Matematiikan ja tilastotieteen laitos

Introdu tion to Algebrai Topology

Fall 2011
Exer ise 5
10.10-15.10.2011

1. Suppose (X, A) is a CW- omplex S and S (Xi , Ai ), i ∈ T

I is a T
olle tion of
sub omplexes of X . Prove that ( i∈I Xi , i∈I Ai) and ( i∈I Xi , i∈I Ai ) are
both sub omplexes of X .

Solution: ( The proof of) Lemma 4.1.4 implies that (Y, B) is a sub omplex
of (X, A) if and only if
1) B = {α ∈ A | eα ∩ Y 6= ∅},
and
2) eα ∩ Y 6= ∅ implies that eα ⊂ Y .

Now [
eα ∩ ( Xi ) 6= ∅ ⇔ eα ∩ Xi 6= ∅ for some i ∈ I ⇔
i∈I
[
⇔ α ∈ Ai for some i ∈ I ⇔ α ∈ Ai ,
i∈I
S
and eα ∩ ( i∈I Xi ) 6= ∅ implies that eα ∩ Xi 6= ∅ for some i ∈ I whi h implies

that eα ⊂ Xi for some i ∈ I . In parti ular

[
eα ⊂ Xi .
i∈I
S S
Hen e ( i∈I Xi , i∈I Ai ) is a sub omplex.
T
Let us he k the same for the interse tion. Suppose eα ∩ i∈I Xi 6= ∅. Then
in parti ular eα ∩ Xi 6= ∅ for all i ∈ I , so
eα ⊂ Xi
T
for all i ∈ I . Hen e eα ⊂ Ti∈I Xi . In parti ular ondition
T 2) above is satised
and if α ∈ {α ∈ A | eα ∩T i∈I Xi 6= ∅}, then α ∈ i∈I Ai .
Conversely suppose α ∈ i∈I Ai . Then α ∈ Ai for all i ∈ I , so
eα ⊂ Xi
for all i ∈ I , sin e (Xi , Ai ) is a sub omplex for all i ∈ I . Hen e
[
eα ⊂ Xi .
i∈I
S T
In parti ular eα ∩ i∈I Xi 6= ∅, so α ∈ {α ∈ A | eα ∩ i∈I Xi 6= ∅}.

2. a) Suppose X is a CW- omplex and A is a path- omponent of X . Prove that

A is a sub omplex of X .
2

b) Suppose X is a CW- omplex. Prove that the following laims are equi-
valent:
1) X is onne ted.
2) X is path- onne ted.
3) X 1 is path- onne ted.
4) Every two verti es in X 0 an be joined by a path that lies in X 1 .

Solution: a) Suppose eα ∩ A 6= ∅ and let x ∈ eα ∩ A. Now eα is a path-

onne ted subset of X (sin e it is a ontinuous image of the path onne -
n
ted spa e B for some n ∈ N, whi h ontains x. By the denition of path-
omponent, eα ⊂ A.
By Lemma 4.1.4 A is a sub omplex of A.

b) 1) ⇔ 2):
All path- onne ted spa e are onne ted, so 2) implies 1) trivially.
Suppose X is a onne ted CW- omplex and let A be the set of all path-
onne ted omponents of X . If A is empty, X is empty, so it is trivially
path- onne ted. Otherwise x a path- omponent A ∈ A and dene
[
B= B.
B∈A,B6=A

Then A ∪ B = X , A ∩ B = ∅. A is a sub omplex by a), so it is in parti ular

losed in X . Also every B ∈ B, B 6= A is a sub omplex for the same reason.
By the exer ise 1 B is a sub omplex, hen e also B is losed in X . Hen e
B must be empty, sin e otherwise A|B would be a separation of onne ted
spa e X . Hen e there is only one path- omponent A, whi h means that X is
path- onne ted.

