[DRAFT] A Peripatetic Course in Algebraic Topology

Julian Salazar
jsalazar01@[Link] • [Link]

July 22, 2016

Abstract
These notes are based on lectures in algebraic topology taught by Peter May and
Henry Chan at the 2016 University of Chicago Math REU. They are loosely chrono-
logical, having been reorganized for my benefit and significantly annotated by my
personal exposition, plus solutions to in-class/HW exercises, plus content from read-
ings (from May’s Finite Book), books (e.g. May’s Concise Course, Munkres’ Elements of
Algebraic Topology, and Hatcher’s Algebraic Topology), Wikipedia, etc.

I Foundations + Weeks 1 to 3 3

1 Topological notions 3
1.1 Topological spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Separation properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Continuity and operations on spaces . . . . . . . . . . . . . . . . . . . . . . 4

2 Algebraic notions 5
2.1 Rings and modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Tensor products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Categorical notions 11
3.1 Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 Natural transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 [DRAFT] Universal properties . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.5 Adjoint functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1
4 The fundamental group 21
4.1 Connectedness and paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Homotopy and homotopy equivalence . . . . . . . . . . . . . . . . . . . . . 22
4.3 The fundamental group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.5 [TODO] Van Kampen’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . 28

5 Higher homotopy groups 28

5.1 Mapping spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 Operations on pointed spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3 Higher homotopy groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

6 Simplicial complexes 34
6.1 Simplicial complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2 Abstract simplicial complexes . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.3 Ordered simplicial complexes . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.4 Simplicial approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

7 Chain complexes and simplicial homology 39

7.1 Simplicial chain complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.2 Simplicial homology groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.3 Maps of chain complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.4 Tensor products of chain complexes . . . . . . . . . . . . . . . . . . . . . . . 43

8 [TODO] Simplicial objects 44

8.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

9 CW complexes 44
9.1 Motivation and definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
9.2 Operations on CW complexes . . . . . . . . . . . . . . . . . . . . . . . . . . 46
9.3 [TODO] Cellular approximation, etc. . . . . . . . . . . . . . . . . . . . . . . 47

II Week 4 onwards 47

10 Topics covered 47
10.1 July 11 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
10.2 July 12 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
10.3 July 13 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
10.4 July 14 (Chan) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
10.5 July 15 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
10.6 July 18 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

2
 1 TOPOLOGICAL NOTIONS

10.7 July 19 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

10.8 July 20 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
10.9 July 21 (May) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Part I
Foundations + Weeks 1 to 3
1 Topological notions
Remark. This is a review of material from point-set topology. See Munkres, or the readings
[1.1-1.5] (from which some of the definitions are directly taken).

1.1 Topological spaces

Definition 1.1. A topological space (X, U ) is a set X, along with a set of subsets U of X
we call a topology on X. One requires that ∅, X ∈ U , and that U is closed under finite
intersection and arbitrary union. If the topology is apparent, one simply uses X to refer
to the space itself.
Definition 1.2. The elements of U are called the open sets of X (in the topology U ). The
complements of open sets in X are called closed sets. A neighborhood of a point x ∈ X is
an open set U such that x ∈ U .
Definition 1.3. The closure of a subset A ⊆ X, denoted A, is the intersection of all closed
sets containing A. Equivalently, A is the union of A and its limit points (points x for
which every neighborhood intersects A at some a , x).
Definition 1.4. A basis for a topology on X is a set B of subsets such that every x ∈ X is
in some B ∈ B, and if x ∈ B0 ∩ B00 then there exists B ∈ B such that x ∈ B ⊆ B0 ∩ B00 . (Hence,
every finite intersection in B is the union of these extant B ∈ B as well.) In this way, the
unions of the sets in B define the topology U generated by basis B on X. Equivalently,
we say that B is a basis for the topology U .
Definition 1.5. A subbasis for a topology on X is a cover S of X. By including the finite
intersections of S, we get a basis B on X. This basis generates a topology U on X, and we
say that S is a subbasis for the topology U .
Remark. Observe that a subbasis S of a topology U is a weaker version of a basis B. When
a subbasis is available, however, it is a efficient description of a topology (based on the
number of sets we have to describe; by the measure of simplicity of operations, a basis is
more “efficient” because we only have to consider unions). In both the basis and subbasis
cases, we can think of the topology U as the smallest topology containing S or B.

3
1.2 Separation properties 1 TOPOLOGICAL NOTIONS

1.2 Separation properties

Definition 1.6. The separation properties are the following hierarchy of types of a topo-
logical space:
• A T0 -space is one where “points are distinguished”, i.e., for every two points, one of
the points has an open neighborhood not containing the other.

• A T1 -space is one where each point is a closed subset.

• A T2 -space or Hausdorff space if any two points have disjoint open neighborhoods.
Proposition 1.7. T2 implies T1 implies T0 .
Proof. The first implication follows from fixing your desired point x, and taking the union
of all the open neighborhoods disjoint from it. By the Hausdorff property, that union
includes all points except x. The second implication follows from noting the complement
of a point is the desired open neighborhood distinguishing it from the other point.

Proposition 1.8. Metric spaces are Hausdorff.

Proof. Two distinct points x, y have a non-zero distance , which means they are separated
by the disjoint basis elements B(x; 2 ), B(y; 2 ).

1.3 Continuity and operations on spaces

Definition 1.9. A continuous map of topological spaces (X, U ) and (Y , V ) is a function
f : X → Y such that f −1 (V ) ∈ U for all V ∈ V . That is, the preimages of open sets are open.
Remark. It suffices to check that f −1 (V ) is open for each V in a basis or a subbasis of V , as
preimages preserve union and intersection.
Definition 1.10. X and Y are homeomorphic if there exists a continuous map f : X → Y
(called a homeomorphism) with a continuous inverse.
Remark. Homeomorphisms play the role of isomorphisms for topological spaces. Two
spaces that are isomorphic (e.g., (a, b) and R) have the same topological properties. Intu-
itively, a homeomorphism is a bijection on the underlying set, and a bijection on the open
sets that preserves the relations between the open sets in the topology.
Definition 1.11. The subspace topology on A ⊆ X is given by {A∩U | U open in X}. More
abstractly, let i : A → X be an injection. Then the subset topology is given by the set of
i −1 (U ) for open U .
Example 1.12. Consider [0, 1] ⊆ R where R has the usual Euclidean topology. The open
sets of [0, 1] with the subspace topology are the usual open sets, except truncated to [0, . . . )
and (. . . , 1] when they cross these boundaries.

4
 2 ALGEBRAIC NOTIONS

Definition 1.13. Let ∼ be an equivalence relation on (X, U ). The quotient topology is

given on the set of equivalence classes X/ ∼ where the open sets are sets of equivalence
classes whose unions were open in X. That is, the topology U 0 on X/ ∼ is
 

 [ 

0
 
U = U ⊆ X/ ∼ | [a] ∈ U .
 
 

 
 [a]∈U 

Equivalently, let q : X → Y be a surjection. Then we write Y = X/ ∼ and define its open

sets as U ⊆ Y such that q−1 (U ) is open in X.
Example 1.14. The quotient topology enables the general method of gluing, whereby
points in a space are identified together to give a new space. Consider [0, 1] ⊆ R with a
subspace topology, and take the equivalence relation given by the partition:

∼ = {{0, 1}} t {{x} | x ∈ (0, 1)}

One can envision this as gluing the ends of the interval together, and so we expect to
get the space of a circle (with, e.g., its subspace topology in R2 ). One can check that
f (x) = e2iπx is well-defined on [0, 1]/ ∼ and a homeomorphism to S 1 .
Remark. In a sense made precise later, the definitions for subspace and quotient topologies
are dual to each other (they are with respect to an injection and a surjection, respectively).
Definition 1.15. The disjoint union topology on X tY takes as a basis the disjoint unions
of an open set of X and an open set of Y (though they are in fact all the open sets).
Example 1.16. For the disjoint union topology, think about R1 t R1 . The two spaces are
essentially independent (though your set has to be open in both to be called open in the
union topology).
Definition 1.17. The product topology on X ×Y takes as a basis the products U ×V where
Q
U is open in X and V is open in Y . In the case of infinite products Xi , we only allow
basis elements whose components are Xi in all but finitely many places.
Example 1.18. For the product topology, think about R2 ’s metric space topology, which is
equivalent to the product topology. The open sets are not just the (boundaryless) unions
of rectangles that are products of R1 ’s open sets; these merely serve as a basis. The open
sets are arbitrary unions of (arbitrarily small) rectangles; cf., the open disc.

2 Algebraic notions
Remark. This section should be review, hence the terseness. Based on class, Dummit and
Foote’s Abstract Algebra, and [Link]
[Link].

5
2.1 Rings and modules 2 ALGEBRAIC NOTIONS

2.1 Rings and modules

Definition 2.1. A (commutative, unital) ring R is a set equipped with two binary oper-
ations + and · such that R is an abelian group under addition, an abelian monoid (think
group without the inverse requirement) under multiplication, and that distributivity
holds:

a · (b + c) = (a · b) + (a · c).

Example 2.2. We have the rings Z, Q, the polynomial rings R[x] for any ring R, the contin-
uous functions f : R → R, etc., where + and · correspond to addition and multiplication
in the usual way.

Definition 2.3. An R-module M is an abelian group M with an “R-action” (i.e., an action

with respect to both R’s additive group and R’s multiplicative monoid)

· : R × M → M.

Explicitly, we must have for all x, y ∈ M and r, s ∈ R:

r · (x + y) = r · x + r · y
(r + s) · x = r · x + s · x
(rs) · x = r · (s · x)
1R · x = x,

where 1R denotes R’s multiplicative identity.

Definition 2.4. The group of R-module homomorphisms HomR (M, N ) (also known as
R-linear maps) is given by the set of maps f : M → N satisfying for all x, y ∈ M and r ∈ R:

f (x + y) = f (x) + f (y)
f (rx) = rf (x)

These maps “preserve” the structure of the module. More formally, we are demanding
that f is a group homomorphism that is equivariant (with respect to R’s multiplicative
action). One can verify that HomR (M, N ) is a group by pointwise addition.

Example 2.5. For R = k a field, the concept of R-module coincides exactly with being a
k-vector space.

Example 2.6. For R = Z, the concept of R-module coincides exactly with being an abelian
group (the action is given by repeated addition: ng = g + · · · + g).

Example 2.7. Polynomial rings R[x] are R-modules.

6
2.2 Tensor products 2 ALGEBRAIC NOTIONS

Definition 2.8. A submodule N of an R-module M is an abelian subgroup of M that is

also closed under multiplication by R.
Example 2.9. For N ⊆ M = R, then N is a submodule exactly when it is an ideal of R.
Definition 2.10. The quotient module M/N given by an R-module M and a submodule
N is the abelian subgroup M/N with the induced R-action r(m + N ) = rm + N . We verify
this is well-defined: if m + N = m0 + N , then m − m0 ∈ N . Then r(m − m0 ) = rm − rm0 ∈ N
since N is a submodule. Then rm + N = rm0 + N as desired.

2.2 Tensor products

The motivation for the tensor product M ⊗R N of modules M, N is that it is the
“minimal” setting from which bilinear maps M × N → P can be studied as linear maps
(i.e., R-module homomorphisms) M ⊗R N → P . For example, if there is a natural notion
of multiplication between M and N , that is a bilinear map and is thus ideally studied in
M ⊗R N .
Definition 2.11. Let f : M × N → P be an R-bilinear module homomorphism, i.e.,

f (m + m0 , n) = f (m, n) + f (m0 , n)
f (rm, n) = rf (m, n)
f (m, n + n ) = f (m, n) + f (m, n0 )
0

f (m, rn) = rf (m, n)

hold. The tensor product M ⊗R N is uniquely defined (up to unique isomorphism) as an

R-module equipped with an R-bilinear map

⊗ : M × N → M ⊗R N

such that any such R-bilinear map f , there exists a unique R-linear (i.e., an R-module
homomorphism) map f˜ such that

M ×N
⊗ / M ⊗R N

f˜
% 
f
P

commutes; that is, f = f˜ ◦ ⊗.

