Help File

The attached pages provide information on some exercises in the text, and they are posted
since the contents may be useful for working some of these exercises. This represents one graduate
student’s efforts, working without authorization from the author or instructor. The quality of the
discussions in the document is poor in some (but not all) cases, and due to its inconsistent quality
the document should not be viewed as a model for writing things up. To summarize, the warning,
“Use at your own risk,” applies.
 Solutions to Algebraic Topology∗
Steven V Sam
ssam@mit.edu

August 1, 2008

Contents
1 The Fundamental Group 1
1.1 Basic Constructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Van Kampen’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Covering Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.A. Graphs and Free Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Homology 7
2.1. Simplicial and Singular Homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2. Computations and Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Cohomology 12
3.2. Cup Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3. Poincaré Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.C. H-Spaces and Hopf Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4 Homotopy Theory 16
4.1. Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.2. Elementary Methods of Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3. Connections with Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1 The Fundamental Group

1.1 Basic Constructions
3. Exercise. For a path-connected space X, show that π1 (X) is Abelian if and only if all basepoint-
change homomorphisms βh depend only on the endpoints of the path h.
Solution. Suppose that for any two paths g and h from x0 to x1 , the isomorphisms π1 (X, x0 ) →
π1 (X, x1 ) given by f 7→ g −1 f g and f 7→ h−1 f h are the same. Now choose f, f 0 ∈ π1 (X, x0 ).
We wish to show that f 0 f = f f 0 . Note that f 0 is homotopy equivalent to a composition gh−1 ,
where g and h are paths from x0 to x1 , for the following reason. We can pick any point y on
the path f 0 and let p be a path from y to x1 . Then the path from x0 to y along f 0 composed
with p is the desired g, and p−1 composed with the path from y to x0 along f 0 is the desired
∗
by Allen Hatcher

1
 1.1 BASIC CONSTRUCTIONS 2

h−1 . However, we know that h−1 f h ' g −1 f g, which can be rewritten as gh−1 f ' f gh−1 . Since
f 0 is homotopic to gh−1 , this gives f 0 f ' f f 0 , so π1 (X, x0 ) is Abelian.
Conversely, suppose that π1 (X, x0 ) is Abelian and let g and h be two paths from x0 to x1 .
Then we get two isomorphisms π1 (X, x0 ) → π1 (X, x1 ) via f 7→ g −1 f g and f 7→ h−1 f h, and
we wish to show these two maps are the same. Note that hg −1 is a loop based at x0 , so is an
element of π1 (X, x0 ). For any f ∈ π1 (X, x0 ), we have hg −1 f ' f hg −1 , which can be rewritten
g −1 f g ' h−1 f h, so the two maps are indeed equal.

6. Exercise. We can regard π1 (X, x0 ) as the set of basepoint-preserving homotopy classes of

maps (S1 , s0 ) → (X, x0 ). Let [S1 , X] be the set of homotopy classes of maps S1 → X, with
no conditions on basepoints. Thus there is a natural map Φ : π1 (X, x0 ) → [S1 , X] obtained by
ignoring basepoints. Show that Φ is onto if X is path-connected, and that Φ([f ]) = Φ([g]) if
and only if [f ] and [g] are conjugate in π1 (X, x0 ). Hence Φ induces a one-to-one correspondence
between [S1 , X] and the set of conjugacy classes in π1 (X), when X is path-connected.
Solution. Choose f, g ∈ π1 (X, x0 ). Ignoring the base point, we will show that f gf −1 is
homotopic to g. Without loss of generality, we may assume that f gf −1 traverses f , g, and f −1
on the intervals [0, 1/3], [1/3, 2/3], and [2/3, 1], respectively. Thinking of S1 as R/Z, we can start
at 1/3 and end at 4/3 (this corresponds to a free homotopy that moves the base point). This
means that f gf −1 is free homotopic to gf −1 f , which is homotopic to g, so conjugacy classes
map into homotopy classes of maps S1 → X. Any homotopy class of maps S1 → X can be
represented by some loop in X. Since X is path-connected, this can be extended to a loop based
at x0 , and such a loop will be mapped by Φ to this homotopy class, so Φ is surjective.
To see that Φ is injective, let f, g ∈ π1 (X, x0 ) be elements that are homotopic if we ignore
base points (i.e., Φ(f ) = Φ(g)). Then there is a continuous map H : [0, 1]2 → X such that
H(0, t) = H(1, t) for all t, and H(t, 0) = f (t) and H(t, 1) = g(t). Let h : [0, 1] → X be defined
by h(t) = H(0, t), so that h keeps track of the basepoint change along H. Then h(0) = H(0, 0) =
f (0) and h(1) = H(0, 1) = g(0), so h ∈ π1 (X, x0 ). We claim that hgh−1 ' f . Write

h(3t)
 if 0 ≤ t ≤ 0
f ' H(t, 0) if 0 ≤ t ≤ 1 ,

 −1
h (3t − 2) if 1 ≤ t ≤ 1

and 
h(3t)
 if 0 ≤ t ≤ 31
hgh−1 ' H(3(t − 13 ), 1) if 13 ≤ t ≤ 23 .

 −1
h (3t − 2) if 23 ≤ t ≤ 1
This observation suggests using the following homotopy H 0 (t, s) : [0, 1]2 → X from f to hgh−1 :

h(3t)
 if 0 ≤ t ≤ 3s
0
H (t, s) = H((2s + 1)(t − 3s ), s) if 3s ≤ t ≤ 1 − 3s .

 −1
h (3t − 2) if 1 − 3s ≤ t ≤ 1

Then H 0 (t, 0) = f (t), and H 0 (t, 1) = hgh−1 , and H 0 (0, s) = H 0 (1, s) = h(0) = x0 , so f and g
come from the same conjugacy class of π1 (X, x0 ), and hence Φ is injective.

16. Exercise. Show that there are no retractions r : X → A in the following cases:
 1.2 VAN KAMPEN’S THEOREM 3

(a) X = R3 with A any subspace homeomorphic to S1 .

(b) X = S1 × D2 with A its boundary torus S1 × S1 .
(c) X = S1 × D2 with A the circle shown in the figure (refer to Hatcher p.39).
(d) X = D2 ∨ D2 with A its boundary S1 ∨ S1 .
(e) X a disk with two points on its boundary identified and A its boundary S1 ∨ S1 .
(f) X the Möbius band and A its boundary circle.

Solution. If there is a retraction r : X → A and i : A → X is inclusion, then ri is the identity

on A, and the induced homomorphism r∗ i∗ is the identity homomorphism on π1 (A), so i∗ is
injective.

(a) Since A ∼ = S1 , π1 (A) ∼

= Z. Also, π1 (R3 ) ∼
= 0, and there is no injection Z → 0, so A cannot
be a retraction of R . 3

(b) Since π1 (S1 ) ∼= Z and π1 (D2 ) ∼ = 0, we get π1 (S1 × D2 ) ∼= Z and π1 (S1 × S1 ) ∼= Z×Z
1 2
because both S and D are path-connected. For any homomorphism f : Z × Z → Z, we
have f ((1, 0)) = n and f ((0, 1)) = m for some integers m and n. But then f ((m, 0)) = nm
and f ((0, n)) = nm, so f cannot be injective. Thus, there is no retraction of S1 × D2 to
S1 × S1 .
(c) As above, π1 (S1 × D2 ) ∼ = Z, and since A is homeomorphic to S1 , π1 (A) ∼ = Z. Let x0 be
some point of A. The homomorphism i∗ : π1 (A, x0 ) → π1 (X, x0 ) induced by the inclusion
h i ∗
A → X is given by mapping h : [0, 1] → A to the composition [0, 1] − →A− → X. However,
if h is the generator of π1 (A, x0 ) that loops around A once, then i∗ (h) is nullhomotopic, so
i∗ is not injective. This gives that no retraction of X onto A can exist.
(d) Since D2 is contractible, each copy of D2 can be contracted to the identified point in
D2 ∨ D2 , and thus D2 ∨ D2 has trivial fundamental group. However, the fundamental
group of S1 ∨ S1 is F2 , the free group on 2 generators, by the van Kampen theorem. Since
there is no injection of F2 into the trivial group, there cannot be a retraction of D2 ∨ D2
onto its boundary.
(e) Up to homeomorphism, we may assume that the disk is the unit disk in R2 and that the
two points that are identified are (1, 0) and (−1, 0). There is a homotopy from X to the
circle [−1, 1] on the x-axis via the map ht ((x, y)) = (x, (1 − t)y), so π1 (X) ∼
= Z. However,
π1 (S1 ∨ S1 ) ∼
= F2 . If a and b are the generators of F2 , and f : F2 → Z is a homomorphism,
then f (a) = n and f (b) = m for some integers n and m. Then f (am ) = mn and f (bn ) = mn,
but am 6= bn , so there is no injection F2 → Z, and thus no retract of X onto its boundary.
(f) Let X be the Möbius band and A its boundary. The inclusion i : A → X induces a
homomorphism i∗ : π1 (A) → π1 (X). Both groups are Z, and i∗ (x) = 2x because looping
around the boundary of the Möbius band is the same as looping twice around the Möbius
band itself. This can be seen by letting A be the horizontal sides of a square whose vertical
sides are identified with opposite orientation. If a retraction r : X → A exists, then ri is the
identity on A, so by functoriality, r∗ i∗ is the identity homomorphism on π1 (A). If this were
the case, then i∗ (1) = 2, and r∗ (2) = 1, which implies r∗ (1) + r∗ (1) = 1, but r∗ (1) cannot
have an integer value. Thus there is no retraction of the Möbius band to its boundary.

1.2 Van Kampen’s Theorem

17. Exercise. Show that π1 (R2 \ Q2 ) is uncountable.
 1.3 COVERING SPACES 4

Solution. To see that R2 \ Q2 is path-connected, choose two points (a, b) and (c, d). Either a
or b must be irrational. Same with c and d. If a and d are both irrational, there is straight line
path from (a, b) to (a, d), and then another straight line path from (a, d) to (c, d). If instead c is
irrational, there is a straight line path from (c, d) to (c, d0 ) where d0 is some irrational number,
and this is the previous case. The other cases are similar, so we can compute√ π1 (R2 \ Q2 ) for any
base point
√ we like. For each irrational√number α,√let Bα be the union of √ {(x, 2) √ : −α ≤ x ≤ α},
{(x, − 2) : −α ≤ x ≤ α}, {(α, y) : − 2 ≤ y ≤ 2}, and {(−α, y) : − 2 ≤ y ≤ 2}. Neither of
2 2 2
√ contains a point of Q , so we think of it as a box in R \ Q . Let hα be a loop based
these sets
at (0, 2) that goes along Bα counterclockwise. If α < β are two irrationals, we claim that
hα and hβ are not homotopic to one another. The interior of the loop hα h−1 β can be thought
of as the space outside of hα and√ inside hβ . To be more precise, we mean the set of points
(x, y) ∈ R2 \ Q2 such that |y| < 2 and α < |x| < β. By the denseness of Q in R, there is a
rational number q in between α and β. Consider the inclusion R2 \ Q2 → R2 \ {(0, q)}. This
induces a homomorphism ϕ : π1 (R2 \ Q2 ) → π1 (R2 \ {(0, q)}). Then ϕ(hα h−1 β ) is the same path
in R \ {(0, q)}. This space is homotopic to S and under such a homotopy from R2 \ {(0, q)}
2 1

to S1 , ϕ(hα h−1 1
β ) becomes a nontrivial loop around S , so is not nullhomotopic. Thus hα hβ
−1

cannot be nullhomotopic because ϕ is a homomorphism, so hα and hβ are different elements in

π1 (R2 \ Q2 ). We have exhibited an injection of the irrationals into π1 (R2 \ Q2 ), and since the
set of irrational numbers is uncountable, we have the desired result.

1.3 Covering Spaces

1. Exercise. For a covering space p : X e = p−1 (A). Show that
e → X and a subspace A ⊂ X, let A
e → A is a covering space.
the the restriction p : A
Solution. For each point x ∈ A, there is a neighborhood U in X such that p−1 (X) is the
disjoint union of open sets Ui in Xe each of which gets mapped homeomorphically to U . Also,
e form a disjoint union of p−1 (U ∩ A). Each Ui ∩ A
U ∩ A is an open set, and the Ui ∩ A e is mapped
homeomorphically to U ∩ A, so A is a covering space of A.
e

e1 → X1 and p2 : X
2. Exercise. Show that if p1 : X e2 → X2 are covering spaces, so is their product
p1 × p2 : X1 × X2 → X1 × X2 .
e e
Solution. Choose (x1 , x2 ) ∈ X1 × X2 . Then there is a neighborhood Ui of xi in Xi such that
p−1
i (Ui ) is a disjoint union of open sets Vi,α in Xi which map homeomorphically to Ui . So U1 ×U2
e
−1
is an open set of X1 ×X2 such that (p1 ×p2 ) (U1 ×U2 ) is a disjoint union of products V1,α ×V2,β
each of which maps homeomorphically to U1 × U2 .