2)⇔ 3):
By Lemma 4.2.1 ) the in lusion i : X 1 → X indu es an isomorphism i∗ : H0 (X 1 ) →
H0 (X), hen e in parti ular

H0 (X 1 ) ∼
= H0 (X)

Sin e a spa e Y is path- onne ted if and only if H0 (Y ) ∼ = Z, it follows that

X is path- onne ted if and only if X 1 is path- onne ted.
3)⇔ 4):
Every point x of X 1 an be joined by the path to a vertex a ∈ X 0 , sin e
1
x ∈ eα for some fα : B , where eα is path- onne ted and eα interse ts X 0 (in
subset fα (S 0 )).
Hen e X 1 is path- onne ted if and only if all verti es (i.e. points of X 0 ) be-
long to the same path- omponent of X 1 .

3. Suppose K is a simpli ial omplex and a, b are verti es of K . An edge-path

from a to b is a nite sequen e of verti es a = a0 , . . . , an = b of K su h that
for all i = 0, . . . , n ai and ai+1 belong to the same 1-simplex τi . In this ase
also the sequen e τ0 , . . . , τn−1 is also alled an edge-path from a to b.
Prove that |K| is onne ted if and only if for every pair of verti es a, b ∈ K
3

there is an edge-path from a to b.

Solution: Suppose |K| is onne ted and let a be a vertex of a. Dene

sub omplex L, N of K as following. A simplex σ ∈ K belongs to L if and
only if all verti es of σ an be joined to a via an edge-path. A simplex σ ∈ K
belongs to N if and only if none of the verti es of σ an be joined to a via an
edge-path.

Suppose σ ∈ K is a simplex and σ ∈ / N . Then one vertex v of σ an be

joined to a via edge-path a = a0 , a1 , . . . , an = v . Let v ′ be any other vertex
of σ . Then 1-simplex with verti es v and v ′ is a fa e of σ , so belongs to K .
Hen e a = a0 , a1 , . . . , an = v, an+1 = v ′ is an edge-path from a to v ′ . Hen e
σ ∈ L, so K = L ∪ N .
It follows that |K| = |L| ∪ |N|. Clearly |L| ∩ |N| = ∅. Sin e L, N are
sub omplex of K , |L| and |N| are losed in |K|. Sin e |L| is non-empty
(a ∈ |L|), and |K| is onne ted, N must be empty, sin e otherwise |L|||N| is
a separation of |K|.
In parti ular all verti es are in L, so every vertex an be joined to a by an
edge-path.

Conversely suppose for every pair a, b of verti es there is an edge-path from

a to b. Sin e edge-path learly denes a ontinuous path from a to b in |K|, all
verti es belong to the same path- omponent of |K|. Sin e every point x ∈ |K|
belong to some simplex σ ∈ K , whi h is path- onne ted and ontains at least
one vertex, every point belongs to the path- omponent of some vertex. These
observations now easily imply that |K| is path- onne ted, in parti ular on-
ne ted.

4. Suppose g ∈ N (g ≥ 1). Show that Mg (Ng ) is a onne ted ompa t 2-

manifold without boundary, whi h an be triangulated.

Solution: Clearly all these surfa es are ompa t and onne ted spa es, sin e
they are quotient spa es of ompa t and onne ted polygon.

M0 = S 2 and N1 = RP 2 are known tobe 2-manifolds without boundary, so

may assume g ≥ 1 (g ≥ 2).

Let X be 4g (2g )-polygon and p : X → Mg (Ng ) be a anoni al quotient pro-

je tion. Then the restri tion of p to the interior of X is a homeomorphism to
its image (for instan e be ause by the exer ise 11.6 it is a quotient mapping,
whi h is also inje tive), whi h is also open in Mg (Ng ) so the points in the
image of interior have neighbourhoods homeomorphi to B2 .
Suppose x is an interior point of an edge τ in X . Then p−1 (p(x)) = {x, x′ },
where x′ is an interior point of another edge, identied with τ . Both have
small enough neighbourhoods in X , whi h do not interse t other edges and
are homeomorphi to {(x, y) ∈ B 2 | y ≥ 0}, where the homeomorphism maps
{(x, 0) ∈ B 2 } and only points in that set to the edge. Clearly both homeo-
morphisms an be hosen so that they an be tted together as an embedding
4

B 2 → Mg (Ng ) with image being an open neighbouhood of Mg (Ng ). Hen e

p(x) has a neighbourhood homeomorphi to B 2 .

