MATHEMATICS 210A: HOMEWORK 1

LAURA LYMAN

For all exercises, assume that X is the underlying topological space and I = [0, 1] unless otherwise
specified.
Hatcher Section 1.1.
(1) Show that the composition of paths satisfies the following cancellation property: If f0 ·g0 ' f1 ·g1
and g0 ' g1 then f0 ' f1 .
Proof. First we prove the following lemma about inverse paths.
Lemma. Let f, g : I → X be paths such that f ' g. Let f , g : I → X be the corresponding
inverse paths as defined in Hatcher in the proof of Proposition 1.3; that is,
f : I → X, f (s) = f (1 − s)
g : I → X, g(s) = g(1 − s).
Then f ' g as well.
Proof. Let H : I×I → X be a homotopy between f and g, so H(s, 0) = f (s) and H(s, 1) = g(s).
Then consider the map
H : I × I → X s.t. H(s, t) = H(1 − s, t).
Then H is continuous since H is continuous (since H is the composition of continuous func-
tions). Furthermore, H(s, 0) = H(1 − s, 0) = f (1 − s) = f (s) and similarly H(s, 1) = g(s).
Ht
Letting Ht = H( · , t), we have that (since f ' g) Ht (0) = x0 and Ht (1) = x1 for some
x0 , x1 ∈ X independent of t. Thus
H(0, t) = H(1, t) = Ht (1) = x1 and H(1, t) = H(0, t) = Ht (0) = x0
meaning the endpoints of H(s, t) are fixed in time. Hence H( · , t) is a homotopy between f
and g, so f ' g. 
Now suppose f0 · g0 ' f1 · g1 for paths f0 , f1 , g0 , g1 : I → X with f0 (1) = g0 (0), f1 (1) = g1 (0)
(so the path compositions are defined). Furthermore, suppose g0 ' g1 . Consider the inverse
paths g0 , g1 . Since g0 ' g1 , by the lemma above, g0 ' g1 .
Let c : I → X be the constant path such that c(s) = f0 (1). Then, as seen in the proof of
Proposition 1.3, f0 · c is a reparametrization of f0 , and thus f0 ' f0 · c. Similarly, f1 ' f1 · d
where d(s) = f1 (1) is a constant path. Furthermore, within the same proof, it is shown that
g0 · g0 ' e
c where e c(s) = g0 (0) for all s ∈ I. However, since g0 (0) = f0 (1), we have that c = e c.
Similarly d = d where g1 · g1 ' d. So by substituting, and by the symmetry and transitivity of
e e
', we have that
f0 ' f0 · c = f0 · e c
' f0 · (g0 · g0 ) [since g0 · g0 ' e
c]
' (f0 · g0 ) · g0 [path composition is associative]
' (f1 · g1 ) · g0 [since f0 · g0 ' f1 · g1 ]
' (f1 · g1 ) · g1 [since g0 ' g1 ]
' f1 · (g1 · g1 )
' f1 · de = f1 · d ' f1
which concludes the proof. 
(2) Show that change-of-basepoint homomorphism βh depends only on the homotopy class of h.
1
 2 LAURA LYMAN

Proof. The change-of-basepoint homomorphism βh : π1 (X, x1 ) → π1 (X, x0 ) is defined for a

path h : I → X by
βh [f ] = [h · f · h].
Suppose h1 ' h2 for paths h1 , h2 . By the lemma in problem (1), h1 ' h2 ; that is, the inverse
paths are homotopic. Then for any [f ] ∈ π1 (X, x1 ), f · h1 ' f · h2 . Since h1 ' h2 , we then
have that
h1 · (f · h1 ) ' h2 · (f · h2 ).
So h1 · f · h1 and h2 · f · h2 are in the same homotopy class, meaning [h1 · f · h1 ] = [h2 · f · h2 ].
Therefore,
βh1 [f ] = [h1 · f · h1 ] = [h2 · f · h2 ] = βh2 [f ]
when h1 ' h2 . Hence, βh only depends on the homotopy class of h. 
∼
(10) From the isomorphism π1 (X × Y, (x0 , y0 )) = π1 (X, x0 ) × π1 (Y, y0 ) it follows that loops in
X × {y0 } and {x0 } × Y represent commuting elements of π1 (X × Y, (x0 , y0 )). Construct an
explicit homotopy demonstrating this.
Proof. Let ` : I → X × {y0 } and `0 : I → {x0 } × Y be loops each based at the point (x0 , y0 ).
Then their composition `0 · ` is defined since `(0) = (x0 , y0 ) = `0 (1). Define the projections of `
and `0 onto their components in X and Y respectively; that is, for any s ∈ I the loops `proj (s)
and `0proj (s) are defined such that
`(s) = (`proj (s), y0 ) and `0 (s) = (x0 , `0proj (s)).
Consider the maps