Remark. By uniquely defined up to unique isomorphism, we mean that for two tensor
products T , T 0 and associated ⊗, ⊗0 , one can prove there is a unique isomorphism i : T  T 0
which satisfies ⊗0 = i ◦ ⊗. This is why we can safely write M ⊗R N . This characteristic is
typical of universal properties, of this definition is an example.

7
2.2 Tensor products 2 ALGEBRAIC NOTIONS

However, this does not guarantee that a tensor product actually exists; it just states
that if one did exist, it would be unique in this manner. The following rectifies that:

Proposition 2.12. Given M × N , there exists a tensor product M ⊗R N .

M ⊗R N = F(M × N )/L

where F(M × N ) denotes the free module over the set M × N , and L is the submodule generated
by

(m + m0 , n) − ((m, n) + (m0 , n))

(m, n + n0 ) − ((m, n) + (m, n0 ))
c(m, n) − (cm, n)
c(m, n) − (m, cn)

(implicitly working in the image of ι : M × N → F(M × N )).

Proof. See Garrett’s notes (the universal property of a free module also comes into play).
The intuitive idea is that the submodule L encodes the exact relations that enforce bilin-
earity on the “obvious” inclusion ι.

Example 2.13. Let V , W be vector spaces over F. They have bases {vi }, {wj }. The map ⊗ is
given by the bilinear map on the basis (vi , wj ) → vi ⊗ wj , then extended by linearity. Hence
the induced f˜ (see the universal property) must satisfy

f˜(vi ⊗ wj ) = f (vi , wj ).

Example 2.14. By the tensor product relations, we have in M ⊗ N :

m ⊗ 0 = m ⊗ (0 + 0) = (m ⊗ 0) + (m ⊗ 0),

which implies m ⊗ 0 = 0 for all m ∈ M. The same reasoning gives 0 ⊗ n = 0 for all n ∈ N .

Example 2.15. Note that

Z/(2) ⊗Z Z/(3)  0.

To see this, note that 3a = a for a ∈ Z/(2) but 3b = 0 for b ∈ Z/(3). 3(a, b) = (a, b) = (a, 0).
Meanwhile, 4(a, b) = (a, b) = (0, b). Hence (a, b) = (a, 0) = (0, b) = (0, 0).

Exercise 2.16. Let I, J be ideals of R. Show that

R/I ⊗R R/J  R/(I + J).

8
2.2 Tensor products 2 ALGEBRAIC NOTIONS

Proof. Observe that

φ : R/I × R/J → R/(I + J), φ(a + I, b + J) = (a + I)(b + J) = ab + (I + J)

is a bilinear map. For example,

φ(a + a0 + I, b + J) = (a + a0 + I)(b + J) = (a + a0 )b + (I + J)
= φ(a + I, b + J) + φ(a0 + I, b + J).

Hence, it factors through a linear map

φ̃ : R/I ⊗R R/J → R/(I + J)

of R-modules. Certainly φ̃ is surjective, since we now know the following is well-defined:

φ̃(m + I ⊗ 1 + J) = (m + I)(1 + J) = m + (I + J).

To show injectivity, we first note that every element in R/I ⊗R R/J can be a written as a
simple tensor of a particular form. Namely, we can rewrite a general tensor as:
 
X X X 
(mi + I ⊗ ni + J) = mi ni (1 + I ⊗ 1 + J) =  mi ni  (1 + I ⊗ 1 + J)
i
i   i
X  
=  mi ni  + I ⊗ 1 + J  .
i

Then every element of R/I ⊗R R/J is of the form m + I ⊗ 1 + J. Then if

0 + (I + J) = φ̃(m + I ⊗ 1 + J) = m + (I + J)

implies m ∈ I + J. Hence we can write m = a + b where a ∈ I and b ∈ J and get

m + I ⊗ 1 + J = a + b + I ⊗ 1 + J = b + I ⊗ 1 + J = b(1 + I ⊗ 1 + J) = 1 + I ⊗ b + J = 1 + I ⊗ J
= 0,

which shows the kernel is trivial.

Example 2.17. We have

Z/(m) ⊗Z Z/(n)  Z/((m) + (n)) = Z/(gcd(m, n)).

We also have

k[x]/(f ) ⊗k[x] k[x]/(g)  k[x]/((f ) + (g)),

and if k a field then the right side is equal to k[x]/(gcd(f , g)).

9
2.2 Tensor products 2 ALGEBRAIC NOTIONS

Exercise 2.18. Verify the following properties: that R is unital with respect to ⊗R , that ⊗R is
associative, and that ⊗R is commutative. That is:

R ⊗R M  M  M ⊗R R
M ⊗R (N ⊗R P )  (M ⊗R N ) ⊗R P
M ⊗R N  N ⊗R M.

Proof. Let r, r 0 ∈ R, m, m0 ∈ M, n, n0 ∈ N , p, p0 ∈ P . Then:

• Observe that φ : (r, m) 7→ rm is a bilinear map:

φ(r + r 0 , m) = (r + r 0 )m = rm + r 0 m = φ(r, m) + φ(r 0 , m)

φ(r, m + m0 ) = r(m + m0 ) = rm + rm0 = φ(r, m) + φ(r, m0 ).

It thus factors through a map φ̃ : R ⊗R M → M. We define the map ψ̃(m) = 1 ⊗R m. One

checks these are inverses; e.g.,

ψ̃(φ̃(r ⊗R m)) = ψ̃(rm) = 1 ⊗R rm = r ⊗R m.

• The following strategy is due to this comment. Observe that φp : (m, n) 7→ (m, n ⊗R p) is
a bilinear map for fixed p:

φ(m + m0 , n) = (m + m0 , n ⊗R p) = (m, n ⊗R p) + (m0 , n ⊗R p) = φ(m, n) + φ(m0 , n)

φ(m, n + n0 ) = (m, (n + n0 ) ⊗R p) = (m, n ⊗R p) + (m, n0 ⊗R p) = φ(m, n) + φ(m, n0 )

It thus factors through φ̃p : M ⊗R N → M ⊗R (N ⊗R P ). Let a, a0 ∈ M ⊗R N . Then

ψ : (a, p) 7→ φ̃p (a) is a bilinear map:

ψ(a + a0 , p) = φ̃p (a + a0 ) = φ̃p (a) + φ̃p (a0 ) = ψ(a, p) + ψ(a0 , p)

ψ(a, p + p0 ) = φ̃p+p0 (a) = φ̃p (a) + φ̃p0 (a) = ψ(a, p) + ψ(a, p0 )

where the equality φ̃p+p0 = φ̃p + φ̃p0 can be verified by writing the general sum a =
P
i mi ⊗ ni and using the linearity of φ̃p on M ⊗R N . Hence, ψ factors through

ψ̃ : (M ⊗R N ) ⊗R P → M ⊗R (N ⊗R P ).

One can construct an R-module homomorphism in the other direction in a similar

manner, and verify that they are inverses.

• Observe that φ : (m, n) 7→ n⊗R m and ψ : (n, m) 7→ m⊗R n are bilinear maps. For example,

φ(m + m0 , n) = n ⊗R (m + m0 ) = n ⊗R m + n ⊗R m0 = φ(m, n) + φ(m0 , n)

φ(m, n + n0 ) = (n + n0 ) ⊗R m = n ⊗R m + n0 ⊗R m = φ(m, n) + φ(m, n0 ).

10
 3 CATEGORICAL NOTIONS

They thus factor through φ̃ : M ⊗R N → N ⊗R M and ψ̃ : N ⊗R M → M ⊗R N respectively.

One confirms they are inverses; e.g.,

ψ̃(φ̃(m ⊗R n)) = ψ̃(n ⊗R m) = m ⊗R n.

Example 2.19. Consider the following sets and elements f and g:

(f : M × N → P ) ∈ HomR (M × N , P )
(g : M → HomR (N , P )) ∈ HomR (M, HomR (N , P )).

One might hope HomR (M × N , P ) and HomR (M, HomR (N , P )) are equal, e.g., via the “ob-
vious” currying map

(g(m))(−) := f (m, −).

However, this is not the case; the reason is that f (m, −) is not guaranteed to be an R-
module homomorphism (i.e., an element of HomR (N , P ))! In fact, the requirement for
f (m, −) to be a homomorphism (and for the f 7→ g map itself to be a homomorphism) is
exactly the requirement for f (−, −) to be bilinear:

(g(m))(r−) = f (m, r−) = rf (m, −) = r((g(m))(−))

(g(rm))(−) = f (rm, −) = rf (m, −) = (rg(m))(−).

Making this rigorous, one concludes that

HomR (M ⊗R N , P )  HomR (M, HomR (N , P ))

is the correct statement. This is the tensor-hom adjunction, which is a special case of
adjoint functors that we will revisit in the section on category theory.

3 Categorical notions
3.1 Categories
Remark. This treatment is mostly from class. [Link]
matema/docums/[Link] was used as an additional reference.
Definition 3.1. A category C consists of
• A class of objects ob(C).

• A class of morphisms homC (X, Y ) for each pair of objects X, Y ∈ ob(C). Here, X is
called the source and Y is called the target.

11
3.1 Categories 3 CATEGORICAL NOTIONS

• A composition map homC (X, Y ) × homC (Y , Z) → homC (X, Z) for every three objects
X, Y , Z ∈ ob(C). The composition of f , g is written g ◦ f .

such that the following axioms

• Associativity: If f ∈ homC (W , X), g ∈ homC (X, Y ), f ∈ homC (Y , Z), then

h ◦ (g ◦ f ) = (h ◦ g) ◦ f .

• Identity: For each object X there is an identity morphism idX ∈ homC (X, X), such that
every morphism f ∈ homC (Y , Z) satisfies

f ◦ idY = f = idZ ◦ f .

hold.

Remark. Think of categories as classes of dots and arrows between dots, where if two
arrows occur in sequence, they (implicitly) define a composed arrow. Associativity holds,
and every dot has an implicit identity arrow to itself.

Definition 3.2. An endomorphism is a morphism with the same source and target, i.e.,
an element of homC (X, X) for some object X.

Definition 3.3. An isomorphism is a morphism f ∈ homC (X, Y ) if there exists an inverse

morphism g ∈ homC (Y , X) such that f ◦ g = idY and g ◦ f = idX . The inverse morphism is
necessarily unique.

Definition 3.4. A category C is:

• Small if ob(C) and all homC (X, Y ) are sets.

• Large if it is not small.

• Locally small if all homC (X, Y ) are sets.

Example 3.5. We can formulate the different types of mathematical objects we’ve encoun-
tered thus far as living and interacting in their respective categories. The following are
all locally small categories:

Set Top (Top? ) Grp (Ab) Vectk , R-Mod

objects all sets top spaces (with basepoint) (abelian) groups ···
morphisms functions (pointed) cont functions homomorphisms ···
isomorphisms bijection (pointed) homeomorphism isomorphism ···

12
3.2 Functors 3 CATEGORICAL NOTIONS

Definition 3.6. The category associated to a poset X is the small category where the
objects are elements of X, and the morphisms are

(x ≤ y) if x ≤ y


X(x, y) = 
∅
 otherwise.

Composition is given by (y ≤ z) ◦ (x ≤ y) = (x ≤ z) and identity is given by idx = (x ≤ x).

Definition 3.7. The category associated to a monoid is the small category with one object
∗. The set of (endo)morphisms is homC (∗, ∗) = G, where the composition law is given by
h ◦ g = gh.

Definition 3.8. The category associated to a group, denoted CG for a group G, is the
category associated to it as a monoid. However, in this case every morphism is an isomor-
phism, since group elements have inverses.

Exercise 3.9. Conversely, show that a small category with one object such that every morphism
is an isomorphism, is a group.

Proof. Consider the set homCG (∗, ∗). The composition map gives a closed binary operation

homCG (∗, ∗) × homCG (∗, ∗) → homCG (∗, ∗)

and the axioms ensure associativity. The morphism id∗ satisfies the properties of a group
identity, and since every f ∈ homCG (∗, ∗) is an isomorphism, there exists g such that f ◦ g =
g ◦ f = id∗ , which is exactly the inverse property for a group.