14. Exercise. Find all the connected covering spaces of RP2 ∨ RP2 .
Solution. Let X = RP2 ∨ RP2 , and let X1 and X2 denote the first and second copy of RP2 .
Since π1 (RP2 ) = Z/2 (this is done in example 1.43 of Hatcher), using the van Kampen theorem,
π1 (X) = Z/2 ∗ Z/2. Let a and b be the generators of Z/2 ∗ Z/2. To understand the connected
covering spaces of X, we classify the proper subgroups of Z/2 ∗ Z/2. We describe them in
terms of their generators. The first observation is that every element of Z/2 ∗ Z/2 is a word of
alternating a and b. The words that start and end with the same letter are precisely the set
of elements of order 2. The other words are of the form (ab)n and (ba)n , and these two are
inverse to one another, so without loss of generality, if (ba)n is in a generating set, it can be
replaced by (ab)n . For every n ≥ 0, there is a cyclic subgroup generated by (ab)n . In particular
the subgroup generated by ab is cyclic, and hence all of its subgroups are cyclic, so any set of
generators {(ab)n1 , (ab)n2 , . . . } can be replaced with a single generator (ab)n for some n. The
 1.A. GRAPHS AND FREE GROUPS 5

other subgroups have generating sets {(ab)n , g} where g is an element of order 2 and n ≥ 0.
Note that if g and g 0 are both of order 2, then gg 0 = (ab)m for some m, so a set {(ab)n , g, g 0 }
generates the same subgroup as {(ab)k , g} for some k, and the same applies for infinitely many
elements of order 2 (similar argument as for the case of generators {(ab)n1 , (ab)n2 , . . . }, so we
have described all of the subgroups of Z/2 ∗ Z/2.
The universal covering space X e of X is the infinite chain of S2 shown in Figure 4. We number
them with Z, and map the S2 with odd numbering to X1 and the others to X2 via the canonical
map S2 → RP2 . In each of the following cases, we will use this same map (we define the
numbering in Figure 4). The covering space associated to the subgroup generated by (ab)n for
n > 0 is a chain of S2 of length 2n. To get covering spaces associated to the subgroups of order
2, we can take one copy of RP2 and attach an infinite chain of S2 to one end. How we number
the S2 and where we map the RP2 gives rise to different subgroups of order 2. Also, we can
choose different base points. Of course, X itself corresponds to Z/2 ∗ Z/2. Finally, for the group
with generators {(ab)n , g} with n > 0 and g is an element of order 2, we can take a copy of
RP2 , attach a chain of S2 of length 2(n − 1), and attach to the end another copy of RP2 . The
covering map sends the first RP2 to X1 and the second one to X2 . If we number the S2 in
the chain, then the odd ones go to X2 and the even ones to X1 via the antipode identification.
Depending on which base point we choose, we can get the subgroups for various g.

26. Exercise. For a covering space p : Xe → X with X connected, locally path-connected, and
semilocally simply-connected, show:

(a) The components of X e are in one-to-one correspondence with the orbits of the action of
π1 (X, x0 ) on the fiber p−1 (x0 ).
(b) Under the Galois correspondence between connected covering spaces of X and subgroups
of π1 (X, x0 ), the subgroup corresponding to the component of X
e containing a given lift x
e0
of x0 is the stabilizer of x
e0 .

Solution.

(a) Choose z0 , z1 ∈ p−1 (x0 ). If z0 and z1 are in different components of X, e π1 (X, x0 ) cannot
map one to the other because there is no path connecting them. So we need to show that
π1 (X, x0 ) acts transitively on each of the components of X e to get the bijection. Since
X is assumed locally path-connected, X is locally path-connected. Thus, the notions of
e
connected components and path-connected components are the same. If z0 and z1 are in the
same component, let γ be a path joining them. Then pγ is an element of π1 (X, x0 ) whose
action on p−1 (x0 ) maps z1 to z0 (by Hatcher’s definition), and this gives the transitivity.
Then the set of elements in p−1 (x0 ) in a given component form an orbit, and this gives the
desired bijection.
(b) Choose a given lift xe0 of x0 in some component X 0 of X. e Under the Galois correspondence,
the subgroup of π1 (X, x0 ) associated to X 0 is the image of G = π1 (X 0 , x e0 ) in the inclusion
p∗ : G → π1 (X, x0 ). Any loop γ ∈ p∗ G then lifts back to a loop in X 0 by the unique lifting
property, so γ sends x e0 to itself, and is an element of the stabilizer of x e0 . Conversely, if
β ∈ π1 (X, x0 ) is in the stabilizer of x
e0 , then the lift β̄ of β is a loop from x e0 to itself, so
β̄ ∈ G, which means β ∈ p∗ G. This gives that p∗ G is the stabilizer of x e0 .

1.A. Graphs and Free Groups

3. Exercise. For a finite graph X define the Euler characteristic χ(X) to be the number of vertices
1.A. GRAPHS AND FREE GROUPS 6

minus the number of edges. Show that χ(X) = 1 if X is a tree, and that the rank (number of
elements in a basis) of π1 (X) is 1 − χ(X) if X is connected.
Solution. If X is a tree, then by Hatcher’s definition, it is contractible to a point, so must be
connected. Furthermore, for any two vertices v and w, there is a unique path from v to w. If
not, going along one and then backwards along the other gives a loop that is not nullhomotopic,
which contradicts the contractibility. Suppose X has n vertices. We claim that X has n − 1
edges. If n = 1, this is clear. For a graph on n vertices, remove any edge e. Then the remaining
space has two connected components. If not, the endpoints of e have another path connecting
them, which is a contradiction. The connected components have k and n − k vertices, and are
trees. This follows because X can be contracted to any vertex, so the connected components
are also contractible. By induction, the connected components have k − 1 and n − k − 1 edges,
so X has (k − 1) + (n − k − 1) + 1 = n − 1 edges, and χ(X) = 1.
Let T be a maximal tree in X. The existence of maximal trees is given by Proposition 1A.1 in
Hatcher. For each edge eα in X \ T , we can choose a small neighborhood Uα of T ∪ eα in X that
deformation retracts onto T ∪ eα . Then {Uα } is a covering of X, and the intersection of any
of them contains a small neighborhood of T , so is path-connected. Since T is contractible to a
point, π1 (T ) = 0, so the van Kampen theorem gives π1 (X) ∼ = ∗α π1 (Uα ). Each Uα deformation
retracts to T ∪ eα , which is homotopy equivalent to S1 because it contains exactly one cycle (this
follows from the uniqueness of paths in T ), so π1 (Uα ) ∼
= Z. Thus π1 (X) is a free group whose
rank is the number of edges of X minus the number of edges of T . If e is the number of edges
of X and v is the number of vertices, then the rank of π1 (X) is e − (v − 1) since T has v − 1
edges, and this is equal to 1 − χ(X).

6. Exercise. Let F be the free group on two generators and let F 0 be its commutator subgroup.
Find a set of free generators for F 0 by considering the covering space of the graph S1 ∨ S1
corresponding to F 0 .
Solution. Let a and b denote the generators of F . Construct a graph X e whose vertices are
the integer points Z in the plane, with an edge in between (x, y) and (x0 , y 0 ) if and only if
2

|x − x0 | + |y − y 0 | = 1. Let the base point be x e0 = (0, 0). Any edge either connects (x, y) to
(x + 1, y), or (x, y) to (x, y + 1) for some values of x and y. In both cases, orient the edge away
from (x, y). In the first case, label this edge with an a, and with the second, label it with a b.
We show part of this graph in Figure 5. Then this orientation is well-defined, and every vertex
has exactly one a edge coming in, one a edge coming out, and the same with b, so (X, e x
e0 ) is
1 1
a covering space of (X, x0 ) where X = S ∨ S and x0 . The covering map p maps each a to a
loop going around one of the copies of S1 and each b to a loop going around the other copy. For
m ∈ Z with m > 0, define am to be the edges of the form {(x, m), (x + 1, m)} where 0 ≤ x < m
and define bm to be the edges of the form {(m, y), (m, y + 1)} where 0 ≤ y < m. Also, define
a−m to be the edges making up the reflection of am across the y-axis, and b−m to be the edges
making up the reflection of bm across the x-axis. Then for any m, n ∈ Z \ {0}, define Xm,n to
be the union of an , bm , and the edges on the x-axis and y-axis. We illustrate this in Figure 5.
It is clear that the union of the Xm,n is X,e and their intersection is the union of the x-axis and
the y-axis, whose fundamental group is trivial. Also, each triple intersection is path-connected
since it is the union of the x-axis and the y-axis and possibly also some sets am and bn for
some numbers m and n. Note that π1 (Xm,n , x e0 ) is a free group generated by the loop that goes
from (0, 0) to (0, n) to (m, n) to (m, 0) and then back to (0, 0). Also, (Xm,n , x e0 ) is a covering
space for X, and the subgroup it maps to in π1 (X, x0 ) is the one generated by [an , bm ]. By
van Kampen’s theorem, the image of π1 (X, e x
e0 ) in π1 (X, x0 ) is a free group generated by the
2.1. SIMPLICIAL AND SINGULAR HOMOLOGY 7

elements {[an , bm ] : m, n ∈ Z \ {0}}. The commutator subgroup G of F contains this set, and is
generated by it, so we conclude that it is freely generated by this set.

2 Homology
2.1. Simplicial and Singular Homology
8. Exercise. Construct a 3-dimensional ∆-complex X from n tetrahedra T1 , . . . , Tn by the follow-
ing two steps. First arrange the tetrahedra in a cyclic pattern as in the figure (refer to Hatcher,
p.131), so that each Ti shares a common vertical face with its two neighbors Ti−1 and Ti+1 , the
subscripts being taken mod n. Then identify the bottom face of Ti with the top face of Ti+1
for each i. Show the simplicial homology groups of X in dimensions 0, 1, 2, 3 are Z, Zn , 0, Z,
respectively.
Solution. All of the outer vertices are identified with one another, and the two inner vertices are
also identified, so X has 2 0-cells. Label the outer vertex v0 and the inner vertex v1 . Also, the
n 3-cells each have 4 2-dimensional faces, but they are identified in pairs, so there are 2n 2-cells.
Each of the edges connecting v0 to itself are identified, and there is only one edge connecting
v1 to itself. So each tetrahedron has 4 remaining edges. Each one is identified with an edge of
its neighbor, and then further identified with another edge by identifying the bottom face of Ti
with the top face of Ti+1 , so in total there are only n 1-cells that connect v0 and v1 . Thus, we
compute the homology of the complex

∂3 ∂2 ∂1
0 / Zn / Z2n / Zn+2 / Z2 /0.

We order the faces of X based on the orientation of the edges in the figure in Hatcher. Each
1-cell either connects v0 to itself, v1 to itself, or connects v0 to v1 . In the first two cases, ∂1 maps
these 1-cells to 0, and in the last case ∂1 maps them to v0 − v1 , so ∂1 (Zn+2 ) ∼ = Z, which means
H0 (X) = Z2 /Z ∼ = Z.
Label the bottom face of Ti as fi and label its face on the right side (using counterclockwise
orientation in the figure in Hatcher) fn+i . Also, label the outer edge e and the edge connecting
v1 to itself en+1 . Label the bottom edge of fn+i with ei . For 1 ≤ i ≤ n, we have ∂2 (fi ) =
−ei + ei−1 − e and ∂2 (fn+i ) = −en+1 + ei−1 − ei , where e0 means en . If we order the edges
e, e1 , . . . , en , en+1 , then the image of ∂2 is the subgroup generated by the row vectors of the
following 2n × (n + 2) matrix

−1 −1 0 0 ···
 
0 1 0
−1 1 −1 0 · · · 0 0  0
 
−1 0 1 −1 · · · 0 0  0

..

 

 . 

−1 0 0 0 ··· 1 −1 0 
 . (1)
 0 −1 0 0 ··· 0 1 −1
 
0
 1 −1 0 · · · 0 0 −1
0
 0 1 −1 · · · 0 0 −1
 .. 
 . 
0 0 0 0 ··· 1 −1 −1
 2.1. SIMPLICIAL AND SINGULAR HOMOLOGY 8

For 1 ≤ i ≤ n, we subtract row i from row n + i and see that the resulting last n rows are the
same, so we can reduce to the following (n + 1) × (n + 2) matrix
 
−1 −1 0 0 ··· 0 1 0
−1 1 −1 0 · · · 0 0 0
 
−1 0 1 −1 · · · 0 0 0
.
 
