates how  K  can be pictured as an 
immersed surface in  R3 .
(a)  Find a presentation for the fundamental group of  K .
(b)   Find a double covering map p : T  →
K , where  T   is the torus. Describe the 
induced homomorphism of fundamental groups.

Proof of  (a). π1 (K ) = a, b  aba −1 b = 1  by Theorem 74.2.

 | 
×
Proof of  (b).   We consider T  as [0, 1] [0, 1] with the relations (0, y)  (1, y), (x, 0) ∼ ∼
×
(x, 1), and K   as [0, 1] [0, 1] with the relations (0, y)  ∼
(1, 1 y), (x, 0) −
(x, 1).  ∼
 →
Then, define p : T   K  by

 p(x, y) =
 (2x, y) if  x ∈ [0, 1/2],
−
(2x 1, 1 − y) if  x ∈ [1/2, 1].
This is continuous in each region, and agrees on the boundary since p(1/2, y) =
· − · − −
(2 1/2, y) = (1, y) = (0, 1 y) = (2 1/2 1, 1 y) and p(1, 1 y) = (1, 1 y) = − −
(0, y) = p(0, y).
Now recall that π1 (T ) = α, β   αβα−1 β −1 = 1 . Looking at Figures 74.4 and
   | 
74.11, we see p∗ (α) = a2 , p∗ (β ) = b. Since aba−1 b = 1 bab = a, we see that
⇐⇒
2 2
 p∗ (α) p∗ (β ) = a b = babab = ba = p ∗ (β ) p∗ (α).

75 Homology of Surfaces
Exercise 75.3. Let  X  be the quotient space obtained from an  8-sided polygonal region 
P  by pasting its edges together according to the labelling scheme  acadbcb−1 d.

49
 (a)  Check that all vertices of  P  are mapped to the same point of the quotient space 
X  by the pasting map.
(b)   Calculate  H 1 (X ).
(c)   Assuming  X   is homeomorphic to one of the surfaces given in Theorem  75.5
(which it is), which surface is it? 

a
c d

a b

c
d
b−1

Figure 1: Labeling of edges and identification of vertices in  P .

Proof of  (a).  We have the identification of vertices in P  as shown with solid lines
in Figure 1,  found by identifying heads/tails of arrows with the same label. Thus,
all the vertices of  P  are mapped to the same point of the quotient space X  by the
pasting map.
Solution for  (b).  We apply Lemma 77.1 to the labeling scheme acadbcb−1 d repeatedly
to find an equivalent labeling scheme, where the brackets show our decomposition
[y0 ]a[y1 ]a[y2 ]:

[ ]a[c]a[dbcb−1 d] −1
∼ aac dbcb d −1

[aac−1 ]d[bcb−1 ]d[ ] ∼ ddaac bc b −1 −1 −1

[ddaa]c−1 [b]c−1 [b−1 ] −1 −1

∼ c c ddaab −1 −1
b .

Relabeling the edges, which is allowed by p. 460, we have the labeling scheme
aabbccdd. Thus, by Theorem 74.2, the fundamental group is

π1 (X, x0 ) = a,b,c,d  a2 b2 c2 d2 = 1 .

 | 
Finally, the first homology group is
3
H 1 (X )  a,b,c,d | 2a + 2b + 2c + 2d = 0  Z × (Z/2Z).

50
Solution for  (c).   Theorem 75.4 says that H 1 (P n ) = Zn−1 (Z/2Z). Thus, assuming
×
that X   is homeomorphic to one of the surfaces in Theorem 75 .5, we see that X  is
homeomorphic to P 4  = P 2 #P 2 #P 2 #P 2 by (b).

Exercise 75.4. Let  X  be the quotient space obtained from an  8-sided polygonal region 
P   by means of the labelling scheme  abcdad−1 cb−1 . Let  π : P  →
X   be the quotient 
map.
(a)  Show that  π does not map all the vertices of  P  to the same point of  X .
(b)  Determine the space  A = π(Bd P ) and calculate its fundamental group.
(c)   Calculate  π1 (X, x0 ) and  H 1 (X ).
(d)   Assuming  X  is homeomorphic to one of the surfaces given in Theorem  75.5,
which surface is it? 

c
b d

a a

b−1 d−1
c

Figure 2: Labeling of edges and identification of vertices in  P .

Proof of  (a).  We have the identification of vertices in  P  as shown in Figure 2, found
by identifying heads/tails of arrows with the same label, where the solid lines are
one identification and the dashed lines are another. Thus, π does not map all the
vertices of  P  to the same point of  X .

b
a c
x0 x1
d

Figure 3: Sketch of  A = π(Bd P ).

Solution for  (b). Let x 0  be the point identified by the solid lines in Figure  2,  and x 1
the point identified by the dashed lines. Then, using Figure  2, we see that a connects

51
x0  to itself, c  connects x 1  to itself, and b  goes from x 0 to x 1 , while d  goes from x 1 to
x0 . Thus, we have the sketch in Figure  3  for  A = π(Bd P ).
We now want to calculate its fundamental group. First, we see that we can
homotopically retract the segment d into the point x0 , thereby making x0  coincide
with x1 ; the resulting deformation retract is then the wedge sum of three circles.
Thus, π1 (A, x0 )  π 1 (S 1 S 1 S 1 , x0 ) = Z Z Z  by Theorems 58.3 and 71.1.
 ∨ ∨ ∗ ∗
Solution for  (c).  We proceed as in Exercise  75.3(b):