Definition 3.10. A groupoid is a small category where every morphism is an isomor-

phism.

Remark. In this way, we see that monoids are essentially the same as small categories with
a single object. Groups are the subclass where the single object’s endomorphisms are all
isomorphisms, i.e., a group is a groupoid with only one object.

3.2 Functors
Definition 3.11. A (covariant) functor between categories F : C → D:

• Sends every object in C to an object in D

• Sends every morphism in homC (X, Y ) to a morphism in homD (F(X), F(Y )), such that
F(g ◦ f ) = F(g) ◦ F(f ) and F(idX ) = idF(X) .

Remark. The last property can be viewed as the requirement that commutative diagrams
are preserved:

13
3.2 Functors 3 CATEGORICAL NOTIONS

X goes to F(X)
g◦f F(g◦f )=F(g)◦F(f )
f F(f )
   #
Y /Z / F(Z)
g F(Y )
F(g)

Example 3.12. Let G be a group with an action φ on a set X. This action can be viewed as
a functor

Fφ : CG → Set

given by Fφ (∗) = X, which maps morphisms from CG (∗, ∗) to Set(X, X). Hence the action is
given by

φ : G × X → X, φ(g, x) = (Fφ (g))(x).

Similarly, F : CG → Top describes a G-space, and F : CG → Vectk describes a G-representation.

Example 3.13. Taking the free group is a functor F : Set → Grp. Set maps become group
homomorphisms between free groups.

Example 3.14. A functor CG → CH between categories associated to two groups are group
homomorphisms G → H.

Example 3.15. A functor between the categories associated to two posets can be viewed
as an order-preserving function.

Exercise 3.16. If F : C → D is a functor and f ∈ homC (X, Y ) is an isomorphism, then F(f ) is

an isomorphism.

Proof. Let f ∈ homC (X, Y ) an isomorphism with inverse g. Then

F(f ) ◦ F(g) = F(f ◦ g) = F(idF(Y ) )

F(g) ◦ F(f ) = F(g ◦ f ) = F(idF(X) )

which shows F(f ) is an isomorphism with inverse F(g).

Remark. The contrapositive of the above is what makes invariants a powerful concept.
For example, if the fundamental groups of two topological spaces are not isomorphic,
then they cannot be homeomorphic spaces.

Definition 3.17. A forgetful functor is a covariant functor that loses some of the algebraic
properties of the source category.

14
3.3 Natural transformations 3 CATEGORICAL NOTIONS

Example 3.18. There are canonical forgetful functors from each of Grp, Top, Vectk to
Set (forget the structure implied on each object’s underlying set by the restricted class
of morphisms). These functors are not as trivial as one might expect. Consider Grp →
Set. Group homomorphisms become non-privileged subsets of the set maps. Meanwhile,
multiple groups will map to their shared underlying set.
Definition 3.19. A contravariant functor between categories F : C → D:
• Sends every object in C to an object in D
• Sends every morphism in homC (X, Y ) to a morphism in homD (F(Y ), F(X)), such that
F(g ◦ f ) = F(f ) ◦ F(g) and F(idX ) = idF(X) .
Remark. The last property can be viewed as the requirement that commutative diagrams
are preserved, but with arrows in the opposite direction:

X goes to F(X)
O c
g◦f F(g◦f )=F(f )◦F(g)
f F(f )
 
/Z
Y g F(Y ) o F(Z)
F(g)

Remark. One can still study contravariant functors in the same realm as (covariant) func-
tors as follows: let the opposite category C op of a category C be the category where the
source and target of every morphism in C is reversed. Then contravariant functors C → D
are exactly covariant functors C → Dop .
Example 3.20. A map f : V → W of k-vector spaces has a corresponding dual map that
goes in the opposite direction
f ∗ : Hom(W , k) → Hom(V , k).
That is, Hom(−, k) : Vectk → Vectk reverses the direction of morphisms, making it a con-
op
travariant functor (i.e., a functor Vectk → Vectk ). A quick way to see this is that passing
to a map’s dual is equivalent to taking the transpose of its matrix, and that (AB)T = BT AT .

3.3 Natural transformations

Definition 3.21. A natural transformation is a family of morphisms written α : F =⇒ G,
where F, G : C → D, where each Z ∈ ob(C) has a morphism αZ ∈ homD (F(Z), G(Z)) such
that for all φ ∈ homC (X, Y ), the following diagram commutes:
F(φ)
F(X) / F(Y )

αX αY
 
G(X) / G(Y )
G(φ)

15
3.3 Natural transformations 3 CATEGORICAL NOTIONS

Remark. Intuitively, natural transformations are like morphisms of functors. They trans-
form one functor to another while respecting the associated categories. The commutative
diagram expresses that every φ : X → Y turning into F(φ) : F(X) → F(Y ), are fixed mor-
phisms αX , αY away from being G(φ) : G(X) → G(Y ).

Example 3.22. Consider the functors id, (−)∗∗ : Vectk → Vectk , where (−)∗∗ is the “double-
dual functor” given by

V 7→ V ∗∗ = (V ∗ )∗ = Hom(V , k)∗ = Hom(Hom(V , k), k),

and take αV : V → V ∗∗ given by v 7→ (f 7→ f (v)). One can verify that

f
V /W
αV αW
 
V ∗∗ / W ∗∗
f ∗∗

commutes. Hence we have a natural transformation α : id =⇒ (−)∗∗ .

Definition 3.23. α : F =⇒ G is a natural isomorphism if αC is an isomorphism for all

objects C ∈ ob(C).

Example 3.24. If one restricts themselves to the category of finite vector spaces over k,
then α : id =⇒ (−)∗∗ is a natural isomorphism. This reflects the idea that there are canoni-
cal isomorphisms V  V ∗∗ for all V (they are consistent with respect to morphisms V → W
and the respective V ∗∗ → W ∗∗ ). Meanwhile, although V  V ∗ as well, this requires a non-
canonical choice of basis in each case making it impossible for the required diagrams to
commute everywhere.

Remark. There is a generalization of categories known as 2-categories, where in addition

to objects and morphisms between objects, there is a notion of 2-morphisms between
morphisms. For example, Cat is a 2-category whose objects are small categories, whose
morphisms are functors, and whose 2-morphisms are natural transformations.

Exercise 3.25. Let I be the interval category

I = {0 → 1}

(it has two objects and one non-trivial morphism). Prove that a natural transformation is
precisely a functor

η : C×I → D

such that the diagram of functors

16
3.4 [DRAFT] Universal properties 3 CATEGORICAL NOTIONS

C × {0}
_
F
 η "
C ×O I /D
<

? G
C × {1}
commutes (this diagram lives in Cat). Note that homC×I ((x, s), (y, t)) is given by homC (x, y) ×
homI (s, t).
Remark. This exercise is more compelling when done after learning about homotopies. It
asserts that natural transformations are their categorical analogue.
Definition 3.26. Categories C, D are equivalent categories if there are functors F : C → D
and G : D → C such that F ◦ G =⇒ 1D and G ◦ F =⇒ 1C are natural isomorphisms of
functors.
Remark. Continuing the analogy from the exercise, an equivalence of categories is the
categorical analogue of being homotopy equivalent. This notion is more useful than a
potential notion of isomorphism (e.g., where F ◦ G = 1D exactly). Most importantly, an
equivalence still preserves “categorical concepts” such as monomorphisms, limits, etc.
(see [Link]
Example 3.27. Let C be a category whose objects are non-negative integers and whose
homomorphisms are
homC (m, n) = {n × m matrices in R}.

Then C turns out to be equivalent to the category Vectfinite

R . Take F as a functor from
m
m to the “standard” R and G as a functor from an n-dimensional R-vector space to n.
Matrices are taken to the corresponding linear maps, and vice versa.
Remark. This example exhibits that an equivalence of categories does not even ensure a
bijection of objects. However, categorical concepts are preserved, which is what matters
(e.g., think about the relative importance of isomorphism over equality).
Example 3.28. In algebraic geometry, the category of affine schemes is equivalent to the
opposite category of commutative rings. Take F as a functor from an affine scheme to its
ring of global sections, and take G as a functor from a commutative ring to its spectrum.

3.4 [DRAFT] Universal properties

Remark. This section is based on [1.5], Wikipedia, and the previously mentioned refer-
ence. We reframe via the notions of initial and final topologies as discussed by Wikipedia
(not covered in class).

17
3.4 [DRAFT] Universal properties 3 CATEGORICAL NOTIONS

A universal property can be thought of as a description (with respect to other

objects/morphisms) that uniquely defines an object in a category up to isomorphism,
whereby the object can be thought of as the most “efficient solution” to a diagram (i.e.,
universal with respect to the diagram) such that all other solutions factor through it.
This is terribly vague; a technical definition can be found at [Link]
org/wiki/Universal_property. It is best to proceed by providing examples.
Example 3.29. (TODO: Universal property of the free group) [Link]
org/wiki/Free_group
Example 3.30. We saw that the tensor product M ⊗R N of modules was defined by the
universal property of being factored through by any R-bilinear maps from M × N .
Definition 3.31. The pushout or fibered sum X ∪Z Y of morphisms f : Z → X and g :
Z → Y is given by an object P and morphisms i1 : X → P , i2 : Y → P such that
g
Z /Y

f i2
 
X /P
i1

commutes, and that (P , i1 , i2 ) is universal for this diagram; i.e., for any other solution
(Q, j1 , j2 ) making such a diagram commute, there exists a unique u that makes the follow-
ing diagram commute:
g
Z /Y

f i2
  j2
X /P
i1
u

j1 . Q.

Example 3.32. (TODO: pushout of sets)

Example 3.33. (TODO: pushout of groups is amalgamated product)
Example 3.34. (TODO: pushout of pointed spaces is wedge product)
Definition 3.35. The pullback or fiber product X ×Z Y of morphisms f : X → Z and
g : Y → Z is given by an object P and morphisms p1 : P → X, p2 : P → Y such that
p2
P /Y
p1 g
 
X /Z
f

18
3.4 [DRAFT] Universal properties 3 CATEGORICAL NOTIONS

commutes, and that (P , p1 , p2 ) is universal for this diagram; i.e., for any other solution
(Q, j1 , j2 ) making such a diagram commute, there exists a unique u that makes the follow-
ing diagram commute:

Q j2
u
 p2 
P /Y
j1
p1 g
"  
X / Z.
f

Example 3.36. (TODO: pullback of sets)

Example 3.37. (TODO: pullback of rings)
Definition 3.38. Let fi : X → Yi be a family of functions. The initial topology on X is
the coarsest topology such that all fi are continuous. It is characterized by the following
property: a function g : Z → X is continuous if and only if fi ◦g : Z → Yi are all continuous.
(TODO: diagram)
Example 3.39. The subspace topology is the initial topology with respect to an inclusion
map A → X.
(TODO: diagram)
Example 3.40. The product topology is the initial topology with respect to the projection
maps (the underlying space is the set product).
(TODO: diagram)
Definition 3.41. Let fi : Xi → Y be a family of functions. The final topology on Y is
the finest topology such that all fi are continuous. It is characterized by the following
property: a function g : Y → Z is continuous if and only if g◦fi : Xi → Z are all continuous.
(TODO: diagram)
Example 3.42. The quotient topology is the final topology with respect to a surjective
map X → Y .
(TODO: diagram)
Example 3.43. The disjoint union topology is the final topology with respect to the in-
clusion maps (the underlying space is the disjoint set union).
(TODO: diagram)

19
3.5 Adjoint functors 3 CATEGORICAL NOTIONS

3.5 Adjoint functors

Definition 3.44. We say that S : D → C and T : C → D are an adjoint pair of functors
(where S is left adjoint to T , or T is right adjoint to S) if for all X ∈ ob(D) and Y ∈ ob(C),
we have bijections ηX,Y via which
homC (SX, Y )  homD (X, T Y )
such that the implicit transformations are natural in X and Y . Explicitly, for all X and Y ,
the pair of (covariant and contravariant, respectively) functors
homC (SX, −)  homD (X, T −)
homC (S−, Y )  homD (−, T Y )
are naturally isomorphic.
Example 3.45. Consider the functors − ⊗R N : R-Mod → R-Mod and Hom(N , −) : R-Mod →
R-Mod. The tensor product is left adjoint to the Hom map; note that
homR-mod (SX, Y ) = homR-mod (X ⊗R N , Y )  homR-mod (X, Hom(N , Y )) = homR-mod (X, T Y )
via the tensor-hom adjunction we saw earlier.
Remark. This is a consequence of R-mod equipped with
⊗R : R-mod × R-mod → R-mod
being a closed (symmetric) monoidal category. Such categories are characterized by the
property that the associated “bifunctor” (in this case ⊗R ), when restricted to be a functor
− ⊗R N for some fixed N , has a right adjoint functor denoted [N , −]. We will see other
adjunctions of this type when discussing mapping spaces.
Example 3.46. Let S : Set → Grp be the functor that takes a set to its free group, and let
T : Grp → Set be the forgetful functor (taking the group to its underlying set). Then S is
left adjoint to T . To see this, the desired assertion is that for any X ∈ Set and Y ∈ Grp,
homGrp (SX, Y ) = homGrp (FX , Y )  homSet (X, |Y |) = homSet (X, T Y ),
and the isomorphism follows directly from the universal property of the free group.
Exercise 3.47. Let  : T S =⇒ 1D and η : 1C =⇒ ST be natural transformations. Show that
the diagrams

ST
; SX T : ST Y
ηS S Tη T
$ $
SX
id / SX TY
id / TY

commute for all X, Y if and only if T , S are adjoint.