 ..

 . 

−1 0 0 0 · · · 1 −1 0 
1 0 0 0 · · · 0 0 −1

Now for 1 ≤ i ≤ n, add the first i − 1 rows to the ith row to get
 
−1 −1 0 0 ··· 0 1 0
 −2 0 −1 0 · · · 0 1 0
 
 −3 0 0 −1 · · · 0 1 0
.
 
 ..

 . 

−n 0 0 0 ··· 0 0 0
1 0 0 0 ··· 0 0 −1

Note that e and en+1 are the only edges mapped to 0 by ∂1 and every other edge is mapped to
v0 − v1 , so ker ∂1 is generated by e, en+1 , and all differences ei − ej where 1 ≤ i < j ≤ n. These
differences can be generated by just en −ei for 1 ≤ i ≤ n−1, so we have ker ∂1 = he, en+1 , en −ei i =
he, e − en+1 , −ie + en − ei i where the second equality is a result of adding e some number of
times to each generator. From this, it is easy to see that H1 (X) = ker ∂1 / image ∂2 ∼ = Z/n.
For any a1 f1 + · · · + a2n f2n ∈ ker ∂2 , reading off from (1), we see that a1 + · · · + an = an+1 + · · · +
a2n = 0 and that for 1 < i ≤ n, we have ai + an+i = ai−1 + an+i−1 and a1 + an+1 = an + a2n .
This implies that a1 + an+1 = a2 + an+2 = · · · = an + a2n ; the sum of all of these terms is 0, so
ai = −an+i for 1 ≤ i ≤ n. We claim that a1 f1 + · · · + a2n f2n ∈ image ∂3 . Since

∂3 (b1 T1 + · · · + bn Tn ) = b1 (−fn+1 + f2n − fn + f1 ) + · · · + bn (−f2n + f2n−1 − fn−1 + fn )

= (b1 − b2 )f1 + (b2 − b3 )f2 + · · · + (bn − b1 )fn
+ (b2 − b1 )fn+1 + (b3 − b2 )fn+2 + · · · + (b1 − bn )f2n ,

it is enough to find b1 , . . . , bn such that bi − bi+1 = ai where bn+1 means b1 . To do this, we can
pick any b1 , and the other bi are determined. The only thing to check is that bn − b1 = an , but
this follows because

bn − b1 = −((b1 − b2 ) + (b2 − b3 ) + · · · + (bn−1 − bn )) = −(a1 + · · · + an−1 ) = an .

This gives image ∂3 = ker ∂2 , so H2 (X) = 0.

For each 1 < i ≤ n, ∂3 (Ti ) = −fn+i +fn+i−1 −fi−1 +fi , and ∂3 (T1 ) = −fn+1 +f2n −fn +f1 . Any
2-cell fj appears in two neighboring 3-cells, say Ti and Ti+1 . The coefficient of fj in ∂3 (Ti ) and
∂3 (Ti+1 ) appear with opposite sign. Thus, if a1 T1 + · · · + an Tn ∈ ker ∂3 , then a1 = a2 = · · · = an .
So ker ∂3 ∼= Z, which gives H3 (X) ∼= Z/0 = Z.

12. Exercise. Show that chain homotopy of chain maps is an equivalence relation.
Solution. Let (C, ∂) and (C 0 , ∂ 0 ) be two chain complexes. If f, g : C → C 0 are two chain
0
maps, write f ∼ g if f is chain homotopic to g, i.e., there are maps s : Cn → Cn+1 such that
 2.1. SIMPLICIAL AND SINGULAR HOMOLOGY 9

f − g = s∂ + ∂ 0 s. By choosing s = 0, we see that f ∼ f . If f ∼ g via the map s, then g ∼ f via

the map −s:
g − f = −(f − g) = −(s∂ + ∂ 0 s) = (−s)∂ + ∂ 0 (−s).
Finally, if f ∼ g via s and g ∼ h via t, then

f − h = (f − g) + (g − h) = (s∂ + ∂ 0 s) + (t∂ + ∂ 0 t) = (s + t)∂ + ∂ 0 (s + t),

so f ∼ h via s + t. Thus, chain homotopy is an equivalence relation.

18. Exercise. Show that for the subspace Q ⊂ R, the relative homology group H1 (R, Q) is free
Abelian and find a basis.
Solution. The long exact sequence on homology gives

f
··· / H1 (R) / H1 (R, Q) / H0 (Q) / H0 (R) / ··· .

Since R is contractible, H1 (R) = 0, so H1 (R, Q) = ker f . For aL space X, H0 (X) is a direct

sum of Z with one copy for each connected component. So f : x∈Q Z → Z is induced by
the inclusion Q ,→ R and the images of each generator of H0 (Q) are all homologous to a
generator of H0 (R). This implies that ker f is the set of elements a1 ep1 + · · · + ar epr such that
a1 + · · · + ar = 0, and this set is generated by elements of the form ep − eq where ep denotes
the identity element in the copy of Z corresponding to p ∈ Q (reason: a1 ep1 + · · · + ar epr =
a1 (ep1 − ep2 ) + (a1 + a2 )(ep2 − ep3 ) + · · · + (a1 + a2 + · · · + ar−1 )(epr−1 − epr ) if a1 + · · · + ar = 0).
We claim that a basis for the kernel is B := {e0 − ep : p ∈ Q \ {0}}. It is clear that B is linearly
independent: if there is a relation a1 (e0 − ep1 ) + · · · + ar (e0 − epr ), then a1 e0 + · · · + ar e0 = 0
since there are no relations among e0 and the epi , so ai = 0 for all i. To see B generates ker f ,
pick any ep − eq . Then (e0 − eq ) − (e0 − ep ) = ep − eq , so any element generated by the ep − eq
can be generated by B.

27. Exercise. Let f : (X, A) → (Y, B) be a map such that both f : X → Y and the restriction
f : A → B are homotopy equivalences.

(a) Show that f∗ : Hn (X, A) → Hn (Y, B) is an isomorphism for all n.

(b) For the case of the inclusion f : (Dn , Sn−1 ) → (Dn , Dn \{0}), show that f is not a homotopy
equivalence of pairs – there is no g : (Dn , Dn \ {0}) → (Dn , Sn−1 ) such that f g and gf are
homotopic to the identity through maps of pairs.

Solution.

(a) By naturality of the long exact sequence of homology (p.127 of Hatcher), for all n, the
diagram

Hn (A) / Hn (X) / Hn (X, A) / Hn−1 (A) / Hn−1 (X)

f∗ f∗ f∗ f∗ f∗
    
Hn (B) / Hn (Y ) / Hn (Y, B) / Hn−1 (B) / Hn−1 (Y )

commutes. Since f : X → Y and f : A → B are homotopy equivalences, the first two and
last two vertical arrows in the above diagram are isomorphisms. Also, the top and bottom
rows are exact, so by the five-lemma, the middle vertical arrow is also an isomorphism.
2.2. COMPUTATIONS AND APPLICATIONS 10

(b) Let g : (Dn , Dn \ {0}) → (Dn , Sn−1 ) be a map of pairs. Since Sn−1 is a closed set, g −1 (Sn−1 )
is closed. By assumption, Dn \ {0} ⊆ g −1 (Sn−1 ). Also, 0 is in the closure of Dn \ {0}, so
g(0) ∈ Sn−1 , and thus Dn \ {0} ,→ Dn → Sn−1 is a factorization of g. We have an induced
map on homology Hn (Dn \ {0}) → Hn (Dn ) → Hn (Sn−1 ) whose composition is g∗ . Since
Hn (Dn ) = 0 (here we assume n > 0, but if n = 0, then D0 \ {0} = ∅, so f won’t exist),
we conclude that g∗ = 0. Thus g does not induce an isomorphism on homology because
Hn (Sn−1 ) = Z, so g cannot be a homotopy equivalence of pairs. Since g was arbitrary, we
see that f is also not a homotopy equivalence of pairs.

2.2. Computations and Applications

2. Exercise. Given a map f : S2n → S2n , show that there is some point x ∈ S2n with either
f (x) = x or f (x) = −x. Deduce that every map RP2n → RP2n has a fixed point. Construct
maps RP2n−1 → RP2n−1 without fixed points from linear transformations R2n → R2n without
eigenvectors.
Solution. Suppose there is a map ϕ : S2n → S2n such that ϕ has no fixed points and ϕ(x) 6= −x
for all x ∈ S2n . Since ϕ has no fixed points, the line segment (1−t)ϕ(x)−tx never passes through
the origin, so we can define a homotopy from ϕ to the antipodal map with F : I × S2n → S2n by
(1 − t)ϕ(x) − tx
(t, x) 7→ ,
|(1 − t)ϕ(x) − tx|
so deg ϕ = (−1)2n+1 = −1. Similarly, the line segment (1 − t)ϕ(x) + tx never passes through
the origin since ϕ(x) 6= −x for all x ∈ S2n , so we can define a homotopy from ϕ to the identity
map with F : I × S2n → S2n by
(1 − t)ϕ(x) + tx
(t, x) 7→ .
|(1 − t)ϕ(x) + tx|
Then deg ϕ = 1, which is a contradiction. Thus, there must exist x ∈ S2n such that either
ϕ(x) = x or ϕ(x) = −x.
Now let f : RP2n → RP2n be any map. Composing it with the canonical map π : S2n → RP2n ,
we get f 0 : S2n → RP2n . Since S2n is a covering space of RP2n , f 0 lifts (via the lifting criterion
since S2n has trivial fundamental group) to a map g : S2n → S2n . In other words, the diagram

jjj5 S
2n
g jjjjj
jjjjjjj π
jjj
j π / f 
S2n RP2n / RP2n

commutes. From above, there is a point x ∈ S2n such that either g(x) = x or g(x) = −x. Then
f (π(x)) = π(g(x)) = π(±x) = π(x),
so π(x) is a fixed point of f .
Let T : R2n → R2n be the linear transformation given by (x1 , x2 , . . . , x2n ) 7→ (−x2n , x1 , x2 , . . . , x2n−1 ).
Then T 2n = −I, where I is the identity map on R2n , so x2n + 1 divides the characteristic poly-
nomial of T , and hence is the characteristic polynomial since it has degree 2n. However, it
has no real roots, so T has no real eigenvalues, and hence no eigenvectors. Notice that T acts
on S2n−1 ⊆ R2n , and this action is a continuous map. Since T has no eigenvectors, we have
T (x) 6= x and T (x) 6= −x for all x ∈ S2n . Also, since T (−x) = −T (x), T gives a well-defined
map RP2n−1 → RP2n−1 which has no fixed points.
 2.2. COMPUTATIONS AND APPLICATIONS 11

8. Exercise. A polynomial f (z) with complex coefficients, viewed as a map C → C, can always
be extended to a continuous map of one-point compactifications fˆ: S2 → S2 . Show that the
degree of fˆ equals the degree of f as a polynomial. Show also that the local degree at fˆ at a
root of f is the multiplicity of the root.
Solution. Let z1 , . . . , zr be the distinct roots of f (z) with multiplicities m1 , . . . , mr . We can find
disjoint neighborhoods U1 , . . . , Ur in S2 for each zi . Each Ui is mapped into some neighborhood
Vi of 0. Consider the induced map on relative homology fˆ∗ : H2 (Ui , Ui \ {zi }) → H2 (Vi , Vi \ {0}).
Both groups are isomorphic to Z, so this map is given by multiplication by some number, which
is the local degree of fˆ at zi (see Hatcher p.136). The restriction of fˆ to any small neighborhood
of zi will be a mi -to-1 mapping onto some open neighborhood of 0 contained in its image. This
implies that the local degree is mi since a generator for H2 (Ui , Ui \ {zi }) is mapped to mi times
a generator of H2 (Vi , Vi \ {0}).
Since the local degree of fˆ at zi is mi , we see that deg fˆ = deg fˆ|zi =
P P
i mi = deg f ,
i
where the first equality is by Proposition 2.30 of Hatcher and the last equality follows from the
fundamental theorem of algebra.
17. Exercise. Show the isomorphism between cellular and singular homology is natural in the
following sense: A map f : X → Y that is cellular – satisfying f (X n ) ⊂ Y n for all n – induces
a chain map f∗ between the cellular chain complexes of X and Y , and the map f∗ : HCW n (X) →
CW
Hn (Y ) induced by this chain map corresponds to f∗ : Hn (X) → Hn (Y ) under the isomorphism
HCW ∼
= Hn .
n
Solution. Since f : X → Y is a cellular map, for all n ≥ 0, the restriction of f to the n-skeleton
of X gives a map of pairs (X n , X n−1 ) → (Y n , Y n−1 ), which induces a map on relative homology
f∗ : Hn (X n , X n−1 ) → Hn (Y n , Y n−1 ). But cellular chain groups are defined as these relative
homology groups, so f induces a chain map between the cellular chain complexs of X and Y .
Also, f∗ induces a map on homology f∗CW : HCW CW
n (X) → Hn (Y ). By abuse of notation,
f : X → Y induces a map on homology f∗ : Hn (X) → Hn (Y ). Let iX : Hn (X) → HCW n (X)
CW
and iY : Hn (Y ) → Hn (Y ) be the isomorphism between cellular and singular homology given
by Theorem 2.35 of Hatcher. We wish to show that the diagram
f∗
Hn (X) / Hn (Y ) (2)
iX iY
 f∗CW