We are left with the point that orrespond to all verti es of the polygon.
For every vertex we hoose a small enough neighbourhood of a vertex, that do
not ontain any pairs of identied points, i.e. p is inje tion restri ted to that
neighbourhood. Then we sti k them together as the pi ture below indi ated
for Mg .
x2
x1 x1 x0
D
x0
C B A x1
C
x0
B
E E b
D
yn
A x1 yn
x0 x2
The ase of a vertex for Ng is done the same way.

It remains to prove that Mg and Ng are trianguable. It is enough to trian-

gulate them using some ∆- omplex (). Choose a vertex a0 in the interior
of a polygon. Let a1 , . . . , am be verti es of the polygon. Then 2-simpli es
[a0 , ai , ai+1 ], where an+1 = a1 and their fa es form a triangulation of a poly-
gon with ordering of simpli es indi ated by their indi es. This triangulation
denes a ∆- omplex triangulation for any spa e obtained from a polygon by
identifying edges in an obvious way.
5. Suppose K is a 2-dimensional simpli ial omplex and τ ∈ K is a 1-simplex
whi h is a fa e of exa tly n 2-simpli es. Suppose x is an interior point of τ .
Prove that
H1 (|Lk(x)|) ∼
= Zn−1 .

Solution: Lk(x) onsists of all 1-fa es (and their verti es) of simpli es σi ,
that are not τ , so |Lk(x)| is homeomorphi to the spa e obtained from n dis-
joint opies I1 , I2 , . . . , Im of the unit interval I = [0, 1] by identifying all points
0 ∈ Ii , i = 1, . . . , m to a single point and also all points 1 ∈ Ii , i = 1, . . . , m
to a single point - see the pi ture below.
a1
b

b x

a0
This spa e an be triangulated as a polyhedron |K| of a ∆- omplex K ,
that onsists of n 1-simpli es τ1 , . . . , τn , with their orresponding verti es a, b
5

identied. We order every edge as a simplex [a, b], then in C1 (K)

∂τi = b − a
for all i = 1, . . . , n. Hen e
∂(m1 τ1 + m2 τ2 + . . . + mn τn ) = (m1 + m2 + . . . + mn )(b − a) = 0
if and only if mn = −m1 − m2 − . . . mn−1 . It follows that
{τ1 − τn , τ2 − τn , . . . , τn−1 − τn }
is a basis of the free abelian group Ker ∂1 . Sin e there are no 2-simpli es, for
the simpli ial homology we have
H1 (K) = Ker ∂1 ∼
= Zn−1 .
The laim follows sin e the simpli ial homology is isomorphi to the singular
homology.

6. Suppose K is a nite simpli ial omplex su h that |K| is an n-dimensional

manifold, possibly with boundary. Prove that |Lk(x)| has the homotopy type
of S n−1 , if x ∈ |K| is an interior point and ontra tible if x is the boundary
point.
Assuming n = 2 prove that |K| is 2-dimensional as simpli ial omplex and
every 1-simplex of K is a fa e of two or one 2-simplex. Moreover if L is a
sub omplex of K generated by 1-simpli es that are fa es of exa tly one 2-
simplex, then |L| = ∂|K|.
n
Solution: For every x ∈ |K| exists a homeomorphism f : B → U , U on-
tains an open neighbourhood of x in |K| and f (y) = x if x is an interior point
for y = 0, f (y) = x for some y ∈ S n−1 if x is a boundary point. Triangulate
n
B as |K(σ)|, where σ is an n-simplex. If x is a boundary point, we may as-
sume that y above is a vertex point of σ .
Now Proposition 4.3.5 implies that |Lk(y)| and |Lk(x)| have the same homo-
topy type. If x is an interior point |Lk(y)| is homeomorphi to S n−1 . If x is
a boundary point, Lk(y) is a sub omplex generated by n − 1-fa e of σ oppo-
site to y , so |Lk(y)| is ontra tible. This on ludes the proof of the rst laim.