20
 4 THE FUNDAMENTAL GROUP

4 The fundamental group

Remark. Corresponding readings are [2.1-2.2], though these are not comprehensive. The
presentation here is primarily from class.

4.1 Connectedness and paths

Definition 4.1. X is connected if X cannot be written as the disjoint union of two non-
empty open sets.

Definition 4.2. X is path-connected if for all x, y ∈ X, there exists a continuous function

p : [0, 1] → X with p(0) = x and p(1) = y (a path between x and y).

Proposition 4.3. Path-connectedness implies connectedness.

Proof. Suppose X is path-connected but not connected, with X = U t V disjoint non-

empty open sets. Let x ∈ U and y ∈ V , and let p : [0, 1] → X be the path connecting them.
Then

[0, 1] = p−1 (X) = p−1 (U t V ) = p−1 (U ) t p−1 (V ),

a disjoint non-empty union of open sets. This contradicts the fact that [0, 1] is connected.

Proposition 4.4. The existence of a path is an equivalence relation on points in X.

Proof. Reflexivity is given by taking p : [0, 1] → X to be the constant map to x. Transitivity

is given by gluing paths together (re-parameterizing so that the domain is still [0, 1]).
Symmetry is given by taking q(x) := p(1 − x).

Proposition 4.5. A union of (path-)connected spaces with a point in common are (path-
)connected. The image of a (path-)connected space under a continuous map is (path-)connected.

Proof. For connectedness:

a) Let X, Y be connected and consider X ∪ Y with a common element x (implicit here is

that the subspace topologies on X and Y give their original topology). The key is to
argue that for one of the connected Z = X, Y , the disjoint union (A ∩ Z) t (B ∩ Z) is a
non-empty (and necessarily disjoint, open by relativity) partition of Z, a contradiction.
There exists a common element, it is WLOG in A (and thus it is in every A ∩ Z); then,
choose the other element in B ∩ Z (which is non-zero for some Z).

b) The disjoint non-empty open partition of a space would also partition its pre-image.

For path-connectedness:

21
4.2 Homotopy and homotopy equivalence 4 THE FUNDAMENTAL GROUP

a) Construct paths in the new space by passing through the shared point as needed.

b) The image of a path is a path.

These observations prove the corresponding statements.

Example 4.6 (Topologist’s sine curve). The converse is not true, even in a metric space.
Furthermore, path-connectedness is not preserved by closure (though connectedness is).
Consider the graph Γ = {(x, sin(1/x)) | x ∈ (0, 1]} ⊆ R2 . Its closure is

Γ = Γ ∪ ({0} × [−1, 1]).

is closed and bounded and also compact. Pre-closure, we had the graph of a continuous
function, which is path-connected (e.g., parameterize with (0, 1]). However, its closure
remains connected, but it is not locally connected nor path-connected. Problems oc-
cur around (0, 0). See [Link]
[Link] for details.

Definition 4.7. The equivalence relation of path-connectedness induces a partition of X

into path-connected components. We denote the set of path components as π0 (X).

4.2 Homotopy and homotopy equivalence

Definition 4.8. Two continuous maps f , g : X → Y are homotopic, written f ' g, if there
exists a continuous h : X × I → Y (called a homotopy) such that h(x, 0) = f (x) and h(x, 1) =
g(x).

Exercise 4.9. Homotopy is a equivalence relation on continuous maps X → Y .

Example 4.10. Imagine two maps f , g from X = [0, 1] to two curves in a space Y . Then
we have a homotopy which continuously transforms the image from the first curve to the
second. Note that this can fail when there is a “hole”.

Definition 4.11. A function f is null-homotopic if it is homotopic to a constant function.

Definition 4.12. Two spaces X, Y are homotopy equivalent (written X ' Y ) if there exist
continuous maps f : X → Y and g : Y → X such that f ◦ g ' idY and g ◦ f ' idX .

Remark. Compare this definition with X, Y being homeomorphic, where we need f ◦ g =

idY and g ◦ f = idX .

Exercise 4.13. Homotopy equivalence is a equivalence relation on topological spaces.

22
4.2 Homotopy and homotopy equivalence 4 THE FUNDAMENTAL GROUP

Example 4.14. Let X = R2 , Y = {0}. We want to show the two are homotopy equivalent.
Let

f : X → Y, f (x, y) = 0
g: Y → X, g(0) = (0, 0).

Then f ◦g = idY , so f ◦g ' idY . Meanwhile, (g ◦f )(~

v ) = (0, 0). We need g ◦f ' idX . Consider
the following null-homotopy:

H : X × [0, 1] → X, v , t) = t~
H(~ v.

Then H(~ v , 0) = 0~
v = (0, 0) and H(~
v , 1) = idX , exhibiting that g ◦ f ' idX . We can visualize
this as the continuous contraction of the space to the origin.

Intuitively, homotopy equivalence indicates that two spaces can be transformed into
each other by bending, shrinking, and expanding. Homeomorphism implies homotopy
equivalence, but not vice versa (the disk and the point are homotopy equivalent by shrink-
ing along radial lines, but not homeomorphic since there is certainly no bijection).

Definition 4.15. X is contractible if it is homotopy equivalent to the space of one ele-

ment. One can show this is equivalent to idX : X → X being null-homotopic.

Example 4.16. Let X = R2 \{(0, 0)} and Y = S 1 = {(x, y) ∈ R2 | x2 + y 2 = 1}. Note that Y ⊆ X.
We want to show these are homotopy equivalent. Let

g : Y → X, v ) = v~
g(~
v~
f : X → Y, f (~
v) = .
v~

v~
Then f ◦ g = idY and (g ◦ f )(~
v) = . We show the latter is homotopic to idX via:
||v~||

v~
H : X × [0, 1] → X, v , t) = t~
H(~ v + (1 − t) .
v~

v~
Then H(~
v , 0) = and H(~
v , 1) = v~.
||v~||
Remark. Note that homotopy equivalence is weaker than homeomorphism. In this previ-
ous example, we saw the homotopy equivalence of a non-compact space and a compact
space. What let this argument pass through is that in some sense, homotopy equivalence
preserves the number of holes but not other topological properties.

23
4.3 The fundamental group 4 THE FUNDAMENTAL GROUP

4.3 The fundamental group

Definition 4.17. Let x ∈ X. A loop at b is a path p : [0, 1] → X with p(0) = p(1) = b.
Remark. A Moore loop is a variant definition of a loop which attaches a domain R/nZ to a
loop. The necessary conclusions (e.g., the fundamental group) turn out to be equivalent.
Remark. We adopt the following convention: A homotopy of loops p, q at b has to fix the
endpoints at b. Thus,

H : [0, 1] × [0, 1] → X

(the first [0, 1] represents the path, the second [0, 1] represents time), so that

H(x, 0) = p(x), H(x, 1) = q(x), H(0, t) = H(1, t) = b.

Definition 4.18. The fundamental group π1 (X, ?) is the group produced by taking the
set of loops at ?, modulo the equivalence relation of loop homotopy.
• The group operation is induced via loop concatenation:

1
p(2t) if 0 ≤ t ≤ 2


(p ∗ q)(t) = 
q(2t − 1) if 12 ≤ t ≤ 1.


• This only gives e.g., associativity if we work modulo homotopy. One can check that

(p ∗ q) ∗ r ' p ∗ (q ∗ r)

explicitly, as desired.
• Our identity element is (the equivalence class) of the constant loop at ?, which we call
c? . One can construct a homotopy p ∗ c? ' p, for example.
• The inverse element p−1 of a loop p can be defined as p−1 (t) := p(1 − t). One can con-
struct a homotopy whereby one traverses less and less of the path until it becomes the
constant map [HW1, Q1].
If [p] denotes the equivalence class of p, then it turns out

[p] · [q] := [p ∗ q]

is a well-defined group operation. That is, loop concatenation ∗ gives a group operation
up to homotopy.
Definition 4.19. Let f : X → Y where p is a loop at ? ∈ X. Then f ◦ p is a loop at f (?) ∈ Y .
In this way, f induces the induced homomorphism written

f∗ : π1 (X, ?) → π1 (Y , f (?)), [p] 7→ [f ◦ p].

24
4.3 The fundamental group 4 THE FUNDAMENTAL GROUP

Remark. See [HW1, Q6] where we check that the fundamental group and the induced
homomorphism have the desired properties:
1) is the assertion that [p][q] = [p ∗ q] is well-defined.

2) is the assertion that f∗ ([p]) = [f ◦ p] is well defined.

3) is the assertion that f∗ is a homomorphism (i.e., f∗ ([p][q]) = f∗ ([p])f∗ [q])).

4) is the assertion that f ' g implies f∗ = g∗ .

Corollary 4.20. The following hold from the property f ' g implies f∗ = g∗ :
• If f : X → Y is a homotopy equivalence, then f∗ is an isomorphism (i.e., bijective group
homomorphism) [HW1, Q7]

• Homotopy-equivalent spaces have the same fundamental group.

• A contractible space has trivial fundamental group.

Example 4.21 (Fundamental group of S 1 ). We claim that

π1 (S 1 , (1, 0))  Z.

We sketch the proof visually: consider the real line coiled over S 1 , such that Z is the
preimage of (1, 0). Following the loop along the real line, we end at one of these integers.
If we class the loops based on which of Z our loop ends up at, this turns out to construct
an isomorphism between the homotopy classes of loops and Z. See [Link]
[Link]/˜may/VIGRE/VIGRE2011/REUPapers/[Link], which formalizes this via
path lifting and covering spaces.
Example 4.22. The fundamental group is a functor

π1 : Top* → Grp.

The morphisms are mapped via

f : (X, x) → (Y , y) 7→ f∗ : π1 (X, x) → π1 (Y , y)

and composition is preserved since (g ◦ f )∗ = g∗ ◦ f∗ .

Definition 4.23. The fundamental groupoid ΠX of a space X is the category whose ob-
jects are points in X, and whose morphisms are

ΠX(x, y) = {paths from x to y}/{homotopy}.

This is a groupoid since the existence of return paths exhibits isomorphism. Note that
ΠX(x, x) forms a group isomorphic to π1 (X, x).

25
4.4 Applications 4 THE FUNDAMENTAL GROUP

4.4 Applications
Remark. The following are Theorems 1.8 and 1.9 in Hatcher.

Theorem 4.24 (Fundamental theorem of algebra). A polynomial p(z) = zn + an−1 zn−1 + · · · +

a1 z + a0 has a root in C if n ≥ 1.