HCW / HCW (Y )
n (X) n

commutes. In fact, the diagram without f∗ comes from the commutative diagram
Hn (X)
l5
lll
lll
Hn (X n )
k5 RRRjX
kkk RRR
kkk R(
Hn+1 (X n+1 , X n ) / Hn (X n , X n−1 ) / Hn−1 (X n−1 , X n−2 )

 f∗  f∗  f∗
Hn+1 (Y n+1 , Y n ) / Hn (Y n , Y n−1 ) / Hn−1 (Y n−1 , Y n−2 )
SSS jY lll6
SSS l
S) lll
Hn (Y n )
RRR
RRR
R)
Hn (Y )
 3.2. CUP PRODUCT 12

which is an augmentation of the one found on p.139 of Hatcher. Here the isomorphisms iX and
iY are induced by jX and jY . If f∗ : Hn (X) → Hn (Y ) fills in the above diagram, then we are
done. But now this is just a question of filling in ? with f∗ in the following diagram

Hn+1 (X n+1 , X n ) / Hn (X n ) / Hn (X)


f∗

  ?
Hn+1 (Y n+1 , Y n ) / Hn (Y n ) / Hn (Y )

which is the long exact sequence on homology of good pairs. By the naturality of the long exact
sequence on homology, we conclude that f∗ does fill in ?, so (2) commutes. We conclude that
the isomorphism between cellular homology and singular homology is natural.
20. Exercise. For finite CW complexes X and Y , show that χ(X × Y ) = χ(X)χ(Y ).
Solution. Given two finite CW complexes X and Y with some given CW structure, let an and
bn be the number of n-cells in X and Y , respectively. By the isomorphism of cellular homology
and singular homology, we have χ(X) = n (−1) an and χ(Y ) = n (−1)n bn . The product
n
P P
X × Y has a CW structure whose cells are given by em n m
α × eβ where eα ranges
P over
P the cellsnof X
n
and eβ ranges over the cells of Y (see Hatcher p.8). This gives χ(X ×Y ) = n i+j=n (−1) ai bj ,
and this is also the product χ(X)χ(Y ).
21. Exercise. If a finite CW complex X is the union of subcomplexes A and B, show that χ(X) =
χ(A) + χ(B) − χ(A ∩ B).
Solution. Now suppose X is a finite CW complex that is the union of two subcomplexes A
and B. Let an , bn , and cn denote the number of n-cells in A, B, and A ∩ B, respectively. By
inclusion-exclusion, the number of n-cells in X is then an + bn − cn . So we have the following
equalities:
X
χ(X) = (−1)n (an + bn − cn )
n
X X X
= (−1)n an + (−1)n bn − (−1)n cn
n n n
= χ(A) + χ(B) − χ(A ∩ B).

22. Exercise. For X a finite CW complex and p : X e → X an n-sheeted covering space, show that
χ(X) = nχ(X).
e
Solution. Now suppose that X is a finite CW complex and p : X e → X is an n-sheeted covering
space. Then X has a CW complex structure where the i-cells are the lifts of i-cells of X. More
e
specifically, every i-cell σ is equipped with a characteristic map fσ : Di → X which lifts to a
unique map feσ : Di → X once the image of any point is specified. Since p is n-sheeted, we can
get n different lifts, so the number of i-cells of Xe is n times the number of i-cells of X. This
gives the formula χ(X) e = nχ(X), which follows directly from the alternating sum of number of
cells.

3 Cohomology
3.2. Cup Product
1. Exercise. Assuming as known the cup product structure on the torus S1 × S1 , compute the
3.2. CUP PRODUCT 13

cup product structure in H∗ (Mg ) for Mg the closed orientable surface of genus g by using the
quotient map from Mg to a wedge sum of g tori.
Solution. From the universal coefficient theorem, we have the following exact sequence

0 / Ext(H0 (Mg ; Z), Z) / H1 (Mg ; Z) / Hom(H1 (Mg ; Z), Z) /0.

We know that H1 (Mg ; Z) ∼ = Z2g , so Hom(H1 (Mg ; Z), Z) ∼ ∼ Z, so

= Z2g , and H0 (Mg ; Z) =
1 ∼ 2g
Ext(H0 (Mg ; Z), Z) = 0, which all implies that H (Mg ; Z) = Z . Similarly, we have the short
exact sequence

0 / Ext(H1 (Mg ; Z), Z) / H2 (Mg ; Z) / Hom(H2 (Mg ; Z), Z) /0

which implies that H2 (Mg ; Z) ∼

= Z. By Example 3.15 of Hatcher, H∗ (T 2 ; Z) ∼= Z2 is the
V
exterior algebra on two generators where T 2 = S1 × S1 . By Example 3.13 of Hatcher, there is
an isomorphism of reduced cohomology rings
g g
T 2 ; Z) ∼
_ Y
e ∗(
H = e ∗ (T 2 ; Z).
H
i=1 i=1
Wg
Now let f : Mg → i=1 T 2 be the quotient map illustrated in Exercise 3.2.1 of Hatcher. This
induces a graded homomorphism of cohomology rings
g
_
∗ ∗
f :H ( T 2 ; Z) → H∗ (Mg ; Z).
i=1

Denote the 2g generators of H1 (Mg ; Z) as α1 , . . . , αQg , β1 , . . . , βg where αi , βi correspond to the

g 1 2
generators of the ith coordinate in the product W i=1 H (T ; Z), and the one generator of
2 1 g 2
H (Mg ; Z) as γ. Also denote the generators of H ( i=1 T ; Z)Was a1 , . . . , ag , b1 , . . . , bg where
ai , bi correspond to the ith component, and the generators of H2 ( gi=1 T 2 ; Z) as c1 , . . . , cg . With
an appropriate choice of labels, we see that f ∗ (ai ) = αi , f ∗ (bi ) = βi , and f ∗ (ci ) = γ for all i.
From this direct product description, it is immediately clear that for i 6= j,

αi ^ αj = βi ^ βj = αi ^ βj = βi ^ αj = 0.

From the fact that each H∗ (T 2 ; Z) is the exterior algebra of Z2 , we also verify that αi ^ αi =
βi ^ βi = 0. Finally, since ai ^ bi = −bi ^ ai = ci , we see that f ∗ (ai ) ^ f ∗ (bi ) = −f ∗ (bi ) ^
f ∗ (ai ) = f ∗ (ci ), which implies that αi ^ βi = −βi ^ αi = γ.

4. Exercise. Apply the Leftschetz fixed point theorem to show that every map f : CPn → CPn
has a fixed point if n is even, using the fact that f ∗ : H∗ (CPn ; Z) → H∗ (CPn ; Z) is a ring
homomorphism. When n is odd show there is a fixed point unless f ∗ (α) = −α, for α a generator
of H2 (CPn ; Z).
Solution. The cohomology ring of CPn is H∗ (CPn ; Z) = Z[x]/(xn+1 ) where x has degree 2
by Theorem 3.12 of Hatcher. So each cohomology group in even degree ≤ n has rank 1, and
each cohomology group in odd degree is 0. By the naturality of the universal coefficient theorem
(Hatcher p.196) and the discussion of trace in Hatcher p.181, the trace of f ∗ : Hi (CPn ; Z) →
Hi (CPn ; Z) is the same as the trace of f∗ : Hi (CPn ; Z) → Hi (CPn ; Z) for any map f : CPn →
CPn .
 3.3. POINCARÉ DUALITY 14

Given such a map, the induced map f ∗ : H0 (CPn ; Z) → H0 (CPn ; Z) has trace 1. The induced
map on the second cohomology groups is x 7→ ax for some a ∈ Z. By naturality of cup product,
this means that the map on the 2ith cohomology groups is xi 7→ ai xi . Collecting these remarks,
the Lefschetz number is
n
(
1−an+1
1−a , a 6= 1 .
X
2i i
τ (f ) = (−1) a =
i=0
n + 1, a=1

This last number is nonzero unless a = −1 and n is odd. In particular, we have shown that
if n is even, then f must have a fixed point by the Lefschetz fixed point theorem, and for n
odd, we have shown the same except in the case that f ∗ (x) = −x where x is the generator of
H2 (CPn ; Z).

11. Exercise. Using cup products, show that every map Sk+` → Sk × S` induces a trivial homo-
morphism Hk+` (Sk+` ) → Hk+` (Sk × S` ), assuming k > 0 and ` > 0.
Solution. By the Künneth formula, H∗ (Sk × S` ; Z) = ∼ H∗ (Sk ; Z) ⊗ H∗ (S` ; Z). In particular,
the kth and `th cohomology groups of Sk × S` are Z because Hk (Sk ; Z) = H` (S` ; Z) = Z by
the universal coefficient theorem. On the other hand, the kth and `th cohomology groups of
Sk+` are trivial. Thus, any map f : Sk+` → Sk × S` induces the zero map on the kth and `th
cohomology groups. Any element of Hk+` (Sk × S` ; Z) can be written as a product of elements
in Hk (Sk × S` ; Z) and H` (Sk × S` ; Z), so the induced map f ∗ : Hk+` (Sk × S` ; Z) → Hk+` (Sk+` ; Z)
is also zero by naturality of cup product. Finally, by the naturality of the universal coefficient
theorem (Hatcher p.196), the following diagram
ϕ
/ Ext(H k+` ; Z), Z) / Hk+` (Sk+` ; Z) / Hom(H k+` ; Z), Z) /0
0 k+`−1 (S
O O k+` (S
O
(f∗ )∗ f∗ (f∗ )∗
ϕ
/ Ext(H k ` / Hk+` (Sk × S` ; Z) / Hom(H k × S` ; Z), Z) /0
0 k+`−1 (S × S ; Z), Z) k+` (S

commutes and the horizontal rows are exact. Since f ∗ is the zero map the surjectivity of
the maps ϕ imply that the (f∗ )∗ on the right is also zero. This map is the dual of the map
f∗ : Hk+` (Sk+` ; Z) → Hk+` (Sk × S` ; Z), and hence f∗ is also the zero map.

3.3. Poincaré Duality

8. Exercise. For a map f : M → N between connected closed orientable n-manifolds, suppose
there is a ball B ⊂ N such that f −1 (B) is the disjoint
P union of balls Bi each mapped home-
omorphically by f onto B. Show the degree of f is i εi where εi is +1 or −1 according to
whether f : Bi → B preserves or reverses local orientations induced from given fundamental
classes [M ] and [N ].
Solution. Let x be a point in the interior of B, and let xi be the point in Bi that maps to x.
Also, let r be the number of balls Bi . Similar to the discussion of the degree of a map from a
 3.3. POINCARÉ DUALITY 15

sphere to itself on p.136 of Hatcher, we have that the following diagram

f∗
Hn (Bi , Bi \ {xi }) / Hn (B, B \ {x})
=iiiii
∼ iii
iiii ki ∼
=
tiiii p  f∗ 
Hn (M, M \ {xi }) o / Hn (N, N \ {x})
i
Hn (M, M \ {x1 , . . . , xr })
jUUUU O O
UUUU
UUUU ∼
∼ UUUU j =
= UU f∗
Hn (M ) / Hn (N )

commutes, where the ki and pi are induced by inclusions. Taking the generator [M ] ∈ Hn (M ),
we know that pi j([M ]) = µxi , the local orientation at xi , by the commutativity of the lower
triangle. By excision, the middle term Hn (M, M \ {x1 , . . . , xr }) is isomorphic to the direct sum
of the groups Hn (Bi , Bi \ {xi }) ∼
= Z, and ki is the inclusion mapPinto the ith summand. Since
the pi is projection onto the ith summand, we see that j([M ]) = ri=1 ki (µxi ) by commutativity
of the upper triangle.
By the commutativity of the upper square, we deduce that f∗ (ki (µxi )) = εi where εi = ±1
depending on whether f preserves or reverses the orientation of Bi whenP
mapping to B. Finally,
r
by
Prcommutativity of the lowerPr square, we conclude that f∗ ([M ]) = ( i=1 f∗ (ki (µxi )))[N ] =
( i=1 εi )[N ]. Thus, deg f = i=1 εi .
17. Exercise. Show that a direct limit of exact sequences is exact. More generally, show that
homology commutes with direct limits: If {Cα , fαβ } is a directed system of chain complexes,
with the maps fαβ : Cα → Cβ chain maps, then Hn (lim Cα ) = lim Hn (Cα ).
−→ −→
Solution. There is a canonical map ϕi : Ci → lim Cα for all i which induces a map on homology
−→
Hn (Ci ) → Hn (lim Cα ). By the functoriality of homology, these induced maps are compatible
−→
with the maps Hn (Ci ) → Hn (Cj ), i.e., the following diagram