Suppose n = 2. If K would have a maximal simplex σ of dimension m > 2,

then the interior point of this simplex would have link homeomorhi to S m−1 .
However S m−1 does not have the same homotopy type of S n−1 neither it
is ontra tible. This is a ontradi tion with what we already proved. Hen e
dim K ≤ 2. Sin e K is not empty, there is at least one vertex a , and there has
to be at least one 1-simplex that ontains this vertex, sin e otherwise a would
be an open dis rete point in |K|, whi h is impossible for 2-manifold. Hen e it
is enough to prove that every 1-simplex of K is a fa e of two or one 2-simplex,
sin e that would also imply that there is at least one 2-simplex, so dim K = 2.

Let τ be 1-simplex and x be an interior point of τ . Suppose τ is a fa e

of exa tly m 2-simpli es. Then previous exer ise shows that H1 (|Lk(x)|) ∼ =
Z m−1 ∼
. On the other hand we already know that H1 (|Lk(x)| = Z, if x is an in-
terior point and H1 (|Lk(x)) = 0, if x is a boundary point. Hen e m−1 = 0, 1,
6

i.e. m = 1 or m = 2.

Suppose L is a sub omplex of K whi h onsists of all 1-simpli es of K ,

that are fa es of exa tly one 2-simplex and all their verti es. We laim that
|L| = ∂|K|.
Suppose τ is a 1-simplex that is a fa e of exa tly one 2-simplex σ ∈ K and let
x be an interior point of τ . Now x has learly has a neighbourhood whi h is
ontained entirely in σ and does not interse t other fa es of σ . Moreover we
an hoose this neighbourhood to be homeomorphi to {(x, y) ∈ B 2 | y ≥ 2}.
Hen e x ∈ ∂|K|.
If, one the other hand x is a vertex in L, its arbitrary neighbourhood inter-
se ts interior of some 1-simplex of L, whi h we already proved to be ontained
in ∂|K|. Hen e x ∈ ∂|K|. But ∂|K| is losed, so x ∈ ∂|K|. We have shown
that |L| ⊂ ∂|K|.

To prove the opposite that x be a boundary point of |K|. There are 3-

possibilities - x is a vertex, x is an interior point of 1-simplex or x is an
interior point of 2-simplex. In the latter ase x learly has small neighbour-
hood homeomorphi to B 2 , so x is not a boundary point. Suppose x is an
interior point of 1-simplex τ . If τ is a fa e of 2 simpli es, their union is ho-
meomorphi to the square, with x being an interior point of square. Hen e
as above we see that x is an interior point of the manifold, ontradi ition.
Hen e τ fa es only one triangle, so is an element of L. In this ase x ∈ |L|.
We are left with the ase x is a vertex. Let
[
St(x) = {int σ | x ∈ σ}.
Then St(x) is an open neighbourhood of x (Lemma 1.2.7), that does not
ontain any other vertex of K , ex ept x itself. Suppose all simpli es of K
that ontain x are not in L. Then all points of St(x) ex ept x are interior
points of 2-simpli es or 1-simpli es that fa e two 2-simpli es. But we alrea-
dy showed above that su h points are not in the boundary of |K|. Hen e
St(x) ∩ ∂|K| = {x}, so x is a dis rete point of ∂|K|, sin e St(x) was open.
But this is not possible, sin e ∂|K| is a 1-manifold (exer ise 10.6). This ont-
radi tion shows that there is at least one 1-simplex in L that ontains x, so
x ∈ |L|.

Bonus points for the exer ises: 25% - 1 point, 40% - 2 points, 50% - 3 points,
60% - 4 points, 75% - 5 points.