x0
 s ∈ [0, 2t ]
f : I × I → X s.t. f (s, t) = `proj (2s − t) s ∈ [ 2t , 1+t
2 ]
1+t

x0 s ∈ [ 2 , 1]


y0
 s ∈ [0, 1−t 2 ]
g : I × I → Y s.t. g(s, t) = 0 1−t
`proj (2s + t − 1) s ∈ [ 2 , 2−t 2 ]
2−t

y0 s ∈ [ 2 , 1].


First we verify that f is continuous by showing that f (s, ·) is continuous for any fixed s ∈ I and
f (·, t) is continuous for any fixed t ∈ I. Clearly f (·, t) is continuous on [0, 2t ] and [ 1+t
2 , 1] since it
is constant on those intervals. On [ 2t , 1+t 2 ], f (·, t) is also continuous since it is the composition
of continuous functions (`proj is a path and thus continuous). At the first boundary point,
f ( 2t , t) = x0 = `proj (0) = `proj (2( 2t ) − t) since `(0) = (x0 , y0 ). In addition for s = 1+t 2 ,
`proj (2s − t) = `proj (1) = x0 . Therefore, f (·, t) is continuous on I.
Similarly, fix s ∈ I and consider f (s, ·). Then any input t ∈ I satisfies at least one of the
following: 0 ≤ t ≤ 2s − 1, 2s − 1 ≤ t ≤ 2s, or 2s ≤ t ≤ 1. For this first interval, f (s, t) = x0
since t ≤ 2s − 1 ⇔ t+1 1+t
2 ≤ s ≤ 1 (since s ∈ [0, 1]) ⇔ s ∈ [ 2 , 1]. Similarly, f (s, t) = x0 on the
last interval. For 2s − 1 ≤ t ≤ 2s, s ≤ 2 (first inequality) and s ≥ 2t (second inequality),
1+t

so f (s, t) = `proj (2s − t) on this interval of t, which is continuous. Finally, for the boundary
points t = 2s − 1 and t = 2s, we have that f (s, t) = `proj (2s − (2s − 1)) = `proj (1) = x0 and
f (s, t) = `proj (2s − 2s) = `proj (0) = x0 respectively. Hence, f (s, ·) is continuous as well. Hence
f is continuous on I × I.
By an analogous argument, g is continuous on I × I. Thus the proposed homotopy
F : I × I → X s.t. F (s, t) = (f (s, t), g(s, t))
is continuous. Also, the endpoints of F are fixed in time since
F (0, t) = (f (0, t), g(0, t)) = (x0 , y0 ) and F (1, t) = (f (1, t), g(1, t)) = (x0 , y0 ).
Finally, by definition of path composition
( (
0 `(2s) s ∈ [0, 12 ] (`proj (2s), y0 ) s ∈ [0, 12 ]
` · ` (s) = 0 =
` (2s − 1) s ∈ [ 12 , 1] (x0 , `0proj (2s − 1)) s ∈ [ 21 , 1]
 MATHEMATICS 210A: HOMEWORK 1 3