Hn (Ci ) / Hn (Cj )
q q
ϕi∗ qqqqq
 xqqq ϕj∗
Hn (lim Cα )
−→
commutes for all i and j for which there is a map fij . By the universal property of direct limit,
this induces a map ϕ : lim Hn (Cα ) → Hn (lim Cα ) such that the following diagram
−→ −→
Hn (Ci ) / Hn (Cj )
;;MMM q
;; MMMλi λj qqq 
;; MMM qq  
q 
;; M& xqqq 
;; 
; lim H (C ) 
ϕi∗ ;;− → n α  ϕj∗
;; 
;; ϕ 
; 

Hn (lim Cα )
−→
commutes for all i and j for which there is a map fij and where λi denotes the canonical map
into a direct limit. We claim that ϕ is an isomorphism.
For surjectivity, choose x ∈ Hn (lim Cα ). Then x is a cycle, and hence ∂x = 0 where ∂ is the
−→
differential in lim Cα . Pick a representative xi ∈ Ci of x, i.e., ϕi (xi ) = x. Then ϕi∗ (∂i xi ) = 0, and
−→
 3.C. H-SPACES AND HOPF ALGEBRAS 16

hence there exists some j such that fij (∂i xi ) = 0. This means that fij (xi ) ∈ Hn (Cj ). Setting
y = λj (fij (xi )), we have ϕ(y) = x.
To see injectivity, suppose x ∈ lim Hn (Cα ) is mapped to 0 by ϕ. Choose a representative
−→
xi ∈ Hn (Ci ) of x, i.e., λi (xi ) = x. Then ϕi∗ (xi ) is a boundary of some element, i.e., there exists
y such that ∂y = ϕi∗ (xi ) where ∂ is the differential of lim Cα . Then we can find a representative
−→
yj ∈ Cj of y for some j with ∂j yj = 0. But ∂j yj is also a representative of x, so x = 0.
Therefore, ϕ is an isomorphism, so direct limits commute with homology. The statement about
direct limits preserving exact sequences follows because exactness is equivalent to homology
being trivial.

32. Exercise. Show that a compact manifold does not retract onto its boundary.
Solution. Let M be a compact manifold and suppose that there is a retraction r : M → ∂M .
Let i : ∂M ,→ M be the inclusion, so that r ◦ i is the identity on M . This implies that the
induced map on homology i∗ : Hn−1 (∂M ; Z/2) → Hn−1 (M ; Z/2) is injective by functoriality of
homology. By the long exact sequence of relative homology,

Hn (M, ∂M ; Z/2)
∂ / Hn−1 (∂M ; Z/2) i∗ / Hn−1 (M ; Z/2)

is exact. This implies that ∂ = 0 because i∗ is injective. But this contradicts exercise 3.3.31
of Hatcher, which says that ∂ sends a fundamental class of (M, ∂M ) to a fundamental class of
∂M . Thus, M does not retract onto its boundary.

3.C. H-Spaces and Hopf Algebras

5. Exercise. Show that if (X, e) is an H-space then π1 (X, e) is Abelian.
Solution. Choose f, f 0 , g, g 0 ∈ π1 (X, e) and suppose that H is a homotopy f ' f 0 and H 0 is
a homotopy g ' g 0 . We claim that H ∗ H 0 : I × I → X defined by (s, t) 7→ H(s, t) ∗ H 0 (s, t)
is a homotopy f ∗ g ' f 0 ∗ g 0 where f ∗ g : I → X is defined by s 7→ f (s) ∗ g(s). Indeed,
(H∗H 0 )(0, t) = H(0, t)∗H 0 (0, t) = e∗e = e, and similarly, (H∗H 0 )(1, t) = e. Also, (H∗H 0 )(s, 0) =
H(s, 0) ∗ H 0 (s, 0) = f (s) ∗ g(s) and similarly (H ∗ H 0 )(s, 1) = f 0 (s) ∗ g 0 (s). Since H ∗ H 0 is the
composition of continuous maps, it is continuous, and thus the desired homotopy. Now pick any
h, h0 ∈ π1 (X, e). Let 1 denote the constant path with base point e. Then h ∗ h0 ' (h · 1) ∗ (1 · h0 ).
Since (h·1)(s) is h(2s) if 0 ≤ s ≤ 1/2 and is e otherwise, and (1·h0 )(s) is h0 (2s−1) if 1/2 ≤ s ≤ 1
and e otherwise, we get (h · 1) ∗ (1 · h0 ) ' h · h0 . By the same reasoning, h · h0 ' (1 · h0 ) ∗ (h · 1),
and this is homotopic to h0 · h. This gives that h · h0 ' h0 · h, so π1 (X, e) is Abelian.

4 Homotopy Theory
4.1. Homotopy Groups
2. Exercise. Show that if ϕ : X → Y is a homotopy equivalence, then the induced homomorphisms
ϕ∗ : πn (X, x0 ) → πn (Y, ϕ(x0 )) are isomorphisms for all n.
Solution. First, the technique of p.341 shows an analogue of Lemma 1.19 for higher homotopy
groups. Let ψ : Y → X be a homotopy inverse of ϕ. Then ψϕ ' 1X implies that ψ∗ ϕ∗ is an
isomorphism because it is equal to a change of base point isomorphism as described in p.341.
Similarly, ϕ∗ ψ∗ is an isomorphism, so we conclude that ϕ∗ is an isomorphism for all n.
 4.1. HOMOTOPY GROUPS 17

3. Exercise. For an H-space (X, x0 ) with multiplication µ : X × X → X, show that the group
operation in πn (X, x0 ) can also be defined by the rule (f + g)(x) = µ(f (x), g(x)).
Solution. Writing x = (x1 , . . . , xn ) ∈ I n , the sum on the left-hand side is defined to be
(
f (2x1 , x2 , . . . , xn ), x1 ∈ [0, 1/2]
(f + g)(x1 , . . . , xn ) = .
g(2x1 − 1, x2 , . . . , xn ), x1 ∈ [1/2, 1]

Letting 1 denote the constant map I n → x0 , we have f + g is homotopic to

(
µ(f (2x1 , x2 , . . . , xn ), 1), x1 ∈ [0, 1/2]
,
µ(1, g(2x1 − 1, x2 , . . . , xn )), x1 ∈ [1/2, 1]

which in turn is homotopic to

µ(f (x1 , x2 , . . . , xn ), g(x1 , x2 , . . . , xn )).

5. Exercise. For a pair (X, A) of path-connected spaces, show that π1 (X, A, x0 ) can be identified
in a natural way with the set of cosets αH of the subgroup H ⊂ π1 (X, x0 ) represented by loops
in A at x0 .
Solution. By definition, π1 (X, A, x0 ) is the set of homotopy classes of paths in X from a varying
point in A to x0 ∈ A. Define a map π1 (X, x0 ) → π1 (X, A, x0 ) by thinking of a loop at x0 as an
element of π1 (X, A, x0 ). Since A is path-connected, every element of π1 (X, A, x0 ) is homotopic
to a loop at x0 , so this map is surjective. Note that two loops γ0 , γ1 ∈ π1 (X, x0 ) are homotopic
relative to A if and only if γ0−1 γ1 is represented by a loop in A, so we can identify π1 (X, A, x0 )
with the set of cosets αH.
10. Exercise. Show that the ‘quasi-circle’ described in (Ex. 1.3.7) has trivial homotopy groups but
is not contractible, hence does not have the homotopy type of a CW complex.
Solution. Let Y be the quasi-circle. Since Y has infinite length, the image of any map I n → Y
must live in some region homeomorphic to the unit interval, or the disjoint union of two copies
of the unit interval with their midpoints identified. Both such spaces are contractible, so Y has
trivial homotopy groups. However, this space is not contractible, because identifying the part
of the graph of y = sin(1/x) to a single point gives the circle.
11. Exercise. Show that a CW complex is contractible if it is the union of an increasing sequence of
subcomplexes X1 ⊂ X2 ⊂ · · · such that each inclusion Xi ,→ Xi+1 is nullhomotopic, a condition
sometimes expressed by saying Xi is contractible in Xi+1 . An example is S∞ , or more generally
the infinite suspension S ∞ X of any CW complex X, the union of the iterated suspensions S n X.
Solution. By Whitehead’s theorem, it is enough to show that all of the homotopy groups of
X = i≥0 Xi are trivial to show that it is contractible. Let ϕ : Sn → X be a map. By cellular
S
approximation, we may assume that is cellular. Since the image of ϕ is compact, it intersects
finitely many n-cells of X, so the image lives inside some Xk . Since Xk ,→ Xk+1 ,→ X is
nullhomotopic, ϕ is also nullhomotopic, so πn (X) = 0.
12. Exercise. Use the extension lemma to show that a CW complex retracts onto any contractible
subcomplex.
Solution. Let X be a CW complex with a contractible subcomplex A. Since A is contractible,
it is path-connected. The identity map A → A can be extended to a map X → A since all of
the homotopy groups of A are trivial (Lemma 4.7). This extension is the desired retraction.
 4.1. HOMOTOPY GROUPS 18

13. Exercise. Use cellular approximation to show that the n-skeletons of homotopy equivalent CW
complexes without cells of dimension n + 1 are also homotopy equivalent.
Solution. Let f : X → Y be a homotopy equivalence between two CW complexes X and Y
with homotopy inverse g. By cellular approximation, we may assume that both f and g are
cellular maps, i.e., define maps X n → Y n and Y n → X n . The homotopy h : X × [0, 1] → X
from f to g can also be replaced by a cellular map, the image lies inside of X n+1 = X n since X
has no cells of dimension n + 1. Hence gf restricted to X n is homotopic to the identity map.
Similarly, f g restricted to Y n is homotopic to the identity map, so X n ' Y n .

14. Exercise. Show that every map f : Sn → Sn is homotopic to a multiple of the identity map by
the following steps.

(a) Reduce to the case that there exists a point q ∈ Sn with f −1 (q) = {p1 , . . . , pk } and f is an
invertible map near each pi .
(b) For f as in (a), consider the composition gf where g : Sn → Sn collapses the complement
of a small ball about q to the basepoint. Use this to reduce (a) further to the case k = 1.
(c) Finish the argument by showing that an invertible n × n matrix can be joined by a path of
such matrices to either the identity matrix or the matrix of a reflection.

Solution. By Theorem 2C.1, f is homotopic to a map that is simplicial with respect to some
iterated barycentric subdivision of Sn . Hence there exists a point q ∈ Sn such that f −1 (q) =
{p1 , . . . , pk }, and f is a piecewise linear map around each pi and hence is locally invertible. For
each i, we can intersect the images of the neighborhoods around pi on which f is invertible to
get a small ball around q. Let g be the map which collapses the complement of this small ball to
the basepoint. For each i, we then get a map fi : Sn → Sn such that fi−1 (q) = {pi } by identifying
the neighborhood around pi with Sn (by collapsing its boundary to a point) and letting fi be
the restriction of gf . Then f is homotopic to the sum of the fi , so we can reduce to the case
k = 1. Thinking of p and q as points at infinity, and using the fact that f is linear, we can
think of Sn \ p → Sn \ q as an invertible n × n matrix. Using Gaussian elimination, we can
find a piecewise linear path from such a matrix either to the identity matrix, or the matrix of a
reflection, depending on the sign of its determinant. Such a path gives a homotopy of f either
to the identity map or the reflection, which is −1 times the identity map.

19. Exercise. Consider the equivalence relation 'w generated by weak homotopy equivalence:
X 'w Y if there are spaces X = X1 , X2 , . . . , Xn = Y with weak homotopy equivalences Xi →
Xi+1 or Xi ← Xi+1 for each i. Show that X 'w Y if and only if X and Y have a common CW
approximation.
Solution. A CW approximation of X comes with a weak homotopy equivalence, so if X and
Y have a common CW approximation, then X 'w Y by definition. Conversely, suppose that
X 'w Y . We wish to show that X and Y have a common CW approximation. Without loss
of generality, we may assume that we have a weak homotopy equivalence g : X → Y . Let X 0
and Y 0 be CW approximations for X and Y . Then by Proposition 4.18, there exists a map
h : X 0 → Y 0 such that the diagram
f1
X0 /X

h g
 f2 
Y0 /Y
 4.2. ELEMENTARY METHODS OF CALCULATION 19

commutes up to homotopy. On homotopy groups, f1∗ , g, and f2∗ are isomorphisms, so h∗ is

also an isomorphism. By Whitehead’s theorem, h is a homotopy equivalence, so X and Y have
a common CW approximation.