[ ]a[bcd]a[d−1 cb−1 ] −1 −1 −1 −1
∼ aad c b d cb−1
[aad−1 c−1 ]b−1 [d−1 c]b−1 [ ] −1 −1
∼ b b aad −1 −1 −1
c c d
[b−1 b−1 aad−1 ]c−1 [ ]c−1 [d] −1 −1
∼ c c b b −1 −1
aad−1 d
c−1 c−1 b−1 b−1 aad−1 d −1 −1
∼ c c b b −1 −1
aa
c−1 c−1 b−1 b−1 aa ∼ aabbcc
where we cancel  d −1 d in the penultimate step and relabel in the last step as allowed
on p. 460. Thus, our fundamental group is

π1 (X, x0 ) = a,b,c  a 2 b2 c2 = 1 ,
 | 
which has the first homology group
2
 |
H 1 (X ) = a,b,c,d  2a + 2b + 2c = 0   Z × (Z/2Z).
Solution for  (d).   Since H 1 (X ) Z2 (Z/2Z), by Theorem 75.4, we have that X  is
  ×
homeomorphic to P 3 = P 2 #P 2 #P 2 , assuming that X  is homeomorphic to one of 
the surfaces in Theorem 75.5.

52
List of Solved Exercises
I General Topology 3 Exercise 23.11   . . . . . . . . . . . 20
24 Connected Subspaces of the
1 Set Theory and Logic 3 Real Line . . . . . . . . . . . . . 21
7 Countable and Uncountable Sets 3 Exercise 24.7 . . . . . . . . . . . 21
Exercise 7.5   . . . . . . . . . . . . 3 Exercise 24.8 . . . . . . . . . . . 22
Exercise 24.12   . . . . . . . . . . . 22
2 Topological Spaces and Continu- 25 Components and Local Con-
ous Functions 3 nectedness   . . . . . . . . . . . . . 26
13 Basis for a Topology . . . . . . . 3 Exercise 25.2 . . . . . . . . . . . 26
Exercise 13.3 . . . . . . . . . . . 3 27 Compact Subspaces of the Real
Exercise 13.5 . . . . . . . . . . . 4 Line . . . . . . . . . . . . . . . . 27
Exercise 13.6 . . . . . . . . . . . 4 Exercise 27.4 . . . . . . . . . . . 27
Exercise 13.7 . . . . . . . . . . . 4 29 Local Compactness   . . . . . . . . 27
16 The Subspace Topology . . . . . 5 Exercise 29.4 . . . . . . . . . . . 27
Exercise 16.8 . . . . . . . . . . . 5 Exercise 29.8 . . . . . . . . . . . 28
Exercise 16.9 . . . . . . . . . . . 6
17 Closed Sets and Limit Points . . 6 4 Countability and Separation Ax-
Exercise 17.2 . . . . . . . . . . . 6 ioms 28
Exercise 17.3 . . . . . . . . . . . 6 30 The Countability Axioms . . . . 28
Exercise 17.5 . . . . . . . . . . . 7 Exercise 30.4 . . . . . . . . . . . 28
Exercise 17.13   . . . . . . . . . . . 7 Exercise 30.5 . . . . . . . . . . . 29
Exercise 17.16   . . . . . . . . . . . 7 Exercise 30.8 . . . . . . . . . . . 29
18 Continuous Functions . . . . . . . 8 Exercise 30.9 . . . . . . . . . . . 29
Exercise 18.1 . . . . . . . . . . . 8 Exercise 30.17   . . . . . . . . . . . 30
Exercise 18.12   . . . . . . . . . . . 8
31 The Separation Axioms . . . . . 30
19 The Product Topology   . . . . . . 9
Exercise 31.3 . . . . . . . . . . . 30
Exercise 19.6 . . . . . . . . . . . 9
32 Normal Spaces . . . . . . . . . . 31
20 The Metric Topology   . . . . . . . 10
Exercise 32.1 . . . . . . . . . . . 31
Exercise 20.4 . . . . . . . . . . . 10
Exercise 32.3 . . . . . . . . . . . 31
Exercise 20.5 . . . . . . . . . . . 12
Exercise 32.4 . . . . . . . . . . . 31
Exercise 20.6 . . . . . . . . . . . 12
Exercise 32.5 . . . . . . . . . . . 32
Exercise 20.8 . . . . . . . . . . . 13
33 The Urysohn Lemma   . . . . . . . 32
21 The Metric Topology (continued) 15
Exercise 21.1 . . . . . . . . . . . 15 Exercise 33.1 . . . . . . . . . . . 32
Exercise 21.2 . . . . . . . . . . . 16 34 The Urysohn Metrization Theo-
Exercise 21.3 . . . . . . . . . . . 16 rem . . . . . . . . . . . . . . . . . 32
22 The Quotient Topology . . . . . . 18 Exercise 34.3 . . . . . . . . . . . 32
Exercise 22.2 . . . . . . . . . . . 18 Exercise 34.5 . . . . . . . . . . . 32
Exercise 22.4 . . . . . . . . . . . 18 36 Imbeddings of Manifolds   . . . . . 33
Exercise 22.6 . . . . . . . . . . . 19 Exercise 36.1 . . . . . . . . . . . 33
Exercise 36.5 . . . . . . . . . . . 33
3 Connectedness and Compactness 20
23 Connected Spaces  . . . . . . . . . 20
Exercise 23.8 . . . . . . . . . . . 20 II Algebraic Topology 35

53