Proof. • Suppose

p(z) = zn + an−1 zn−1 + · · · + a0

has no roots. Define the family of functions

p(re2πis )/p(r)
fr (s) = .
p(re2πis )/p(r)

where r ≥ 0.

• Then |fr (s)| = 1 (the function is well-defined everywhere, and no denominator is zero
by hypothesis) and fr (0) = fr (1) = 1, making fr a loop in S 1 = {z ∈ C | |z| = 1} based at
1 ∈ C.

• We claim that [f0 ] = [fr ] = 0 ∈ π1 (S 1 , 1): First note that f0 (s) = 1, i.e., the constant func-
tion (equivalently, view f0 as the constant loop at 1). Then [f0 ] = [fr ] with the homotopy
exhibited by fr varying from 0 to r (continuous since composition of continuous func-
tions like multiplication, norm, non-zero division, etc.). That is, every fr is homotopic
to the trivial loop.

• Now, define

pt (z) = zn + t(an−1 zn−1 + · · · + a1 z + a0 )

and note that p0 (z) = zn and p1 (z) = p(z), exhibiting a homotopy zn ' p(z). By the
reverse triangle inequality:

|pt (z)| ≥ |z|n − t(an−1 zn−1 + · · · + a1 z + a0 )

• For |z| = r sufficiently large (r ≥ max{|an−1 | + · · · + |a0 | , 1}), we have the inequality

|z|n = r · r n−1 ≥ (|an−1 | + · · · + |a0 |) |z|n−1 ≥ an−1 zn−1 + · · · + a1 z + a0

(the last inequality holds since |z| = r ≥ 1). Then for t ∈ [0, 1], we have |pt (z)| > 0 when
|z| = r.

26
4.4 Applications 4 THE FUNDAMENTAL GROUP

• Since pt (z) , 0 if |z| = r for all t ∈ [0, 1], we fix this r which allows us take the following
homotopy (no divisions by zero anywhere)

p0 (re2πis )/p0 (r) p1 (re2πis )/p1 (r) p(re2πis )/p(r)

=⇒ ' =
p0 (re2πis )/p0 (r) p1 (re2πis )/p1 (r) p(re2πis )/p(r)
=⇒ e2πins ' fr (s).

• The left-hand side as a function of s corresponds to the class n ∈ π1 (S 1 , 1). But we saw
that [fr ] = [f0 ] =∈ π1 (S 1 , 1). Our polynomial is not constant, so n , 0, a contradiction.

Theorem 4.25 (Brouwer’s fixed point theorem). Let

D n = {~
v ∈ Rn | v~ ≤ 1}
S n−1 = {~
v ∈ Rn | v~ = 1}

and take f : D n → D n continuous. Then there exists v ∈ D n such that f (v) = v, i.e., a fixed
point.

Proof. We complete the n = 2 case. Assume f does not fix a point.

−−−−−→
• Define g : D 2 → S 1 such that the g goes to the intersection of the ray f (~ v with S 1 (this
v )~
is well-defined since f (~
v ) , v~ everywhere). Observe that g is continuous (informally, f
is continuous and so the ray produced would vary continuously with v~).

• If v~ ∈ S 1 , we have g(~
v ) = v~. If ι denotes inclusion, then we have

g
S1
ι / D2 / S1

where g ◦ i = idS 1 .

• Passing to fundamental groups, we get

ι∗ g∗
π1 (S 1 ) / π1 (D 2 ) / π1 (S 1 ).

Since π1 (S 1 )  Z and π1 (D 2 )  0 (by contractibility, see [HW1, Q4]), then (g ◦ i)∗ =

(idS 1 )∗ = idZ , but that is impossible since it must factor through the trivial group
π1 (D 2 ). This is a contradiction.

For the cases n > 2, one uses πn or the homology groups (see [Hatcher, Corollary 2.11]).

27
4.5 [TODO] Van Kampen’s theorem 5 HIGHER HOMOTOPY GROUPS

Exercise 4.26. Prove the case n = 1 using the intermediate value theorem.

Proof. We want f : D 1 = [0, 1] → D 1 = [0, 1] continuous. Suppose we have no fixed point;

then g(t) := f (t) − t , 0 on our domain. g(0) = f (0) > 0 and g(1) = f (1) − 1 < 0. By the
intermediate value theorem, there is some value t ∈ [0, 1] for which g(t) = 0, a contradic-
tion. (For a visual picture, draw the graphs of id[0,1] and an arbitrary f , and note that they
always intersect.)

4.5 [TODO] Van Kampen’s theorem

Remark. (TODO: A brief shoutout to Van Kampen’s theorem for fundamental groups,
plus an example of a pushout: [Link]
pdf, or Hatcher)

5 Higher homotopy groups

5.1 Mapping spaces
Remark. The unbased mapping space properties described below are covered in [2.2]. The
conditions from which we get the adjunction relation are given by Hatcher.

Definition 5.1. The mapping space of X, Y is the topological space

Map(X, Y ) = Y X = {continuous functions X → Y }

with the compact-open topology: for every pair of compact sets C ⊆ X and open sets
U ⊆ Y , let W (C, U ) be the set of functions f with f (C) ⊆ U , then take the topology for
which the set of all W (C, U ) is a subbasis.

Remark. If X is compact and Y has a metric, then Y X has a metric

d(f , g) = sup d(f (x), g(x))

x∈X

which induces the same topology as the compact-open topology.

Example 5.2. Let X, Y , Z be spaces, with Y locally compact. Then

Map(X × Y , Z) = Z X×Y  (Z Y )X = Map(X, Map(Y , Z)),

a homeomorphism. The elements of these spaces look like

f : X × Y → Z, g : X → ZY .

28
5.2 Operations on pointed spaces 5 HIGHER HOMOTOPY GROUPS

The idea is that for every x ∈ X fed into f , one induces a function Y → Z. Thus g is a
function from this x to the functions Y → Z. The bijective correspondence is observable
by noting that each can be defined in terms of the other:

(x, y) 7→ (g(x))(y) ↔ x 7→ (y 7→ f (x, y)).

This holds unconditionally if we were considering arbitrary functions. However, we now

work on the mapping spaces with compact-open topologies. It turns out that Y ’s local
compactness is strong enough to ensure that f is continuous if and only if g is continuous,
and that the compact-open topology allows for a bijection that is a homeomorphism.

Remark. This parallels the tensor-hom adjunction, in the sense that

− × Y : Top → Top, Map(Y , −) : Top → Top

are an adjoint pair of functors for Y locally compact. The parallel is more appropriate if
we did not have to make such a specific restriction; instead, if we restricted X, Y , Z to be
objects in the category CGWH of compactly-generated weak Hausdorff spaces (see [Con-
cise, §5] and [Link]
then the bijection holds without restriction (as in the case of tensor-hom).

Definition 5.3. The based mapping space of (X, ?x ), (Y , ?y ) is

Map? ((X, ?x ), (Y , ?y )) = {continuous functions X → Y mapping ?x to ?y }.

equipped with the compact-open topology.

Example 5.4. Analogous to (unbased) mapping spaces, under the compact-open topology
and the constraint that Y is locally compact, one gets the homeomorphism

Map? ((X ∧ Y , ?), (Z, ?z )))  Map? ((X, ?x ), Map? ((Y , ?y ), (Z, ?z )))

where ∧ denotes the smash product defined below.

Remark. As with the unbased case, we can abandon the constraint of Y being locally
compact if X, Y , Z live in a nice category like CGWH* . More formally, ∧ makes CGWH* a
closed symmetric monoidal category (see the section on adjoint functors).

5.2 Operations on pointed spaces

Remark. The following is a synthesis of part of Chapter 0 in Hatcher, along with the
readings and the class’ content.

Notation 5.5. When the existence of a basepoint is obvious and necessary, we might sim-
ply write X to denote the pointed space (X, ?x ).

29
5.2 Operations on pointed spaces 5 HIGHER HOMOTOPY GROUPS

Definition 5.6. The wedge product, written X ∨ Y , of two pointed spaces (X, ?x ) and
(Y , ?y ) is the quotient of the disjoint union X t Y given by identifying ?x ∼ ?y (this is
necessary since working with basepoints isn’t canonical for disjoint unions).

Example 5.7. The wedge product X ∨ Y is the pushout in the following diagram:
?y
{?} /Y

?x i2
 
X / X ∨Y.
i1

Definition 5.8. The smash product, written X ∧ Y , of two pointed spaces (X, ?x ) and
(Y , ?y ) is the quotient of the product space X × Y given by identifying (x, ?y ) ∼ (?x , y) for
all x, y. These are copies of X and Y that only intersect at (?x , ?y ). More precisely,

X ∧ Y := (X × Y )/(X ∨ Y ).

Example 5.9. S 1 × S 1 is the 2-torus, but S 1 ∧ S 1 = S 2 . In general, S m ∧ S n  S m+n . A nice

way to visualize this is to consider the torus as a square with its opposite sides identified.
These sides form two circles intersecting at a point, a copy of S 1 ∨ S 1 . When they are
identified to be one point (that is, (S 1 × S 1 )/(S 1 ∨ S 1 ), the square becomes the sphere S 2 .

Example 5.10. We saw earlier that the smash product gave a homeomorphism of based
mapping spaces, analogous to the tensor product and the tensor-hom adjunction. Sim-
ilarly, the smash product satisfies the universal property that every map from (X, ?x ) ×
(Y , ?y ) that preserve basepoints separately in each variable factors through the smash
product:

(X, ?x ) × (Y , ?y )
∧ / (X ∧ Y , ?)

f˜
f ' 
(Z, ?z )

(see [Link]

Definition 5.11. Given a non-pointed space X, the suspension SX is the quotient of X × I

where X × {0} is collapsed to one point and X × {1} is collapsed to another.

Example 5.12. The motivating example is X = S 1 . Then SX can be viewed as a double-

pointed cone with base X, from which we see that S(S 1 ) = S 2 with the two “suspension
points” as (0, 0, ±1). This argument generalizes in the obvious way for X = S n .

30
5.2 Operations on pointed spaces 5 HIGHER HOMOTOPY GROUPS

Remark. For continuous f : X → Y , we can define Sf : SX → SY to be Sf ([x, t]) = [f (x), t]

(the brackets indicate that we are working modulo equivalence). Then we have the sus-
pension functor

S : Top → Top

which roughly increases the dimension of a space by one.

Definition 5.13. Given a pointed space (X, ?x ), the reduced suspension is given by

ΣX := (X × I)/(X × {0} ∪ X × {1} ∪ {?x } × I),

or equivalently, SX with the line joining the two suspension points also identified. The
identification is done to ensure there is a canonical basepoint, namely the equivalence
class of all the identified points.
Remark. Likewise we have the reduced suspension functor

Σ : Top* → Top*

which roughly increases the dimension of a pointed space by one.

Proposition 5.14. ΣX  X ∧ S 1
Proof. Briefly: taking X ×I and then identifying (x, 0) ∼ (x, 1) gives you X ×S 1 . The further
identification of X × {0} = X × {1} is like collapsing along X, and the further identification
of {?X } × I is like collapsing along S 1 . This gives you

((X × I)/(x, 0) ∼ (x, 1))/(X × {0} ∪ {?x } × I)

 (X × S 1 )/(X × {0} ∪ {?x } × I)
 (X × S 1 )/(X ∨ S 1 ),

which is just X ∧ S 1 .

Definition 5.15. Given a pointed space (X, ?x ), its loop space is given by

ΩX := Map? (S 1 , X),

that is, the based mapping space of maps from S 1 to X. This gives the loop functor on
pointed spaces, whereby for f : X → Y we take Ωf : Map? (S 1 , X) → Map? (S 1 , Y ) to be
(post-)composition with f .
Proposition 5.16.

Map? (ΣX, Y )  Map? (X, ΩY )

Map? (X ∧ I? , Y )  Map? (I? , Map? (X, Y )).