20. Exercise. Show that [X, Y ] is finite if X is a finite connected CW complex and πi (Y ) is finite
for i ≤ dim X.
Solution. Given a map f : X → Y and a cell X 0 of X, there are only finitely many maps
X 0 → X up to homotopy that the restriction f |X 0 could be because one can think of this map
as a composition of the attaching map for X 0 with f . Hence, there are only finitely many maps
f up to homotopy because the homotopies on the individual cells are relative to their boundary,
so this shows that they determine f up to homotopy. Hence [X, Y ] is finite.

4.2. Elementary Methods of Calculation

1. Exercise. Use homotopy groups to show there is no retraction RPn → RPk if n > k > 0.
Solution. The quotient map Sn → RPn is a covering space whose fiber F consists of two
points with the discrete topology. Hence we have isomorphisms πi (Sn ) ∼
= πi (RPn ) (Proposition
4.1) for i > 1. In particular, if there were a retraction r : RPn → RPk , then there is a map
i : RPk → RPn such that r ◦ i is the identity on RPk . On homotopy groups, this becomes the
fact that πi (RPk ) → πi (RPn ) → πi (RPk ) is the identity map. In particular, if i = k, then
πk (RPk ) = Z and πk (RPn ) = 0 for n > k, so this is a contradiction.

4. Exercise. Let X ⊂ Rn+1 be the union of the infinite sequence of spheres Snk of radius k1 and
centerQ( k1 , 0, . . . , 0). Show that πi (X) = 0 for i < n and construct a homomorphism from πn (X)
onto k πn (Snk ).
Solution. Since Si is compact, the image of any map Si → X can only intersect finitely many
of the Snk , so if i < n, the images on each such Snk can be homotoped to the origin, and hence
is homotopic to a constant map. So πi (X) = 0 for i < n. We can divide theWcube I n into parts
Ikn = {(x1 , . . . , xn ) ∈ I n | 2−k ≤ x1 ≤ 2−k−1 } for k ≥ 0. DefineWa map XQ→ k Snk by choosing a
basepoint for each Snk . Then compose this Q with the inclusion k Snk Q
→ k Snk . Call the induced
map on homotopy groups p : πn (X) → k πn (Snk ). An element of k πn (Snk ) is a sequence of
integers (a1 , a2 , . . . ). Define a map f : I n → X by defining the restriction f : Ikn → Snk to be a
degree ak map (here we are identifying Ikn with I n via some homeomorphism that preserves the
boundaries). Then p(f ) = (a1 , a2 , . . . ), so p is surjective.

6. Exercise. Show that the relative form of the Hurewicz theorem in dimension n implies the
absolute form in dimension n − 1 by considering the pair (CX, X) where CX is the cone on X.
Solution. Let X be a (n − 1)-connected space. Since CX is contractible, by the long exact
sequence of homotopy groups of a pair, we see that (CX, X) is n-connected, and πn−1 (X) ∼ =
πn (CX, X). So by the relative Hurewicz, Hi (CX, X) = 0 for i < n and πn (CX, X) ∼
= Hn (CX, X).
Now by the long exact sequence on homology for a pair, Hi (X) ∼= Hi+1 (CX, X), hence we deduce
the absolute Hurewicz in dimension n − 1.

8. Exercise. Show the suspension of an acyclic CW complex is contractible.

Solution. Let X be an acyclic CW complex, i.e., H e i (X) = 0 for all i. This means that X
is a connected space. By the Freudenthal suspension theorem (Corollary 4.24), we have an
isomorphism π0 (X) ∼= π1 (SX), and a surjection π1 (X) → π2 (SX). Since the Abelianization
 4.2. ELEMENTARY METHODS OF CALCULATION 20

of π1 (X) is H1 (X) = 0, and π2 (SX) is Abelian, this implies that π2 (SX) = 0, so SX is 2-

connected. We claim that H e i (X) ∼
=He i+1 (SX) for all i, so SX is also an acyclic CW complex.
To see this, by Proposition 2.22, we have Hi (X × [0, 1], X × {0, 1}) ∼ =H e i (SX) for all i. Then
we have the desired isomorphism by considering the long exact sequence on homology for the
pair (X × [0, 1], X × {0, 1}): the map Hi (X × {0, 1}) → Hi (X × [0, 1]) can be thought of as
Hi (X) ⊕ Hi (X) → Hi (X × [0, 1]) with inclusion for each factor. So this map is surjective, and
its kernel is the subgroup generated by (x, −x), so is isomorphic to Hi (X). So this shows the
claim. Then by the Hurewicz theorem, we must have that πi (SX) = 0 for all i, which means
that SX is contractible by Whitehead’s theorem.

9. Exercise. Show that a map between simply-connected CW complexes is a homotopy equivalence

if its mapping cone is contractible. Use the preceding exercise to give an example where this
fails in the nonsimply-connected case.
Solution. Let f : X → Y be a map of simply-connected CW complexes, and let Mf be its
mapping cylinder with mapping cone Mf /X. Note that Mf is simply-connected since it is
homotopy equivalent to Y . Suppose that Mf /X is contractible. Then the inclusion X → Mf
induces isomorphisms on homology Hn (X) → Hn (Mf ) for all n, so is a homotopy equivalence
(Corollary 4.33). This implies that f is also a homotopy equivalence.
Now let X be a noncontractible acyclic CW complex (Example 2.38). Then the suspension map
f : X → SX gives rise to a contractible mapping cylinder since SX is contractible (Ex. 4.2.8),
but f cannot be a homotopy equivalence.

10. Exercise. Let the CW complex X be obtained from S1 ∨ Sn , n ≥ 2, by attaching a cell en+1 by
a map representing the polynomial p(t) ∈ Z[t, t−1 ] ∼
= πn (S1 ∨ Sn ), so πn (X) ∼
= Z[t, t−1 ]/(p(t)).
0
Show πn (X) is cyclic and compute its order in terms of p(t). Give examples showing that
the group πn (X) can be finitely generated or not, independently of whether πn0 (X) is finite or
infinite.
Solution. Since πn (S1 ∨Sn ) is a free Z[t, t−1 ]-module on one generator, the map πP 0
n (X) → πn (X)
0
is obtained by substituting 1 for t. So the relation p(t) = 0 in πn (X) P becomes ai = 0 where
the ai are the coefficients 0
of p. So πn (X) is cyclic of infinite order if ai = 0, and is cyclic of
P
finite order c if ai = c 6= 0.
If for example p(t) = 0, then πn (X) = Z[t, t−1 ] is not finitely generated, and πn0 (X) = Z is
infinite. On the other hand, if p(t) = t − 1, then πn (X) ∼= Z is finitely generated, but πn0 (X) is
infinite. If p(t) = t, then πn (X) = 0 is finitely generated, and πn0 (X) = 0 is finite. Finally, if
p(t) = t2 +t+1, then πn (X) is not finitely generated since {t−1 , t−2 , . . . } has no finite generating
set, but πn0 (X) = Z/3 is finite.

12. Exercise. Show that a map f : X → Y of connected CW complexes is a homotopy equivalence

e → Ye to the universal covers induces an
if it induces an isomorphism on π1 and if a lift fe: X
isomorphism on homology.
Solution. The commutative diagram

fe
/ e
X
e Y

 f 
X / Y,
 4.2. ELEMENTARY METHODS OF CALCULATION 21

where the vertical maps are the covering maps, induces a commutative diagram

fe∗
πi (X) / π (Y
e)
i
e

 f∗ 
πi (X) / πi (Y ).

The vertical maps are isomorphisms for i > 1 (Proposition 4.1). Since X e and Ye are simply-
connected CW complexes, fe is a homotopy equivalence (Corollary 4.33). Hence fe∗ is an isomor-
phism for all i, which implies that f∗ is an isomorphism for all i > 1 by the commutativity of the
diagram. By assumption, f∗ is also an isomorphism for i = 1, so f is a homotopy equivalence.

13. Exercise. Show that a map between connected n-dimensional CW complexes is a homotopy
equivalence if it induces an isomorphism on πi for i ≤ n.
Solution. Let f : X → Y be a map between connected n-dimensional CW complexes which
induces an isomorphism on πi for i ≤ n. Passing to universal covers, and taking a lift fe: X
e → Ye ,
we get a map which induces isomorphisms on πi for i ≤ n. By the Hurewicz theorem, this implies
e → Hi (Ye ) for i ≤ n. It is also an isomorphism for i > n since
that fe is an isomorphism Hi (X)
the homology vanishes in these degrees (this can be seen by cellular homology). Hence f is a
homotopy equivalence (Ex. 4.2.12).

15. Exercise. Show that a closed simply-connected 3-manifold is homotopy equivalent to S3 .

Solution. Let M be a closed simply-connected 3-manifold. First, M is homotopy equivalent to
a CW complex. Second, M is orientable since it is simply-connected (otherwise, the orientation
covering would be connected). Since π1 (M ) = 0, we have H1 (M ) = 0, so H2 (M ) = 0 by Poincaré
duality, and the top homology is H3 (M ) = Z. Now let f : S3 → M be a map of degree 1. This
exists because π3 (M ) = Z by the Hurewicz theorem. Then f induces isomorphisms on homology,
so is a homotopy equivalence because M and S3 are simply-connected.

16. Exercise. Show that the closed surfaces with infinite fundamental group are K(π, 1)’s by
showing that their universal covers are contractible, via the Hurewicz theorem and results of
section 3.3.
Solution. Let X be a closed surface with infinite fundamental group, and let X e be its universal
cover. Since π1 (X) is infinite, X is not compact. Hence, H2 (X) = 0 (Proposition 3.29). Also,
e e
H1 (X)
e = 0 since π1 (X)e = 0, so X e is contractible since it has the homotopy type of a CW
complex. Hence X is a K(π1 (X), 1).

18. Exercise. If X and Y are simply-connected CW complexes such that H e i (X) and H e j (Y ) are
finite and of relatively prime orders for all pairs (i, j), show that the inclusion X ∨ Y ,→ X × Y
is a homotopy equivalence and X ∧ Y is contractible.
Solution. By the Künneth formula, H e n (X × Y ) ∼=H e n (X) ⊕ H
e n (Y ). Since the is the image of
the map Hn (X ∨ Y ) → Hn (X × Y ) induced by the inclusion, this is an isomorphism. Hence
e e
X ∨ Y ,→ X × Y is a homotopy equivalence. Also, X ∧ Y is contractible since H e n (X ∧ Y ) = 0 by
the long exact sequence of the pair (X × Y, X ∨ Y ) and the fact that X × Y /X ∨ Y = X ∧ Y .

20. Exercise. Let G be a group and X a simply-connected space. Show that for the product
K(G, 1) × X the action of π1 on πn is trivial for all n > 1.
 4.2. ELEMENTARY METHODS OF CALCULATION 22

Solution. An element f ∈ πn is represented by maps f1 : I n → K(G, 1) and f2 : I n → X which

map the boundary to fixed basepoints, and similarly, an element of γ ∈ π1 is represented by
maps γ1 : [0, 1] → K(G, 1) and γ2 : [0, 1] → X which map the boundary to the basepoints. The
action of γ on f is then obtained by shrinking the domain of I n for fi homeomorphically and
inserting γi into the remainder of I n . Since πn (K(G, 1)) = 0 for n > 1, γ1 f1 is homotopic to f1 .
Since γ2 is homotopic to a constant map, γ2 f2 is also homotopic to f2 , so γf is homotopic to
f.

21. Exercise. Given a sequence of CW complexes K(Gn , n), n = 1, 2, . . . , let Xn be the CW

complex formed by the product of the first n of these K(Gn , n)’s. Via the inclusions Xn−1 ⊂ Xn
coming from regarding Xn−1 as the subcomplex of Xn with nth coordinate equal to a basepoint
0-cell of K(Gn , n), we can then form the union of all the Xn ’s, a CW complex X. Show
πn (X) ∼= Gn for all n.
Solution. Let f : Sn → X be a map. By compactness, the image of Sn must lie inside of some
Xm . If m < n, then f is homotopic to a constant map, otherwise, we have πn (Xm ) ∼
= Gn .

22. Exercise. Show that Hn+1 (K(G, n); Z) = 0 if n > 1.

Solution. Let M (G, n) be a Moore space. Since M (G, n) is simply-connected for n > 1, we can
use the Hurewicz theorem to deduce that M (G, n) is (n − 1)-connected. To turn M (G, n) into
a K(G, n), we can attach cells of dimensions ≥ n + 2 to kill the higher homotopy groups. Doing
this does not affect the homology in degrees ≤ n + 1, so we conclude that Hn+1 (K(G, n)) =
Hn+1 (M (G, n)) = 0.