and similarly (
(x0 , `0proj (2s)) s ∈ [0, 12 ]
`0 · `(s) =
(`proj (2s − 1), y0 ) s ∈ [ 21 , 1].
Therefore,
(
(`proj (2s − 0), y0 ) s ∈ [0, 12 ]
F (s, 0) = (f (s, 0), g(s, 0)) = = ` · `0 (s)
(x0 , `0proj (2s + 0 − 1)) s ∈ [ 21 , 1]
and (
(x0 , `0proj (2s + 1 − 1)) s ∈ [0, 21 ]
F (s, 1) = (f (s, 1), g(s, 1)) = = `0 · `(s).
(`proj (2s − 1), y0 ) s ∈ [ 21 , 1]
F
Hence ` · `0 ' `0 · `, meaning loops in X × {y0 } and {x0 } × Y represent commuting elements of
π1 (X × Y, (x0 , y0 )). 
1 1
(12) Show that every homomorphism π1 (S ) → π1 (S ) can be realized as the induced homomor-
phism ϕ∗ of a map ϕ : S 1 → S 1 .
Proof. Without loss of generality, embed S 1 ,→ C in the complex plane. Let wn be a loop in
S 1 based at 1C defined by
wn : I → S 1 s.t. `(s) = e2πins .
In the proof of π1 (S 1 ) ∼
= Z (Theorem 1.7), the map Φ : Z → π1 (S 1 ) defined by Φ(n) =
[wn ] was shown to be an isomorphism. Now any homomorphism Ψ : Z → Z has the form
Ψ(z) = n · z with n ∈ Z. Then by the commutative diagram, observe that Φ(1Z ) = [w1 ] and
Φ ◦ Ψ(1Z ) = Φ(n) = [wn ], meaning any homomorphism f : π1 (S 1 ) → π1 (S 1 ) sends [w1 ] 7→ [wn ].
Furthermore, since 1Z is a generator of Z, by the group isomorphism [w1 ] is the generator of
π1 (S 1 ).
Z Z
1 7→ n · 1
Φ Φ
' '

π1 (S 1 ) π1 (S 1 ).
[w1 ] 7→ [wn ]
Consider the map ϕ : S 1 → S 1 such that ϕ(z) = z n . Then ϕ induces the map
ϕ∗ : π1 (S 1 ) → π1 (S 1 ) s.t. ϕ∗ ([w]) = [ϕw]
which is determined by its image on generator [w1 ]:
ϕ∗ ([ω1 ]) = [ϕ ◦ (s 7→ w1 (s))] = [s 7→ (e2πis )n ] = [wn ] = f ([ω1 ]).
Thus, f = ϕ∗ since these maps are identical on generator [w1 ] for f an arbitrary homomorphism
and ϕ(z) = z n , where n is the number of times f wraps a loop around S 1 . Hence, every
homomorphism π1 (S 1 ) → π1 (S 1 ) can be realized as the induced homomorphism ϕ∗ of a map
ϕ : S1 → S1.

(16) Show there are no retractions r : X → A in the following cases:
(a) X = R3 with A any subspace homeomorphic to S 1
(b) X = S 1 × D2 with A its boundary torus S 1 × S 1
(c) X = S 1 × D2 with A the circle shown in the figure
(d) X = D2 ∨ D2 with A its boundary S 1 ∨ S 1
(e) X a disk with two points on its boundary identified and A its boundary S 1 ∨ S 1
(f) X the Möbius band and A its boundary circle
Proof. For each case, we suppose for contradiction that X retracts onto subspace A. Then
by Proposition 1.17, the homomorphism ι∗ : π1 (A, x0 ) → π1 (X, x0 ) induced by the inclusion
ι : A ,→ X is injective. Hence, if we show that ι∗ : π1 (A, x0 ) → π1 (X, x0 ) cannot be injective,
we will have shown that there are no retractions r : X → A in that case.
4 LAURA LYMAN