31
5.3 Higher homotopy groups 5 HIGHER HOMOTOPY GROUPS

Proof. Both follow from the adjunction property for based mapping spaces that we saw in
the previous section, along with the definitions of Σ, Ω and the symmetric nature of the
smash product.

Example 5.17. We conclude that Σ : Top* → Top* is a left adjoint to the functor Ω :
Top* → Top* .

5.3 Higher homotopy groups

Our observations regarding based mapping spaces and associated operations on
pointed spaces (the wedge and smash product, along with loop and reduced suspension
spaces) let us define the homotopy groups in full generality.

Definition 5.18. The based mapping space up to homotopy for based maps X → Y is
denoted [X, Y ] and given by Map? (X, Y ) modulo homotopy equivalence.

Definition 5.19. The n-th homotopy group is defined by

πn (X) := [S n , X].

This generalizes the definition of π1 (X) as based loops up to homotopy.

Proposition 5.20. Every higher homotopy group is the fundamental group of some space;
namely

πn (X) = π1 (Ωn−1 X).

Proof. Note that

Map? (S n , X)  Map? (S 1 ∧ S n−1 , X)  Map? (S 1 , Map? (S n−1 , X)),

where we use our based mapping homeomorphism and the smash product of spheres.
Taking our mapping spaces modulo homotopy gives

πn (X)  π1 (Ωn−1 X).

in a canonical way (the homeomorphism is induced from the based mapping homeomor-
phism, which expresses a pair of adjoint functors).

Remark. It follows that the higher homotopy groups are also functors

πn : Top* → Grp,

being compositions of the functors Ω and π1 .

Proposition 5.21. πn is an abelian group for n ≥ 2.

32
5.3 Higher homotopy groups 5 HIGHER HOMOTOPY GROUPS

Proof. • In a nice category of spaces (e.g., CGWH), one can show that ΩX is something
known as an H-space, also known as a topological unital magma. The result that the
fundamental group of a topological group is abelian generalizes to H-spaces, and since
πn (X) = π1 (Ωn−1 X) we have the result.

• Without assuming a nice category, note that it suffices to consider n = 2 (since πn (X) =
π2 (Ωn−2 (X))). It suffices to consider n = 2 (since πn (X) = π2 (Ωn−2 (X))). Homotopy
classes of maps S 2 → X are equivalent to classes of maps I 2 → X that take the bound-
ary to the basepoint. The essential idea is that a concatenation f ∗ g : I 2 → X can be de-
formed via the extra dimension into g ∗ f : I 2 → X (see [Link]
com/a/161519).

Example 5.22. For the n-sphere S n , we have

πq (S n ) = 0 for q < n, πn (S n ) = Z.

The last statement is not easy to prove, and is most immediately viewed as a consequence
of a result known as the Hurewicz theorem. There are also:

π4n−1 (S 2n ) = Z ⊕ finite
πq (S n ) = is finite otherwise and non-zero for infinitely many q > n, n ≥ 2.

Definition 5.23. Let f : X → Y . This induces f∗ : πn (X) → πn (Y ) (think S n → X → Y

where the second map is f ). We say that f is a weak homotopy equivalence if f∗ is an
isomorphism for all n ≥ 0.

Proposition 5.24. Homotopy equivalence implies weak homotopy equivalence.

Proof. Suppose X ' Y via maps f : X → Y and g : Y → X. For n = 0, note that homotopy
equivalence preserves the number of path-connected components. For n ≥ 1, since f ◦ g '
idX , we have g∗ ◦ f∗ = (g ◦ f )∗ = idX∗ , we conclude that f∗ is an isomorphism of groups.

Remark. For “nice spaces” like CW-complexes, the converse is true. Furthermore, non-
nice-spaces are weak homotopy equivalent to nice spaces.
Homotopy groups are more sophisticated invariants than the homology groups and
thus better at distinguishing spaces. However, they are hard, if not impossible, to com-
pute. Meanwhile, homology groups will turn out to be independent of the decomposition
we choose for turning a space into a simplicial complex.

33
 6 SIMPLICIAL COMPLEXES

6 Simplicial complexes
Remark. This is a synthesis of Section 4 of May’s Finite Spaces book, Munkres’ Elements of
Algebraic Topology, and things mentioned in class.
Ultimately we want to associate algebraic objects to arbitrary spaces. We start by
considering spaces that are homeomorphic to a net of edges, triangles, tetrahedra, etc.
(e.g., polyhedra). Many spaces (e.g., the torus) can be viewed in this way, leading to the
notion of simplicial complexes.

6.1 Simplicial complexes

Definition 6.1. A set of points {v0 , . . . , vn } is geometrically independent if the vectors
{vi − v0 } are linearly independent. One verifies that this is equivalent to the requirement
that
n
X n
X
ti = 0 and ti vi = 0 =⇒ t0 = t1 = · · · = tn = 0.
i=0 i=0

Definition 6.2. The n-simplex σ spanned by a geometrically independent set {v0 , . . . , vn }

is the convex hull of the points, i.e.,
 
 n n 
X
 X 
σ = t v | t = 1 and t ≥ 0 for all i .

 i i i i 

 
i=0 i=0

Simplices spanned by a subset of {v0 , . . . , vn } are faces of σ . The set {v0 , . . . , vn } is the vertex
set of σ . The dimension of σ is n.

Definition 6.3. A (geometric) simplicial complex K is a collection of simplices in RN

where:

• Every face of a simplex in K is also a simplex in K

• The intersection of two simplices in K is also a simplex in K

The vertices V (K) of a simplicial complex is the union of its simplices’ vertex sets; i.e., it
is the set of 1-simplices in K.

Example 6.4. The hollow tetrahedron with points at (0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1) is
a simplicial complex. The simplices are every proper, non-empty subset of these four
points. The four points are geometrically independent (for example, take v0 = (0, 0, 0))
and thus all the subsets define simplices. Closure under taking faces and intersections is
immediate.

34
6.2 Abstract simplicial complexes 6 SIMPLICIAL COMPLEXES

Definition 6.5. A map of simplicial complexes f : K → L is a map V (K) → V (L) where

f (σ ) is a simplex in L for all simplices σ ∈ K.
Definition 6.6. The n-skeleton K n of a simplicial complex is the simplicial complex de-
termined by K0 , . . . , Kn , i.e., K’s simplices of dimension up to n.
Definition 6.7. The geometric realization |K| of a simplicial complex K is the topological
space where:
• The underlying set is the union of the simplices of K

• The topology is that V ⊆ |K| is closed exactly when V ∩ σ is closed in σ ’s subspace

topology, for all σ ∈ K.
Remark. It turns out that if K is a finite collection of simplices, the subspace topology of
|K| as a subset of RN is exactly the same as the topology defined above. See [Munkres, §2]
to see that |K|’s topology is finer than the subspace topology in general.
Definition 6.8. The category of simplicial complexes, SCxs has geometric simplicial
complexes as its objects and maps of simplicial complexes as its morphisms.
Definition 6.9. The geometric realization functor

|· · · | : SCxs → Top

takes simplicial complexes to their geometric realizations, and a map g : K → L of simpli-

cial complexes to the continuous map
X X
|g| : |K| → |L| , ti vi 7→ ti g(vi ).

6.2 Abstract simplicial complexes

Definition 6.10. An abstract simplicial complex X is comprised of the following data:
• A set X0 whose elements we call vertices,

• Sets Xn of (n + 1)-element subsets of X0 , whose elements we call n-simplices.

The sets of n-simplices are subject to the requirement that every (k + 1)-subset of the
vertices (a face) of an n-simplex (an element of Xn ) is also a k-simplex (an element of Xk )
Proposition 6.11. Every simplicial complex K determines an abstract simplicial complex aK.
Proof. Take the vertices of K as X0 , and take each Xn to be the collection of (n+1)-element
subsets of K that determine n-simplices. The requirement of subsets being abstract sim-
plices is satisfied by the corresponding requirement on simplicial complexes that faces
are also simplexes.

35
6.3 Ordered simplicial complexes 6 SIMPLICIAL COMPLEXES

Example 6.12. Consider an octahedron in R3 with only the top four faces filled in. Num-
bering the vertices from top to bottom and passing to its abstract simplicial complex, we
have:

X0 = {x1 , . . . , x6 }
X1 = {{x1 , x2 }, {x1 , x3 }, {x1 , x4 }, {x1 , x5 },
{x2 , x3 }, {x3 , x4 }, {x4 , x5 }, {x5 , x2 },
{x2 , x6 }, {x3 , x6 }, {x4 , x6 }, {x5 , x6 }}
X2 = {{x1 , x2 , x3 }, {x1 , x3 , x4 }, {x1 , x4 , x5 }, {x1 , x5 , x2 }}.

In this way we can see intuitive geometric relations encoded abstractly as subsets of the
vertex set.
Remark. In the case of finitely many simplices, one can go in reverse. That is, one can
pass from an abstract simplicial complex K to a simplicial complex gK by bijecting from
X0 to any geometrically independent subset of some RN (this induces a bijection on each
Xn ). Then agK is isomorphic to K in the sense of abstract simplicial complexes (there is a
bijection that preserves the relations between each’s simplex sets Xi ).
Definition 6.13. A map of abstract simplicial complexes f : K → L is a map K0 → L0
where f (σ ) is a simplex in L for all simplices σ ∈ K (i.e., σ is a subset of K0 and thus f (σ )
is induced by the map on K0 ; it should be an element of some Li ).
Definition 6.14. The category of abstract simplicial complexes, AbsSCxs has abstract
simplicial complexes as its objects and maps of abstract simplicial complexes as its mor-
phisms.

6.3 Ordered simplicial complexes

Definition 6.15. We have the following types of binary relations on a set S (i.e., subsets
of S × S):
• A preorder is a binary relation that is transitive and reflexive.

• A partial order is a preorder that is antisymmetric (x ≤ y and y ≤ x implies x = y).

• A total order is a partial order that is total (for all x, y, at least one of x ≤ y or y ≤ x
must hold).
Definition 6.16. The category of posets, Posets, has sets equipped with partial orders
(called posets) as its objects, and set maps satisfying

a ≤ b =⇒ f (a) ≤ f (b)

(called order-preserving maps) as its morphisms.

36
6.4 Simplicial approximation 6 SIMPLICIAL COMPLEXES

Definition 6.17. An ordered simplicial complex is an abstract simplicial complex equipped

with a partial ordering on X0 that restricts to a total order on each simplex. One can now
write n-simplices (x0 , . . . , xn ) as having a total order x0 ≤ · · · ≤ xn

Example 6.18. Consider the simplicial complex (of 0 and 1-simplices) given by the tree
structure of PhD advisors and advisees. There is a natural partial ordering on the vertices
as induced by the tree hierarchy. This ordering restricts to a total order on each of the
1-simplices (namely, advisor ≥ advisee).

Definition 6.19. The category of ordered simplicial complexes, OrdSCxs has ordered
simplicial complexes as its objects and maps of abstract simplicial complexes that are
order-preserving on the poset of vertices.

Example 6.20. There is a forgetful functor

OrdSCxs → AbsSCxs

which “forgets” the partial ordering. For example, the two ordered simplicial complexes
with simplices

X0 = {a, b, c},
X1 = {a, b}, {b, c}, {a, c}
X2 = {{a, b, c}}

and orderings a < b < c and c < b < a respectively (the strict < simply indicates that a, b, c
are in fact distinct), are the same object as abstract simplicial complexes.

Example 6.21. There is a functor

K : Posets → OrdSCxs

which takes the poset as the underlying poset for the ordered simplicial complex, and
then takes all finite totally ordered subsets of the poset as simplices. Intuitively, K takes
the “maximal” set of simplices described by the partial order.

6.4 Simplicial approximation

Definition 6.22. Let X be a topological space and K a simplicial complex. Continuous
maps f , g : X → |K| are simplicially close if for all x ∈ X we have f (x), g(x) are both in the
closure of some simplex σx ⊆ |L|.

Proposition 6.23. If f , g are simplicially close, then f ' g.

37
6.4 Simplicial approximation 6 SIMPLICIAL COMPLEXES

Proof. Take the homotopy

h : X × I → |K| ⊆ RN , h(x, t) = tf (x) + (1 − t)g(x).