23. Exercise. Extend the Hurewicz theorem by showing that if X is an (n − 1)-connected CW

complex, then the Hurewicz homomorphism h : πn+1 (X) → Hn+1 (X) is surjective when n > 1,
and when n = 1 show there is an isomorphism H2 (X)/h(π2 (X)) ∼
= H2 (K(π1 (X), 1)).
Solution. First, we can build a K(πn (X), n) from X by attaching cells of dimension ≥ n + 2.
Let Y be the result of attaching all of the (n + 2)-cells of K(πn (X), n) to X. From the naturality
of the Hurewicz homomorphism, the diagram

πn+2 (Y, X) / πn+1 (X) / πn+1 (Y ) / πn+1 (Y, X)

h
   
Hn+2 (Y, X)
∂ / Hn+1 (X) / Hn+1 (Y ) /0

commutes. That Hn+1 (Y, X) = 0 comes from the fact that Y and X have the same (n + 1)-cells,
and hence Cn+1 (Y, X) = 0. From the definition of the Hurewicz homomorphism, the images
of h : πn+1 (X) → Hn+1 (X) and ∂ : Hn+2 (Y, X) → Hn+1 (X) coincide since ∂ sends a relative
cycle in Y to its boundary in X. Note also that Hn+1 (Y ) ∼
= Hn+1 (K(πn (X), n)) by cellular
homology. So from the above, we get Hn+1 (X)/h(πn+1 (X)) ∼ = Hn+1 (K(πn (X), n)). Hence if
n > 1, then by (Ex. 4.2.22), Hn+1 (Y ) = 0, so ∂, and hence h, is surjective. If n = 1, this
becomes H2 (X)/h(π2 (X)) ∼
= H2 (K(π1 (X), 1)).

26. Exercise. Generalizing the example of RP2 and S2 ×RP∞ , show that if X is a connected finite-
dimensional CW complex with universal cover X,e then X and X e × K(π1 (X), 1) have isomorphic
homotopy groups but are not homotopy equivalent if π1 (X) contains elements of finite order.
Solution. It is immediate that X and X e × K(π1 (X), 1) have isomorphic homotopy groups since
e ∼
πi (X) = πi (X) for i > 2, and π1 (X)
e = 0.
 4.2. ELEMENTARY METHODS OF CALCULATION 23

Suppose that π1 (X) has elements of finite order. Then by Proposition 2.45, K(π1 (X), 1) must
be an infinite-dimensional CW complex. By the Künneth formula, K(π1 (X), 1) has infinitely
many nontrivial homology groups (for example, this homology agrees with the group homology
of π1 (X), and the group homology of finite cyclic groups satisfies this property), while X has only
finitely many since it is finite-dimensional. Hence X and X e × K(π1 (X), 1) cannot be homotopy
equivalent spaces.
27. Exercise. From Lemma 4.39 deduce that the image of the map π2 (X, x0 ) → π2 (X, A, x0 ) lies
in the center of π2 (X, A, x0 ).
Solution. By the long exact sequence of homotopy groups for a pair (X, A), the image of the
map π2 (X, x0 ) → π2 (X, A, x0 ) is equal to the kernel of the boundary map ∂ : π2 (X, A, x0 ) →
π1 (A, x0 ). Pick x ∈ ker ∂. Then xbx−1 = e for all b ∈ π2 (X, A, x0 ) by Lemma 4.39, so x is in
the center of π2 (X, A, x0 ).
28. Exercise. Show that the group Z/p × Z/p with p prime cannot act freely on any sphere Sn ,
by filling in the details of the following argument. Such an action would define a covering space
Sn → M with M a closed manifold. When n > 1, build a K(Z/p × Z/p, 1) from M by attaching
a single (n + 1)-cell and then cells of higher dimension. Deduce that Hn+1 (K(Z/p × Z/p, 1); Z/p)
is Z/p or 0, a contradiction.
Solution. Suppose Z/p × Z/p acts freely on Sn . Thinking of Z/p × Z/p as a 0-dimensional
Lie group, we have a free, proper, smooth action on Sn , which gives a smooth submersion
Sn → M = Sn /(Z/p×Z/p). In particular, M is a smooth compact n-manifold. Since dim M = n,
this is a covering space map whose group of deck transformations is Z/p × Z/p. In particular,
π1 (M ) = Z/p × Z/p.
If n = 1, then M must be homotopy equivalent to either S1 or [0, 1] since these are the only
connected compact 1-manifolds, so we have a contradiction since their respective fundamental
groups are Z and 0.
If n > 1, then we can make M a K(Z/p × Z/p, 1) by attaching a single (n + 1)-cell and then
attaching more cells of higher dimension. This shows that Hn+1 (K(Z/p × Z/p, 1); Z/p) is either
Z/p or 0 by cellular homology. But this contradicts the group cohomology of Z/p × Z/p (which
is bigger than Z/p by the Künneth formula).
31. Exercise. For a fiber bundle F → E → B such that the inclusion F ,→ E is homotopic to a
constant map, show that the long exact sequence of homotopy groups breaks up into split short
exact sequences giving isomorphisms πn (B) ∼= πn (E) ⊕ πn−1 (F ). In particular, for the Hopf
3 7 4 7 15 8
bundles S → S → S and S → S → S this yields isomorphisms
πn (S4 ) ∼
= πn (S7 ) ⊕ πn−1 (S3 )
πn (S8 ) ∼
= πn (S15 ) ⊕ πn−1 (S7 )
Thus π7 (S4 ) and π15 (S8 ) contain Z summands.
Solution. The maps πn (F ) → πn (E) in the long exact sequence of homotopy groups for a Serre
fibration are induced by the inclusion F → E, so if this is homotopic to a constant map, then
the induced map is 0. Hence for all n > 0, we have short exact sequences
0 / πn (E) / πn (B) / πn−1 (F ) /0.

Since E → B has the homotopy lifting property with respect to all disks, we can find a section
πn (B) → πn (E) for the induced map πn (B) → πn (E), which means that the above short exact
sequence splits.
 4.3. CONNECTIONS WITH COHOMOLOGY 24

32. Exercise. Show that if Sk → Sm → Sn is a fiber bundle, then k = n − 1 and m = 2n − 1.

Solution. We have the relations n ≤ m and k ≤ m and k +n = m. If k = m, then n = 0, and S0
is not connected, so this contradicts that Sm → Sn is surjective. So k < m, and hence Sk → Sm
is homotopic to a constant map. From (Ex. 4.2.31), we have πi (Sn ) ∼ = πi (Sm ) ⊕ πi−1 (Sk ) for all
i > 0. This shows that k > 0, so m > n. In particular, considering values of i = 1, . . . , n, we see
that πi (Sk ) = 0 if i < n − 1 and πn−1 (Sk ) = Z, so k = n − 1. Hence m = 2n − 1.

33. Exercise. Show that if there were fiber bundles Sn−1 → S2n−1 → Sn for all n, then the groups
πi (Sn ) would be finitely generated free Abelian groups computable by induction, and nonzero
for i ≥ n ≥ 2.
Solution. Assuming that fiber bundles Sn−1 → S2n−1 → Sn exist for all n, we can compute
πi (Sn ) by double induction on i − n and n. Of course, if i − n < 0, then πi (Sn ) = 0, and if
i − n = 0, then πn (Sn ) = Z. Using (Ex. 4.2.31), we have πi (Sn ) ∼ = πi (S2n−1 ) ⊕ πi−1 (Sn−1 ) for
m
all n and i > 0. So if πj (S ) is a finitely generated free Abelian group for all j − m < i − n and
m < n, then this shows that πi (Sn ) is also a finitely generated free Abelian group.

34. Exercise. Let p : S3 → S2 be the Hopf bundle and let q : T 3 → S3 be the quotient map collapsing
the complement of a ball in the 3-dimensional torus T 3 = S1 × S1 × S1 to a point. Show that
pq : T 3 → S2 induces the trivial map on π∗ and H
e ∗ , but is not homotopic to a constant map.
Solution. The only nontrivial homotopy group of T 3 is π3 (T 3 ) ∼ = Z3 . The map q∗ : π3 (T 3 ) →
π3 (S3 ) is zero because any loop that goes around one of the factors of S1 in T 3 can be homotoped
to miss the ball that is used in the quotient map T 3 → S3 . Hence pq induces the trivial
map on all homotopy groups. Similarly, the only nontrivial reduced homology group of S2 is
e 2 (S2 ) = Z. The map that pq induces on homology factors as H
H e 2 (T 3 ) → H
e 2 (S3 ) → H
e 2 (S2 ), but
since He 2 (S3 ) = 0, this composition is 0. Hence pq also induces trivial maps on reduced homology.
Note however, that a homotopy from pq to a constant map would give a homotopy from p to a
constant map, so pq is not homotopic to a constant map.

4.3. Connections with Cohomology

1. Exercise. Show there is a map RP∞ → CP∞ = K(Z, 2) which induces the trivial map on
H
e ∗ (−; Z) but a nontrivial map on He ∗ (−; Z). How is this consistent with the universal coefficient
theorem?
Solution. Note that H e n (RP∞ ; Z) is Z/2 for n odd and 0 otherwise, and that H e n (CP∞ ; Z)
is Z for n > 0 even and 0 otherwise, both of which can be seen by cellular homology and
the fact that RP∞ can be taken to have one cell for each dimension (the attaching map has
degree 2), and CP∞ has one cell for each even dimension. Hence every map RP∞ → CP∞
induces a trivial map on reduced homology. The homotopy classes of maps RP∞ → CP∞ are
in bijection with cohomology classes β ∈ H2 (RP∞ ; Z); in particular, there is a distinguished
class α ∈ H2 (CP∞ ; Z) such that f : RP∞ → CP∞ gives f ∗ (α) = β (Theorem 4.57). Since the
cohomology group H2 (RP∞ ; Z) is nonzero, we can find thus find a map RP∞ → CP∞ which
is nontrivial on cohomology groups.
This is consistent with the universal coefficient theorem because Hom(Z/2, Z) = 0.

2. Exercise. Show that the group structure on S1 coming from multiplication in C induces a
group structure on hX, S1 i such that the bijection hX, S1 i → H1 (X; Z) of Theorem 4.57 is an
isomorphism.
4.3. CONNECTIONS WITH COHOMOLOGY 25

Solution. Given two maps f, g : X → S1 , let their sum f +g be defined by (f +g)(x) = f (x)g(x)
where the multiplication is in C. Suppose that f is homotopic to f 0 via H1 and g is homotopic
to g 0 via H2 . Then f + g is homotopic to f 0 + g 0 via the map H1 + H2 : X × [0, 1] → S1 which
is given by (x, t) 7→ H1 (x, t)H2 (x, t). Hence this is a well-defined (Abelian) group structure on
hX, S1 i.
Let T : hX, S1 i → H1 (X; Z) be the bijection of Theorem 4.57. Then there is a distinguished
class α ∈ H1 (S1 ; Z) such that T ([f ]) = f ∗ (α). It is clear from the definition of pullback that
f ∗ (α) + g ∗ (α) = (f + g)∗ (α), so T is a group isomorphism.

4. Exercise. Given Abelian groups G and H and CW complexes K(G, n) and K(H, n), show
that the map hK(G, n), K(H, n)i → Hom(G, H) sending a homotopy class [f ] to the induced
homomorphism f∗ : πn (K(G, n)) → πn (K(H, n)) is a bijection.
Solution. Surjectivity of the map follows from Lemma 4.31. Now suppose we have two maps
f, g : K(G, n) → K(H, n) such that f∗ = g∗ . In other words, for every basepoint-preserving map
ϕ : Sn → K(G, n), there is a homotopy Hϕ : Sn × [0, 1] → K(H, n) from f ◦ ϕ to g ◦ ϕ. Letting
ϕ vary over the characteristic maps of the n-cells of K(G, n) shows that f is homotopic to g.
More precisely, these define homotopies on the n-skeleton on K(G, n), and the homotopy on the
rest of the cells can be constructed using Lemma 4.7.

5. Exercise. Show that [X, Sn ] ∼

= Hn (X; Z) if X is an n-dimensional CW complex.
Solution. We can build a K(Z, n) from Sn by attaching cells of dimension ≥ n + 2. The
inclusion Sn ,→ K(Z, n) induces a map ϕ : [X, Sn ] → [X, K(Z, n)]. If ϕ(f ) = ϕ(g), then there is
a homotopy H : X × [0, 1] → K(Z, n) between f and g. By cellular approximation, this can be
made to have image inside of the (n + 1)-skeleton of K(Z, n), which is equal to Sn , and hence
f = g, so ϕ is injective. Surjectivity of ϕ also follows from cellular approximation since X is
n-dimensional. Thus, [X, Sn ] ∼
= [X, K(Z, n)] ∼
= Hn (X; Z).