(a) By Proposition 1.14, π1 (X) = π1 (R3 ) ∼

= 0. However, since A is homeomorphic to S 1 and
π1 (S ) ∼
1
= Z (Theorem 1.7), we have that π1 (A) ∼
= π1 (S 1 ) ∼
= Z. So the homomorphism
ι∗ : Z → 0
is injective. This is a contradiction since Z cannot inject homomorphically into 0.
(b) Since S 1 and D2 are path-connected, by Proposition 1.12, π1 (S 1 × D2 ) ∼
= π1 (S 1 ) × π1 (D2 ).
However, by the proof of Theorem 1.9 (Brouwer Fixed Point Theorem), D2 is contractible
i.e. π1 (D2 ) ∼
= 0. By Theorem 1.7, π1 (S 1 ) ∼ = Z. Thus π1 (S 1 ) × π1 (D2 ) ∼
= Z×0 ∼ = Z.
1 1 ∼
Similarly, π1 (A) = π1 (S × S ) = Z × Z. So the homomorphism
ι∗ : Z × Z → Z
is injective, which is impossible since Z×Z does not embed homomorphically into Z. Hence,
there are no retractions r : X → A in this case.
(c) Here A is the circle depicted in the figure below.

As in part (b), π1 (X) ∼

= Z. From the figure we see (since A crosses itself in the front) that
A is homeomorphic to S 1 . Though by tracing A, we see that the homotopy class of any
generating loop in π1 (A) is sent by ι∗ to a homotopy class of a loop in π1 (X) that travels
once around the S 1 factor and then (when reaching the crossing in A) travels again around
S 1 in the opposite direction — which is homotopic to a trivial loop. Hence, the inclusion
ι : A ,→ X induces a trivial map ι∗ : π1 (A) → π1 (X). Thus, ι∗ not injective, meaning there
are no retractions r : X → A in this case.
(d) For A = S 1 ∨ S 1 , by Section 1.2 and Van Kampen’s Theorem, π1 (A) = π1 (S 1 ∨ S 1 ) is
the free group on two generators F2 . However, since D2 is contractible, so is D2 ∨ D2 by
contracting to the identified point. Thus π1 (X) = π1 (D2 ∨ D2 ) ∼
= 0. So
∼0
ι∗ : F2 → π1 (X) =
is injective, which is a contradiction. Hence, there are no retractions r : X → A for
X = D2 ∨ D2 and A = S 1 ∨ S 1 .
(e) Let x0 , x1 be points on the boundary of D2 . Consider an arc on D2 from x0 to x1 , as drawn
in blue in the first figure below. Then D2 can retract onto the blue arc, as seen in the first
three figures, meaning D2 is homotopic to this arc. From there, X is formed by identifying
x0 ∼ x1 , as depicted in the last diagram. Hence, we see that π1 (X) ∼= π1 (S 1 ) ∼
= Z (unlike
∼
in part (d) where π1 (X) = 0).

x0 x0 x0

∼
x0

x1 x1 x1

Since A = S 1 ∨ S 1 , again by Van Kampen’s Theorem, we have that π1 (A) is the free group
on two generators (call them a and b). So we have the induced map
ι∗ : F2 ,→ Z.
Define ι∗ (a) := m and ι∗ (b) = n. Then note that an and bm are distinct elements in F2 .
However, by definition of a group homomorphism (and since Z is abelian),
ι∗ (an ) = ι∗ (a)n = mn = nm = ι∗ (b)m = ι∗ (bm )
 MATHEMATICS 210A: HOMEWORK 1 5

contradicting that ι∗ is injective.

(f) Let M be the Möbius band. Consider the gluing diagram of M depicted below. The sides
in red are glued, and the boundary circle A consists of the sides in gold.

From the figure, observe that M retracts onto the boundary circle A, meaning π1 (M ) ∼ =
π1 (A) ∼= π1 (S 1 ) ∼
= Z. So the induced map ι∗ : Z → Z is a homomorphism on Z and
therefore has the form ι∗ (z) = nz for some n ∈ Z. From the second image above, we can
see that a loop that goes once around the boundary circle A travels twice around the strip
M . Hence, ι∗ (z) = 2z for z ∈ Z.
Now, if a retraction r : M → A existed, since r∗ ι∗ = 1∗ , r∗ ι∗ (1) = 1. However, ι∗ (1) = 2,
meaning r∗ (2) = 1. Though, since r∗ is a homomorphism, r∗ (2) = r∗ (1+1) = r∗ (1)+r∗ (1) =
2r∗ (1), and so r∗ (1) = 21 . Hence r∗ (1) 6∈ Z, which is a contradiction. Thus the Mobius
band does not retract onto its boundary circle.