By simplicial closeness, f (x), g(x) lie in the closure of some simplex σx . By convexity, h(x, t)
is contained in σx which ensures continuity. (In the finite case, we can also observe |K| has
the subspace topology in RN .)

Definition 6.24. A subdivision L of a simplicial complex K, written L ⊆ K, is where every

simplex of L is contained in a simplex of K, and every simplex of K is a finite union of
simplices of L. It follows that |L| = |K|.
1 Pn
Definition 6.25. The barycenter bσ of a geometric simplex σ is given by bσ = n+1 0 (vi ).
One constructs the barycentric subdivision K 0 of a simplicial complex K as the union
K = Ln , where the Ln are defined as follows, starting with L0 = K 0 :
S

• Let Ln−1 be a subdivision of the (n − 1)-skeleton K n−1 .

• For a n-simplex σ in K, take its boundary. This corresponds to an ordered subcomplex

Lσ of Ln−1 .

• We define
[
Ln = Ln−1 ∪ bσ ∗ Lσ .
σ ∈Kn

where the ∗ is the cone operation, where for every simplex τ in Lσ we add an additional
simplex τ ∪ {bσ }.
In this way we get an ordered simplicial complex with vertices {bσ }, where the ordering is
given by bσ ≤ bτ if σ ⊆ τ.
Definition 6.26. The barycentric subdivision procedure gives the functor

X : AbsSCxs → Posets

where K goes to the set of barycenters {bσ } with the partial order described above. The
ordered simplicial complex K 0 is thus the result of the composition

KX : AbsSCxs → OrdSCxs.

Example 6.27. To summarize, we have

|··· |
Top o OrdSCs
9
K
(forgetful)

Posets o
X
AbsSCs

38
 7 CHAIN COMPLEXES AND SIMPLICIAL HOMOLOGY

where the triangle does not commute; following the arrows around from AbsSCxs will
take you from a simplicial complex K to its barycentric subdivision K 0 (with the ordering
forgotten).
Example 6.28. Consider a triangle K (as a simplicial complex). Constructing the barycen-
tric subdivision proceeds as follows:
• We start with three points (the vertices) L0 = {b0 , b1 , b2 }.
• For the original 1-simplices {bi , bj }, we take the barycenter (the midpoint), place a
vertex bij there, and thus get L1 by adding simplices {bi , bij }, {bij , bj }. Hence

L1 = {b0 , b1 , b2 , {b0 , b01 }, {b01 , b1 }, {b0 , b02 }, {b02 , b2 }, {b1 , b12 }, {b12 , b2 }}

• For the original 2-simplex {b0 , b1 , b2 }, we take its barycenter (the centroid), place a
vertex b012 there, which induces the creation of 6 new triangles (and associated edges)
via the cone operation b012 ∗ L1 (since the boundary of the original 2-simplex gives
L1 ⊆ L1 itself).
We are done after the 2-simplex, so K 0 = L2 = L1 ∪ (b012 ∗ L1 ). The barycentric subdivision
K 0 has the natural partial ordering given by b012 ≥ b01 , b12 , b02 and bij ≥ bi , bj . One can see
that this partial order restricts to a total order on each simplex K 0 .
Definition 6.29. The n-th barycentric subdivision, K (n) of K is given by repeated barycen-
tric subdivision; i.e., K (1) = K 0 and K (i) = (K (i−1) )0 .
Theorem 6.30 (Simplicial approximation theorem). For a continuous map f : |K| → |L|
where K and L are simplicial complexes, there exists a number n and a simplicial map g :
K (n) → L such that |g| ' f .
Proof. Not too difficult, but our entire treatment has been too informal to support it. See
[Finite Book, §4].

The point is that if you have any continuous map between two spaces that are geo-
metric realizations of (possibly infinite) simplicial complexes, the map is homotopic to a
simplicial map (albeit now from a more complicated complex K (n) ). Simplicial maps are
nice and piecewise, and we can now approximate many continuous maps with them.
Remark. Historically, the foundations of algebraic topology were established using sim-
plicial complexes. Once we encounter CW complexes however, we will see that that they
are an even more natural setting, especially for homotopy theory and its generalizations
(due to Quillen; see Riehl’s Categorical Homotopy Theory).

7 Chain complexes and simplicial homology

Remark. This section is based on Chapter 12 of Concise, from class, etc.

39
7.1 Simplicial chain complexes
7 CHAIN COMPLEXES AND SIMPLICIAL HOMOLOGY

7.1 Simplicial chain complexes

Definition 7.1. A chain complex (M∗ , ∂∗ ) over R is a sequence of R-modules Mi

∂i+2
··· / Mi+1 ∂i+1 / Mi ∂i
/ Mi−1 ∂i−1 / · · ·

connected by boundary operators ∂i : Mi → Mi−1 , which are homomorphisms that satisfy

∂i ◦ ∂i+1 = 0 for all i.

Notation 7.2. When indices are obvious, one might drop the indices. For example, the
requirement on boundary operators is often simply expressed as ∂2 = 0.

Example 7.3. The chain complex C∗ (K) induced by an ordered simplicial complex K is
a chain complex over Z, given by the sequence of free abelian groups on the n-simplices
of K, i.e.,
nX o
Cn (K) = aj σj | aj ∈ Z, σj is an n-simplex of K ,

where the boundary maps are given by

n
X
∂(σ ) = ∂([x0 , . . . , xn ]) = (−1)i [x0 , . . . , x̂i , . . . , xn ]
i=0

(where x̂i means to remove the i-th component). One can verify that ∂i ◦ ∂i+1 = 0.

Example 7.4. Let Kn denote the set of n-simplices of K. Then in the chain complex C∗ (K),
we have Cn (K) = Z[Kn ]. We can think of Cn as the n-th chain complex functor

Cn : AbsSCxs → Ab, K 7→ Z[Kn ]

where the morphisms are transformed from f : S → T with

X  X
Cn f : Z[Sn ] → Z[Tn ], Cn f ai si = ai f (si ).

7.2 Simplicial homology groups

Definition 7.5. The n-th homology group of a chain complex is given by

Hn := ker(∂n )/im(∂n+1 )

Definition 7.6. The n-th simplicial homology group (over Z) of K, written Hn (K; Z), is
the n-th homology group of the induced chain complex C∗ (K).

40
7.3 Maps of chain complexes7 CHAIN COMPLEXES AND SIMPLICIAL HOMOLOGY

Example 7.7. The n-th simplicial homology group is a functor

Hn : OrdSCxs → Ab.
When we develop singular homology, we will see that this Hn factors through a functor
Top → Ab.
Exercise 7.8. Consider the standard n-simplex ∆sn . Compute that H∗ (∆sn ) = H0 (∆sn )  Z.
Exercise 7.9. Hq (∂∆sn ; Z) is isomorphic to Z at q = 0 and q = n − 1, and is 0 elsewise.
Notation 7.10. One sometimes writes Zn := ker(∂n ) for the set of cycles and write Bn :=
im(∂n+1 ) for the set of boundaries. This leads to the phrase that “homology is cycles mod
boundaries.” The terminology is inspired by the geometric intuition given by simplicial
homology.

7.3 Maps of chain complexes

Definition 7.11. A map of chain complexes f : (M∗ , ∂∗ ) → (M∗0 , ∂0∗ ) is a sequence of maps
fi : M∗ → M∗0 such that the following diagram commutes for all i:
fi
Mi / M0
i

∂i ∂0i
 fi−1 
Mi−1 / M0
i−1

That is, we must have fi−1 ◦ ∂i = ∂0i ◦ fi .

Example 7.12. We have the inclusion of simplicial complexes
∂∆sn → ∆sn .
This induces a map of (simplicial) chain complexes ι which is the identity for i < n. Con-
sider what changes: in the ∆sn case, we now have a single n-simplex. We know that
the space ∆sn is contractible, so it has the homology of a point (0 everywhere except for
H0  Z). However, if Cn (∆sn )  Z and Cn (∂∆sn ) = 0 as observed, then im(∂n ) is empty (vs.
being Z) and hence Hi (∂∆sn )  Z for i = 0, n − 1, while remaining 0 elsewhere.
Exercise 7.13. Work out all the details of the inclusion of the boundary of the 2-simplex into
the 2-simplex.
Definition 7.14. The induced map on homology f∗ is the set of maps
f∗ i : Hi (M∗ , ∂∗ ) → Hi (M∗0 , ∂0∗ ),
where the notation Hi (· · · ) here refers to the i-th (abstract) homology group induced by
the respective chain complex, and where f∗ i is defined in the obvious way (on equivalence
classes in ker(∂n )/im(∂n+1 ); one verifies the map is well-defined).

41
7.3 Maps of chain complexes7 CHAIN COMPLEXES AND SIMPLICIAL HOMOLOGY

Remark. When do two maps of chain complexes induce the same map on homology
(groups)? The answer to ths will turn out to be the notion of homotopy between maps
of chain complexes. This is analogous to homotopic maps of spaces inducing the same
map on homotopy (groups)!

Definition 7.15. A homotopy between maps of chain complexes s : f ' g where f , g :

0
(C∗ , ∂∗ ) → (C∗0 , ∂0∗ ), consists of homomorphisms of abelian groups sn : Cn → Cn+1 that sat-
isfy

∂0 ◦ s + s ◦ ∂ = f − g

for all indices. This has the property that if f ' g, then f∗ = g∗ : H∗ (C∗ ) → H∗ (C∗0 ). (in-
tuitively, f (x) − g(x) “represent the same boundary” when (ds + sd)(x); see [Link]
[Link]/wiki/Homotopy_category_of_chain_complexes for explanation and a
diagram).

We will now describe an example of homotopic maps of chain complexes via the
cone construction.

Definition 7.16. Given an ordered simplicial complex K with geometric realization X

(via passage through SCxs), we define the cone CX as the quotient

CX = (X × I)/(X × {1})

(imagine taking the cylinder X × I and then identifying one of the ends to be a single
point). The cone of an ordered simplicial complex is the ordered simplicial complex CK
(with |CK| = X) is given by taking:

• The vertex set is V (K) plus the cone vertex x which is greater than all elements of V (K)

• For every n-simplex σ , we also create an (n + 1)-simplex given by σ ∪ {x}.

Example 7.17. ∆sn  C∆sn−1 .

Exercise 7.18. The cone is a contractible space (homotopy equivalent to a point).

Example 7.19. We have Cn (CK) as the free abelian group with basis

Kn t {{x} ∪ σ | σ ∈ Kn−1 }.

We will show that id,  : C∗ (CK) → C∗ (CK) are chain homotopic maps, where

n : Cn (CK) → 0 for n , 0
0 : C0 (CK) → Z (generated by 0 (x))

42
7.4 Tensor products of chain7complexes
CHAIN COMPLEXES AND SIMPLICIAL HOMOLOGY

Our chain homotopy is given by

sn : Cn (CK) → Cn+1 (CK), s(σ ) = (x, σ ), s(x, τ) = 0,

(i.e., if a simplex contains the cone vertex, it goes to 0; if it does not, take it to the simplex
with the cone vertex adjoined).
Now observe that

∂s(σ ) = ∂(x, σ ) = σ − (x, ∂σ ), s∂(σ ) = (x, ∂σ ), ∂s(x) + s∂(x) = 0 + 0 = 0.

Altogether, we conclude

∂s + s∂ = id − .

where the first two observations deal with the case n > 0, and the last observation deals
with n = 0.

7.4 Tensor products of chain complexes

Definition 7.20. The tensor product of chain complexes (M∗ , ∂∗ ), (N∗ , ∂0∗ ) over R is the
chain complex (M∗ ⊗ N∗ , d∗ ) given by the sequence of R-modules
M
(M∗ ⊗R N∗ )n = Mi ⊗R Nj ,
i+j=n

where the n-th differential is

d(x ⊗ y) = ∂(x) ⊗ y + (−1)i x ⊗ ∂0 (y).

for x ∈ Mi , y ∈ Nj .