6. Exercise. Use Exercise 4 to construct a multiplication map µ : K(G, n) × K(G, n) → K(G, n)

for any Abelian group G, making a CW complex K(G, n) into an H-space whose multiplication
is commutative and associative up to homotopy and has a homotopy inverse. Show also that
the H-space multiplication µ is unique up to homotopy.
Solution. First note that K(G, n)×K(G, n) ' K(G×G, n). By (Ex. 4.3.4), there is a bijection
hK(G×G, n), K(G, n)i ∼ = Hom(G×G, G). Let µ : K(G, n)×K(G, n) → K(G, n) be a map (well-
defined only up to homotopy) corresponding to the map G × G → G given by (g, g 0 ) 7→ g + g 0 .
From the naturality of this isomorphism, it follows that µ is associative and commutative up to
homotopy. There is a homotopy inverse given by letting i : K(G, n) → K(G, n) correspond to
the homomorphism G → G given by g 7→ −g. If µ0 is another such H-space multiplication on
K(G, n), then it must correspond to the addition map G × G → G by (Ex. 4.1.3).

7. Exercise. Using an H-space multiplication µ on K(G, n), define an addition in hX, K(G, n)i
by [f ] + [g] = [µ(f, g)] and show that under the bijection Hn (X; G) ∼
= hX, K(G, n)i this addition
corresponds to the usual addition in cohomology.
Solution. This follows as in (Ex. 4.3.2).

8. Exercise. Show that a map p : E → B is a fibration if and only if the map π : E I → Ep ,

π(γ) = (γ(0), pγ), has a section.
 4.3. CONNECTIONS WITH COHOMOLOGY 26

Solution. First suppose that π has a section s : Ep → E I . Let gt : X → B be a homotopy, and

ge0 : X → E be a lift of g0 . Then define γ et : X → E by x 7→ (s(e γ0 (x), γt ))(t). Since s is a section
of π, we have p ◦ s(e
γ0 (x), γt ) = γ0 (x), so γ
et is a lift of γt , and hence p is a fibration.
Now suppose that p is a fibration. Given (e, γ) ∈ Ep , i.e., e ∈ E and γ : I → B with γ(0) =
p(e), define s(e, γ) as follows. We have a map ∗ → B given by ∗ 7→ p(e), and a homotopy
ht : ∗ ×[0, 1] → B given by γ. The point e provides a lift e
h0 of h0 , and the (unique) lift e
ht of ht
I
is an element of E which we define to be s(e, γ). It follows immediately that s is a section of
π.

9. Exercise. Show that a linear projection of a 2-simplex onto one of its edges is a fibration but
not a fiber bundle.
Solution. Let ∆ be a 2-simplex, and let I be one of its edges, with p : ∆ → I linear projection.
The map π : ∆I → ∆p given by γ 7→ (γ(0), pγ) has a section s : ∆p → ∆I where s(x, γ) : I → ∆
is the map s(x, γ)(t) = γ(t). So by (Ex. 4.3.8), p is a fibration. However, it is not a fiber
bundle because the fibers over the vertices of the edge are points, while the other fibers are line
segments, and hence not homeomorphic.

11. Exercise. For a space B, let F(B) be the set of fiber homotopy equivalence classes of fibrations
E → B. Show that a map f : B1 → B2 induces f ∗ : F(B2 ) → F(B1 ) depending only on the
homotopy class of f , with f ∗ a bijection if f is a homotopy equivalence.
Solution. Given a fibration p : E → B2 , let f ∗ (E) be the pullback f ∗ (E) → B1 along f , i.e.,

f ∗ (E) /E

p
 f 
B1 / B2

is a pullback diagram. We have to show that if pE : E → B2 and pF : F → B2 are fiber homotopy

equivalent fibrations over B2 , then so are f ∗ (E) and f ∗ (F ). Let g : E → F and h : F → E be
fiber-preserving maps such that gh and hg are homotopic to the identity through fiber-preserving
maps. Let pF,1 : f ∗ (F ) → F , pF,2 : f ∗ (F ) → B1 , pE,1 : f ∗ (E) → E, and pE,2 : f ∗ (E) → B1 be the
respective projection maps. The composition f ∗ (F ) → F → E gives a commutative diagram

f ∗ (F ) /E

pF,2 pE
 f 
B1 / B2

and hence by the universal property of pullback, we have an induced map h∗ : f ∗ (F ) → f ∗ (E).
Similarly, we get an induced map g ∗ : f ∗ (E) → f ∗ (F ). Since these maps fit into diagrams
 4.3. CONNECTIONS WITH COHOMOLOGY 27

consisting of fiber-preserving maps, they are also fiber-preserving. The following diagram

f ∗ (E)
II ∗
II g
II pE,1
II
I$
f ∗ (F )
II
II h∗ h◦pF,1
II
II
I$
pE,1 " 
pE,2 f ∗ (E) /E
h◦pF,2
pE,2 pE
$)  f 
B1 / B2

commutes up to homotopy, so by uniqueness of induced maps, h∗ ◦g ∗ is homotopic to the identity

of f ∗ (E) through fiber-preserving maps. Similarly, g ∗ ◦ h∗ is homotopic to the identity of f ∗ (F )
through fiber-preserving maps. So we have a well-defined function f ∗ : F(B2 ) → F(B1 ). That
f ∗ only depends on the homotopy class of f is the content of Proposition 4.62.
Finally, it is clear that if B1 = B2 and f is the identity map, then f ∗ is also the identity map.
Also, it is clear that (f ◦ g)∗ = g ∗ ◦ f ∗ from the associativity of pullback. So if f : B1 → B2 is
a homotopy equivalence with homotopy inverse g : B2 → B1 , then f ∗ ◦ g ∗ and g ∗ ◦ f ∗ are the
identity maps on F(B1 ) and F(B2 ), respectively. Hence f ∗ is a bijection.

12. Exercise. Show that for homotopic maps f, g : A → B the fibrations Ef → B and Eg → B are
fiber homotopy equivalent.
Solution. Let H be a homotopy from f to g, and let H be the reverse homotopy from g to
f . Define a map Ef → Eg by (a, γ) 7→ (a, H(a) · γ) where the · denotes the path which travels
H(a) first at double speed, and then γ at double speed. Similarly, we can define a map Eg → Ef
by (a, γ) 7→ (a, H(a) · γ). Then it is clear that these maps are fiber-preserving and that their
compositions are homotopic to identity maps through homotopy-preserving maps, so Ef and Eg
are fiber homotopy equivalent.

13. Exercise. Given a map f : A → B and a homotopy equivalence g : C → A, show that the
fibrations Ef → B and Ef g → B are fiber homotopy equivalent.
Solution. Since A is homotopy equivalent to the mapping cylinder Mg , we may assume that
g : C → A is a deformation retract by Corollary 0.21. In this case, Ef g is a deformation retract
of Ef because f (g(C)) is a deformation retract of f (A).

14. Exercise. For a space B, let M(B) denote the set of equivalence classes of maps f : A → B
where f1 : A1 → B is equivalent to f2 : A2 → B if there exists a homotopy equivalence g : A1 →
A2 such that f1 ' f2 g. Show the natural map F(B) → M(B) is a bijection.
Solution. A fibration E → B is an element of M(B), and two fiber homotopy equivalent
fibrations are equivalent as elements of M(B). So we have a natural map F(B) → M(B). This
map is surjective because any map f : A → B is equivalent to the fibration Ef → B since the
natural inclusion A ,→ Ef is a homotopy equivalence. Injectivity follows because if two fibrations
p1 : E1 → B and p2 : E2 → B are homotopy equivalent via g : E1 → E2 such that p1 ' p2 g, then
Ep1 → B and Ep2 → B are fiber homotopic fibrations by (Ex. 4.3.13). Hence E1 → B and
E2 → B are fiber homotopic equivalent (Proposition 4.65).
 4.3. CONNECTIONS WITH COHOMOLOGY 28

15. Exercise. If the fibration p : E → B is a homotopy equivalence, show that p is a fiber homotopy
equivalence of E with the trivial fibration 1 : B → B.
Solution. In this case, p is a fiber-preserving map, and a homotopy inverse q of p can be chosen
to be fiber-preserving by (Ex. 4.3.14).

17. Exercise. Show that ΩX is an H-space with multiplication the composition of loops.
Solution. The identity is the constant loop of ΩX. Since composition of loops is associative up
to homotopy, and composition of a loop with the constant loop is homotopic to itself, this gives
an H-space structure on ΩX.

18. Exercise. Show that a fibration sequence · · · → ΩB → F → E → B induces a long exact

sequence · · · → hX, ΩBi → hX, F i → hX, Ei → hX, Bi, with groups and group homomorphisms
except for the last three terms, Abelian groups except for the last six terms.
Solution. We give hX, Ωn Ki the structure of a group as in (Ex. 4.3.2), where K is any space
and n > 0. If n > 1, then Ωn−1 K is an H-space by (Ex. 4.3.17), so composition of loops in
Ωn K is commutative up to homotopy (Ex. 3.C.5), and hence hX, Ωn Ki has the structure of an
Abelian group.
We first show that if F → E → B is a fibration, then hX, F i → hX, Ei → hX, Bi is exact. It
is obvious that the composition of the two maps is zero. Now let f ∈ hX, Ei be such that pf is
homotopic to a constant map where p : E → B is the projection. Then we can use the homotopy
lifting property to homotope f to a map that lives inside of the fiber of the basepoint of E, and
hence the sequence is exact.
A map of H-spaces E → B preserving the multiplication induces a group homomorphism
hX, Ei → hX, Bi. So we need to show that given E → B, the induced map ΩE → ΩB
preserves the multiplication. Since the multiplication is induced by composition of loops, this
follows.
p
19. Exercise. Given a fibration F → E → − B, define a natural action of ΩB on the homotopy fiber
Fp and use this to show that exactness at hX, F i in the long exact sequence in the preceding
problem can be improved to the statement that two elements of hX, F i have the same image in
hX, Ei if and only if they are in the same orbit of the induced action of hX, ΩBi on hX, F i.
Solution. Pick γ ∈ ΩB and (e, η) ∈ Fp , i.e., η : [0, 1] → B such that η(0) = p(e) and η(1) = b0 ,
where b0 ∈ B is the basepoint. Note that η · γ · η is a homotopy from p(e) to itself, and e is a
lift of p(e), so by the homotopy lifting property, we get a homotopy ηe : [0, 1] → E lifting η · γ · η.
Define γ · (e, η) = (eη (1), η). The endpoint is independent of the lifting chosen. This defines an
action of ΩB on Fp since (γ · γ 0 ) · (e, η) = γ · (γ 0 · (e, η)) by definition.
Two elements f, g ∈ hX, F i have the same image in hX, Ei if and only if there is a homotopy
H : X × [0, 1] → E through basepoint-preserving maps from f to g. Such a homotopy is the
same as the existence of an action of an element of ΩB taking f (x) to g(x) for all x ∈ X, and
hence f and g are in the same orbit of ΩB.

20. Exercise. Show that by applying the loopspace functor to a Postnikov tower for X one obtains
a Postnikov tower of principal fibrations for ΩX.
Solution. Let · · · → X2 → X1 be a Postnikov tower for X. Applying the loopspace functor
gives a Postnikov tower · · · → ΩX2 → ΩX1 for ΩX. By the discussion on p. 409, ΩXn+1 →
ΩXn → ΩXn−1 is a principal fibration for all n > 1.
 4.3. CONNECTIONS WITH COHOMOLOGY 29

23. Exercise. Prove the following uniqueness result for the Quillen plus construction: Given a
connected CW complex X, if there is an Abelian CW complex Y and a map X → Y inducing
an isomorphism H∗ (X; Z) ∼
= H∗ (Y ; Z), then such a Y is unique up to homotopy equivalence.
Solution. Let X → Y 0 be another map which induces isomorphisms on homology such that Y 0
is an Abelian CW complex. Let W be the mapping cylinder of X → Y . Then Hn (W ) ∼ = Hn (X)
via the inclusion X ,→ W by hypothesis that X → Y induces isomorphisms on homology. Hence
Hn+1 (W, X; πn (Y 0 )) = 0 for all n by the long exact sequence on cohomology for the pair (W, X).
By Corollary 4.73, there is a lift W → Y 0 of the map X → Y 0 . In particular, this means that we
have a map Y → Y 0 commuting with the maps X → Y and X → Y 0 . So this map must induced
isomorphisms on homology, and hence is a homotopy equivalence (Proposition 4.74).