1
(17) Construct infinitely many non-homotopic retractions S ∨ S → S . 1 1

Proof. For S 1 ∨ S 1 , denote the circles as S11 and S21 with S21 as the rightmost circle. Without
loss of generality, embed S 1 ,→ C, and let the basepoint x0 ∈ X joining the circles be x0 = 0C ,
so S 1 ∨ S 1 is centered around the origin and the circles are centered around ±1. (This proof
will work for other values of x0 by adjusting the coordinate system). Heuristically, each non-
homotopic retraction will fix the rightmost S21 and twist the leftmost S11 around x0 some integer
number of times before projecting S11 onto S21 .
Formally, define the map fn : S 1 ∨ S 1 → S 1 by
(
z if |z − 1| = 1
fn (z) = n
1 + (z + 1) if |z + 1| = 1
for n ∈ Z. Then fn is clearly continuous and fn |S21 = 1S21 since fn fixes points satisfying
|z − 1| = 1, meaning fn fixes S21 . Hence, fn is a retraction from S 1 ∨ S 1 onto S 1 .
We now show that the {fn }n∈Z are non-homotopic. Let n1 , n2 ∈ Z with n1 6= n2 . Suppose
fn1 ' fn2 . Now {fn }n∈Z ∼ = Z ∼ = π1 (S 1 ) by mapping each fn 7→ n 7→ [wn ]. However, since
fn1 ' fn2 , this forces [wn1 ] = [wn2 ] (which is a contradiction). Hence, the {fn }n∈Z produce
infinitely many non-homotopic retractions.

Hatcher 1.2
(6) Use Proposition 1.26 to show that the complement of a closed discrete subspace of Rn is simply
connected if n ≥ 3.
Proof. Let S ⊆ Rn be a closed discrete subspace of Rn . Note that S is countable, since closed
discrete subspaces of separableWnormal subsets are countable. Now, Rn −S deformation retracts
onto a wedge sum of spheres α S n−1 for index set α with one (n − 1) W sphere per element of
S (so there are at most countably many summands). We can construct α S n−1 by attaching
n-cells (one per S n−1 ) to a common basepoint – and call this constructed space Y . So by
Proposition 1.26,
π1 (Rn − S) ∼= π1 (Y ) ∼ S n−1 ∼
_

= π1 = 0 for n ≥ 3
α
where the third isomorphism follows from Proposition 1.14 (which says that π1 (S n ) ∼
= 0 for
n ≥ 2) and Van Kampen’s Theorem. Furthermore, a topological space that is path-connected
6 LAURA LYMAN

and has trivial fundamental group is said to be simply connected. Since Rn − S is clearly
path-connected and π1 (Rn − S) = 0, we conclude that Rn − S is simply connected. 
1 1
(8) Compute the fundamental group of the space X obtained from two tori S × S by identifying
a circle S 1 × {x0 } in one torus with the corresponding circle S 1 × {x0 } in the other torus.
Proof. Let T1 , T2 ⊆ X be the two tori identified along a circle S 1 × {x0 } in one and the
corresponding S 1 × {x0 } in the other. Without loss of generality, assume this identification
stacks the two tori on top of each other as depicted below.