Exercise 7.21. Verify that d 2 = 0.

Using tensor products, we can motivate the earlier definition of homotopy for chain
complex maps (namely, the ds + sd = f − g requirement) by analogy to the definition of
homotopy for topological spaces.
Example 7.22. Consider I = ∆s1 , which we will also use to denote its chain complex. Let s
be a chain homotopy between maps f , g : M∗ → N∗ . There is a chain map

h : M∗ ⊗ I → N∗ .

such that h(x ⊗ [1]) = f (x) and h(x ⊗ [0]) = g(x).

Exercise 7.23. Let s(x) = (−1)deg x h(x ⊗ [1]). Show the existence of h is equivalent to the
ds + sd = f − g requirement.

43
 9 CW COMPLEXES

Remark. For simplicial (and CW complexes), there is an isomorphism C∗ (X × Y )  C∗ (X) 

C∗ (Y ). Then we can view

C∗ : OrdSCxs → ChCxs

as a functor that preserves the notion of homotopy in the respective categories, via

h : X ×I → Y → C∗ h : C∗ (X × I) = C∗ (X) ⊗ C∗ (I) → C∗ (Y ).

8 [TODO] Simplicial objects

Remark. Content + exercises + etc. for this is scattered across lectures, heaven’t organized
them. Summary for now, see [Finite, §10] or May’s Simplicial Objects for further details.

8.1 Summary
• Degenerate complexes K + (simplices can have repeated vertices); simplicial complexes
as special case of degenerate complexes (which we can think of as simplicial sets)

• Can define differentials and thus C∗+ (K) complexes; this time, C∗+ is itself a functor

• K and K + give same homology: [Link]

pdf

• Defined functor from spaces to simplicial sets

• We now have

Spaces → SimplSets → SimplAb → ChComp → GradedAbGroups

• We also have SimplSets → Spaces by geometric realization.

• (In later lectures: introduced the simplex category and (implicitly) simplicial objects

9 CW complexes
Remark. The following treatment is largely from class, which May admits is a sampling
from Concise Course. Some of the definitional exposition is also from [Hatcher, Chapter
0].

44
9.1 Motivation and definition 9 CW COMPLEXES

9.1 Motivation and definition

The concept of the CW complex was invented by J.H.C. Whitehead. As spaces, they
are more general than simplicial complexes, while still having the nice property that weak
equivalence of CW complexes is homotopy equivalence (Whitehead’s theorem). Further-
more, every space is weakly equivalent to a CW complex (CW approximation theorem).
As to what a CW complex is, they are products of a process that is somewhat reverse
to giving a space a simplicial decomposition. Consider the simplicial complex described
by a tetrahedron. Instead, imagine starting with the 0-simplices (the vertices), between
which we introduce 1-simplices (the edges), along which we introduce 2-simplices (the
faces). In a sense, CW complexes construct a space from the bottom-up, as opposed to
how simplicial complexes were induced by some top-down “saturation”.

Definition 9.1. A CW complex is space X that can be constructed in the following man-
ner: Let X 0 be a set with the discrete topology (we call its elements the vertices). The
space X is then
∞
[
X= Xn
n=0

where the sequence X 0 ⊆ X 1 ⊆ · · · ⊆ X n is defined inductively as follows:

• Given X n−1 , take a set of attaching maps (possibly empty)

jα : S n−1 → X n−1 .

• Along each of these maps, we attach a copy of D n (identifying the boundary of each
D n with each image jα (S n−1 )).

• We define the n-skeleton X n to be the pushout in the following diagram.

F
α jα
S n−1 / X n−1
F
α _

F  n

/ Xn
αD

• That is, it is the quotient space given by the disjoint union X n−1 t α D n under the
F
identification x ∼ jα (x) for every x ∈ ∂Dαn . Intuitively, the boundary of the α-th ball is
identified with the image of the α-th sphere in X n−1 .

• The n-cells of X are ultimately the images of each D n → X n ⊆ X.

45
9.2 Operations on CW complexes 9 CW COMPLEXES

The process may stop at a finite stage (i.e., there are no more attaching maps after some
m < ∞ and thus X = X m ). The largest n for which are n-cells is the dimension of the
CW complex. Otherwise, things might continue indefinitely where the union is given the
weak topology (i.e., a set U is open if and only if U ∩ X n is open for all n).
Example 9.2. Topologically, graphs are exactly 1-dimensional CW complexes (when we
say graph, multiple edges between vertices and loops at a vertex are both allowed). The
points on the graph are X 0 , and the edges are intervals (copies of D 1 ) glued along pairs of
vertices in X 0 (i.e., images of S 0 ) to give X = X 1 .
Example 9.3. The n-sphere S n can be viewed as a CW complex with one 0-cell and one
n-cell. Take X 0 = {?}, then X n−1 = · · · = X 0 , then a single attaching map j : S n−1 → X n−1 as
the constant map to ? to give

X n = {?} t D n / ∼  D n /∂D n ,

since ∼ identifies ∂D n ∼ ? and thus every boundary point to each other. Since S n 
D n /∂D n , we conclude that S n  X = X n .
Example 9.4. (TODO: describe the real projective space CW complex) The point is that
a simplicial triangulation of RP n is annoying (see [Link]
50382/how-to-triangulate-real-projective-spaces-as-simplicial-complexes-in-
mathematic) , while our CW complex construction is rather nice.

9.2 Operations on CW complexes

Example 9.5. Geometric realizations of simplicial complexes are a specialization of CW
complexes:

∂∆tn / K n−1

 
∆tn / Kn

Definition 9.6. The product of CW complexes X and Y is given where the n-skeleton is
[
(X × Y )n = Xp × Y q.
p+q=n

We have
p q p−1 q
Dj × Sk ∪ Sj × Db ⊆ D p × D q
I p × ∂I q ∪ ∂I p−1 × I q ⊆ I p × I q

where I is the CW complex with vertices 0, 1 and a 1-cell glued to them.

46
9.3 [TODO] Cellular approximation, etc. 10 TOPICS COVERED

Definition 9.7. A cellular map between CW complexes is a map that takes n-skeletons
to n-skeletons; i.e., f : X → Y satisfies f (X n ) ⊆ Y n for all n.

Definition 9.8. A cellular homotopy is a homotopy h : X × I → Y which is a cellular map.

9.3 [TODO] Cellular approximation, etc.

Theorem 9.9 (Cellular approximation). (TODO: haven’t done the proof yet, just alluded
to it; requires notion of relative CW complexes, not covered until July 21)

Proof. (TODO: the homotopy extension lifting lemma (HELP) comes in that allows
homotopies to be done cell by cell, then inductively on dimension.)

Corollary 9.10. Any f : X → Y is homotopic to a cellular map. If f is cellular, it induces a

map C# (X) → C# (Y ) of cellular chain complexes. If h : f ' g is any homotopy where f , g are
cellular, then f ' g via a cellular homotopy.

Remark. Intuitively, it suffices to consider cellular maps to study the up-to-homotopy

space [X, Y ].
(TODO: lots of other stuff was said, left out for now)

Part II
Week 4 onwards
Remark. I do have actual day-by-day notes on these, but they’re half TODOs and things I
didn’t understand. Leaving them out for now and just summarizing what was discussed.
Will get to this by the time Alg. Top. lectures resume (if not, then end of August; if not,
then probably never).

10 Topics covered
Remark. References:

• May’s Concise Course: [Link]

pdf

• May’s Finite Book: [Link]

pdf

47
10.1 July 11 (May) 10 TOPICS COVERED

10.1 July 11 (May)

• Quotienting by contractible space (e.g., maximal tree of graph) preserves homology

• Defined cofibrations and fibrations, and the corresponding homotopy extension prop-
erty (HEP) and homotopy lifting property (HLP). Mapping cylinder and co-cylinders.
See [Concise, §6-7].

• We also defined pullbacks (see Categorical Notions section).

• Fibrations generalize unique path lifting of covering spaces.

• Constructed a map that is 0 on homology/homotopy but not nullhomotopic (this one:

[Link]

10.2 July 12 (May)

• Idea of replacing arbitrary maps by cofibration/fibration up to homotopy

• Neighborhood deformation retracts (NDR) [Concise, §6]

• How to recognize fibrations

• The long exact sequence for cofibrations [Concise, §8.4]

10.3 July 13 (May)

• The dual long exact sequence for fibrations [Concise, §8.5]

• The homotopy group long exact sequence [Concise, §9]

• Cohomology, take 1 (via cochains and cochain complexes)

• “Homology is useless”; cohomology has a graded ring structure

• Cohomology, take 2 (via Eilenberg-MacLane spaces)

• “Deduced” the Eilenberg-Steenrod axioms

• Generalized cohomology theories as those satisfying all but dimension axioms

48
10.4 July 14 (Chan) 10 TOPICS COVERED

10.4 July 14 (Chan)

• The simplex category

• Geometric realization of a simplicial object

• The 2-sided bar construction ([Link]

• Classifying spaces of categories via functor B (or posets, or groups)

• Properties of B:

a) B is functorial (i.e., f : G → H induces a map Bf : BG → BH). ((TODO: offhand

comment about homotopy invariance of B with respect to G-equivariance), e.g.,

': ∗ × Gn × ∗ → ∗ × H n × ∗)

b) EG is contractible
c) G ' ΩBG (if G is good)
d) EG → BG is the universal principle G-bundle
e) B(G × H)  BG × BH
f) If G is abelian then BG is a group

• (Proved or at least discussed these by various means)

• One of those proofs used the five lemma from homological algebra ([Link]
[Link]/wiki/Five_lemma)

10.5 July 15 (May)

• Pretty much this diagram:

B
+
TopGps / SimpSpaces / Spaces o T
CWCxs
O g T O 6
C#
N T S
'  (
TopoCat SimpSets / SimpAbGrps / ChCxs
7 O 7 h
N
H∗

Triples SmallCat
g OrdSimpCxs GrdAbGrps
6
K
(forgetful)

Posets o
X
AbsSimpCxs

49
10.6 July 18 (May) 10 TOPICS COVERED

• The nerve functor N and its properties

• The category Triples

• Idea: translate differential topology problems in terms of vector bundles, and then use
homotopy theory (via classifying spaces)

• Defined adjoint functors (see Categorical notions)

• Diagram should commute everywhere? (except barycentric triangle at the bottom?)

10.6 July 18 (May)

• Focused on this part:

SimpSpaces
T / Spaces o CWCxs
g O 8
T S
 T
SimpSets

• Discussed the S, T maps as adjoint pairs of functors

• Mentioned Whitehead’s theorem

• Described the T functor on simplicial sets in detail, e.g., [Finite Book, §10.6]

• Singular chain complexes and cell chain complexes as equivalent

10.7 July 19 (May)

• Diagram for the day

TopGrp SimpSpaces
T / Spaces o CWCxs
7 g O 8
B
T S
& N  T
TopoCat SimpSets
8 g 7

N
Triples Categories

• Initial and terminal objects

• Classifying space of a category BC = |N C|

50
10.8 July 20 (May) 10 TOPICS COVERED

• Topological categories as categories whose hom-sets are topological spaces

• E as a related functor to B

• p : EG → BG as universal principal bundle; every G principal bundle over para-

compact space is a pullback of this: [Link]
space

10.8 July 20 (May)

• Nerve and T functors preserve products for nice spaces

• Simplices as increasing sums ([Link]

coordinates)

• Proved that S, T are homo topy preserving functors

• Idea: do all of homotopy theory in simplicial sets

• By passing to chains, identical homotopy induces identical homology; done without

appeal to topology!

• Constructed the chaotic category ([Link]

for X: X̃. Turns out G̃  EG

10.9 July 21 (May)

• n-equivalences

• Relative CW complexes [Concise, §10.1]

• HELP and Whitehead theorem [Concise, §10.3]

i0
A / Sn × I o A
i1

g
{ h 
o
?Z c e Y_
f
 i0 h̃  g̃ 
X / CS n × I o X
i1

• Power of Whitehead’s theorem (and CW complexes): n-equivalence of CW complexes

with dimension less than n is homotopy equivalence

51