Now, we know π1 (T1 ) is the free group on two generators (call them a = [`a ] and b = [`b ]), and
similarly π1 (T2 ) = hc, di for generating loops `c , `d in T2 . Label the generators such that the
stacking depicted above identifies loops `a and `c .
Consider the neighborhoods N1 , N2 , where N1 is a neighborhood of `a contained entirely in
T1 and N2 is a neighborhood of `c contained entirely in T2 . Let U = T1 ∪ N2 and V = T2 ∪ N1 ,
so U, V, and U ∩ V are all path-connected.
Then U ∩V deformation retracts onto the circle `a (or equivalently `c ), and hence π1 (U ∩V ) ∼ =
π1 (S 1 ) ∼
= Z. Also, since U and V deformation retract onto T1 and T2 respectively (and
π1 (T1 ) = π1 (T2 ) ∼
= Z2 ), we have π1 (U ), π1 (V ) ∼
= Z2 . Let [`Γ ] be a generator of π1 (T1 ∩ T2 ).
Consider the maps induced by inclusion
ι1,∗ : π1 (T1 ∩ T2 ) → π1 (T1 )
ι2,∗ : π1 (T1 ∩ T2 ) → π1 (T2 )
Then ι1,∗ ([`Γ ]) = [`a ] = a and ι2,∗ ([`Γ ]) = [`c ] = c. By Van Kampen’s theorem (since X =
U ∪V ), we have that π1 (X) ∼ = (π1 (U )∗π1 (V ))/N where N is generated by ι1,∗ ([`Γ ])ι2,∗ ([`Γ ])−1 =
ac−1 . Hence,
π1 (X) ∼ = π1 (U ) ∗ π1 (V )/N
= ha, b, c, d : ab = ba, cd = dci/N [since π1 (U ), π1 (V ) are abelian]
∼
= ha, b, c, d : ab = ba, cd = dc, a = ci
∼
= ha, b, d : ab = ba, ad = dai,
completing the calculation. Note that ha, b, d : ab = ba, ad = dai ∼ = Z × F , where F is the free
group on two generators. Hence, π1 (X) ∼ = ha, b, d : ab = ba, ad = dai ∼ = Z × F. 
(9) In the surface Mg of genus g, let C be a circle that separates Mg into two compact subsurfaces
Mh0 and Mk0 obtained from the closed surfaces Mh and Mk by deleting an open disk from each.
Show that Mh0 does not retract onto its boundary circle C, and hence Mg does not retract onto
C. Also show that Mg does retract onto the non separating circle C 0 in the figure below.

Proof. Suppose for contradiction that there is a retraction r : Mh0 → C, so Mh0 retracts onto
its boundary circle C. Now the closed surface Mh (of genus h) has the following cell structure:
one 0 cell, 2h 1 cells, and one 2 cell, similar to the example in Chapter 0 (page 5) of Hatcher.
 MATHEMATICS 210A: HOMEWORK 1 7

The 1 cells correspond to (open arc) loops around every hole of Mh , each of which is then
identified at the same basepoint h0 (corresponding to the 0 cell). The 2 cell is an open 4h-gon,
which is depicted below.

b2
a2

b1

a1
b1
ah a1
bh

This 2 cell provides a relation on π1 (Mh ) by traveling along the perimeter loop; that is, π1 (Mh )
has the presentation
π1 (Mh ) = ha1 , b1 , . . . , ah , bh |a1 b1 a−1 −1 −1 −1
1 b1 · · · ah bh ah bh = 1i.

Let Dc be the open disk whose boundary is the circle C, so Mh0 = Mh \ Dc . Without loss of
generality, choose the representative loops in the 2h homotopy classes (corresponding to the 1
cells) such that they do not intersect Dc ; consequently, we can assume that Dc is contained
entirely in the interior of the 2 cell.
Then Mh0 has the same cell structure of Mh except that Mh0 has an open circle removed from
its 2 cell, meaning the perimeter of the 2 cell deformation retracts onto the boundary of this
circle. Hence, the homotopy class of a generating loop around C corresponds to an element
[a1 b1 a−1 −1 −1 −1 0 0
1 b1 · · · ah bh ah bh ] ∈ π1 (Mh ); that is, the induced map ι∗ : π1 (C) ,→ π1 (Mh ) sends the
∼
generator of π1 (C) = Z to the product of commutators [a1 , b1 ] · · · [ah , bh ] (similar to Example
2.36). However, this means that ι∗ : π1 (C) → abel(π1 (Mh0 )) is trivial since abel(π1 (Mh0 ))
quotients out this product of commutators. Consider the composition
ι∗ r
π1 (C) ,→ π1 (Mh0 ) →
∗
π1 (C)
which is the identity map (since r∗ ι∗ = 1∗ ). As the hint suggests, we abelianize the groups to
yield the composition
ι∗ r
π1 (C) ,→ abel(π1 (Mh0 )) → ∗
π1 (C)
∼
where abel(π1 (C)) = π1 (C) since π1 (C) = Z is already abelian. Therefore, the above compo-
ι∗
sition is the identity map with π1 (C) ,→ abel(π1 (Mh0 )) trivial, which is impossible for group
homomorphisms. Hence, Mh0 does not retract onto C. Since a retract r : Mg → C would yield
a retract of Mh0 to C by restriction, we conclude that Mg does not retract onto C.
We now prove that Mg does retract onto C 0 . Analogous to the diagram above, the 4g-gon
quotients to Mg with the attachment map whose gluing is indicated by the perimeter arrows.
The circle C 0 corresponds to one edge in the perimeter; without loss of generality, identify C 0
with the edge bg . Then we can first contract edges a1 and a−1 1 to a point, contract b1 and b1 ,
−1

and continue in this manner until only edge bg is left (corresponding to C 0 ). Formally speaking,
Mg has CW structure of one 0 cell, 2g W 1 cells (a1 , b1 , . . . , ag , bg ), and one 2 cell (the open 4g-
gon). Hence, the 1 skeleton of Mg is gk=1 (Sa1k ∨ Sb1k ) with 2 cell the product of commutators
[a1 , b1 ] · · · [ag , bg ]. Let q : Mg → M1 be the map that quotients by g−1 1 1
W
k=1 (Sak ∨ Sbk ), which is
clearly a retraction. Consider the map
r : S1 × S1 = ∼ M1 −→ {y} × S 1 = ∼ C 0 such that r(x1 , x2 ) = (y, x2 )

for some point y ∈ S 1 . Then r is continuous, and (letting (y, x2 ) ∈ {y} × S 1 ∼ = C 0 be arbitrary)
r|C 0 (y, x2 ) = (y, x2 ), meaning r|C 0 = idC 0 . So r is a retraction as well. Hence the map
r ◦ q : Mg → C 0
is a retraction. Thus, Mg does retract onto the non separating circle C 0 , as desired. 
 8 LAURA LYMAN

(14) Consider the quotient space of a cube I 3 obtained by identifying each square face with the
opposite square face via the right-handed screw motion consisting of a translation by one unit
in the direction perpendicular to the face combined with a one-quarter twist of the face about
its center point. Show this quotient space X is a cell complex with two 0 cells, four 1 cells,
three 2 cells, and one 3 cell. Using this structure, show that π1 (X) is the quaternion group
{±1, ±i ± j, ±k} of order eight.
Proof. The cube I 3 can be seen as a cell complex with eight 0 cells (vertices), twelve 1 cells
(edges), six 2 cells, and one 3 cell. The identifications described by the right-handed screw
motion leave the 3 cell unchanged and merge opposite pairs of 2 cells (faces). Hence, there is
one 3 cell and 6/2 = 3 total 2 cells in the quotient. The identifications of the 0 and 1 cells in
the quotient space are depicted below. Let i, j, k be group elements whose action is translation
by one unit in the x, y, or z directions. Assume the directions are such that +x is facing the
reader, +y is rightward, and +z is straight up.
e3
a b

e2 e4
e1
b a e2

e1

e4
e4 b a
e3 e3
e1
e2
a b
For example, we see that the horizontal edge e2 (in blue) in the frontmost face is sent, when
acted on by direction i, to an edge on the back of the cube pointing upward. When acted
upon by i again, the edge is sent to the e2 in the left vertical face. Finally, when i acts on this
resultant edge, e2 is sent back to its initial position. Similarly, by acting on 1 cells with j and k,
we get orbits of size 3. Since every 1 cell is contained in an orbit of size 3 (and there are twelve
one cells), the quotient will have 12/3 = 4 distinct 1 cells (e1 , e2 , e3 , e4 ). Furthermore, the
screw motion identifies the 0 cells such that there are only two distinct 0 cells in the quotient;
these cells have been labeled a and b. Thus, the quotient space X is a cell complex with two 0
cells, four 1 cells, three 2 cells, and one 3 cell.
We now analyze π1 (X). Since X is path-connected, the inclusion X 2 ,→ X induces an
isomorphism π1 (X 2 ) ∼ = π1 (X) by Proposition 1.26 (c). So we examine the 2-skeleton π1 (X 2 ).
In order to do so, we determine the 1-skeleton X 1 and then attach the three 2-cells in the
quotient X.
By the identifications drawn in the cube above, X 1 has the form depicted in the figure below.
Then by the map that collapses e1 , we see that X 1 is homeomorphic to the wedge of three
circles with generators e1 · e2 , e1 · e3 , and e1 · e4 . So π1 (X 1 ) ∼
= π1 (S 1 ∨ S 1 ∨ S 1 ) ∼
= Z∗Z∗Z
with generators i := [e1 · e2 ], j := [e1 · e3 ], k := [e1 · e4 ].
e1

e2
a b
e3
e4

Denote the three distinct 2 cells (faces) as f1 , f2 , f3 where f1 is the top horizontal face in the
diagram, f2 is the frontmost (vertical) face, and f3 is the rightmost vertical face. With the
righthand screw motion, f1 has the boundary e1 · e2 · e3 · e4 , as seen in the diagram below. Thus,
 MATHEMATICS 210A: HOMEWORK 1 9

f1 provides the following relation: idπ1 (X 2 ) = [e1 · e2 · e3 · e4 ]. Note that the shared basepoint
of each loop (according to the I 3 diagram above) is colored blue.
a b b a a b
e3 e1 e4
e2 f1 e4 e4 f2 e3 e3 f3 e2

b a a b b a
e1 e2 e1
Similarly, f2 and f3 provide the following relations (noting that each loop begins at the same
basepoint):
idπ1 (X 2 ) = [e3 · e2 · e4 · e1 ]
idπ1 (X 2 ) = [e4 · e2 · e1 · e3 ].
Expressing these relations in terms of the generators,
idπ1 (X 2 ) = [e1 · e2 · e3 · e4 ] = [e1 · e2 ][e3 · (e1 · e1 ) · e4 ] = [e1 · e2 ][e1 · e3 ]−1 [e1 · e4 ] = ij −1 k

idπ1 (X 2 ) = [e3 · e2 · e4 · e1 ] = [e3 · (e1 · e1 ) · e2 · e4 · e1 ] = [e1 · e3 ]−1 [e1 · e2 ][e1 · e4 ]−1 = j −1 ik −1

idπ1 (X 2 ) = [e4 · e2 · e1 · e3 ] = [e4 · (e1 · e1 ) · e2 · e1 · e3 ] = [e1 · e4 ]−1 [e1 · e2 ][e1 · e3 ] = k −1 ij

Hence, since X 2 is obtained from X 1 by attaching the 2 cells f1 , f2 , f3 , by Proposition 1.2.6
(a) we have that π1 (X 2 ) ∼
= π1 (X 1 )/hij −1 k, j −1 ik −1 , k −1 iji. So
π1 (X) ∼ = π1 (X 2 ) ∼= (Z ∗ Z ∗ Z)/hij −1 k, j −1 ik −1 , k −1 iji
∼
= (Z ∗ Z ∗ Z)/hki = j, ij = k, i = jki.
Now observe that
i2 = i(jk) = (ij)k = k 2 = k(ij) = (ki)j = j 2
and
(i2 )2 = i2 j 2 = (k −1 k)i2 j 2 = (k −1 k)i(ij)j = (k −1 k)i(k)j = k −1 (jk)j = k −1 ij = idπ1 (X 2 ) .
So by defining i2 = j 2 = k 2 := −1 and idπ1 (X 2 ) := 1, we can produce the following multiplica-
tion table:
1 i j k -1 -i -j -k
1 1 i j k -1 -i -j -k
i i -1 k -j -i 1 -k j
j j -k -1 i -j k 1 -i
k k j -i -1 -k -j i 1
-1 -1 -i -j -k 1 i j k
-i -i 1 -k j i -1 k -j
-j -j k 1 -i j -k -1 i
-k -k -j i 1 k j -i -1

For example, kj is computed by kj = k(1)j = k(k −1 ij)j = ij 2 = i(−1) = −i, and the other
calculations are similar. Since this multiplication table is identical to that of the quaternions
Q8 , we conclude that π1 (X) is the quaternion group {±1, ±i ± j, ±k} of order eight.