Category Theory

Teja Prabhu

1
Contents

2
1 Categories, Functors, and Natural Transforma-
tions

3
1.1 Functors
Exercise 1
Show how each of the following constructions can be regarded as a func-
tor:
(a) The field of quotients of an integral domain
Proof
Let IntDom be a category with integral domains as objects and injec-
tive homomorphisms as arrows; and F ield be the category of fields, with
arrows obeying ring homomorphism rules. Define:

• To , the object function sends each integral domain to its corresponding

field of quotients.

• Tf , the arrow function sends each morphism f in IntDom to the in-

duced morphism Tf (f ) : a/b 7−→ f (a)/f (b).

Since, arrows in IntDom are injective, b 6= 0 =⇒ f (b) 6= 0. And:

a/b ∼ a0 /b0 =⇒ ab0 = a0 b =⇒ f (a)f (b0 ) = f (a0 )f (b) =⇒ f (a)/f (b) ∼
f (a )/f (b0 ). Therefore, Tf (f ) is well-defined. The functor rules Tf (f ◦g) =
0

Tf (f ) ◦ Tf (g) and Tf (ido ) = idTo (o) are satisfied; hence, the construction is
a functor.
(b) The Lie Algebra of a Lie Group
skipped; don’t know Differential Geometry
Exercise 2
Show that functors 1 → C, 2 → C, and 3 → C correspond respectively
to objects, arrows, and composable pairs of arrows in C.
Proof
Trivial.
Exercise 3
(a) A functor between two preorders is a function which is monotonic
Proof
A preorder is a category P in which, given objects p and p0 , there is at
most one arrow p → p0 .
Monotonicity condition: p ≤ q =⇒ T (p) ≤ T (q). A functor preserves the
order of the domain and codomain. Hence a functor between two preorders
can be interpreted as a monotonic function.

4
 (b) A functor between two groups is a morphism of groups
Proof
A group is a one object category. So there is only one object function
possible. Furthermore, the functor laws correspond exactly to the group
homomorphism laws, i.e., T (f ◦ g) = T (f ) ◦ T (g). This works since the
arrows of the group category correspond to the elements of the group. We
don’t even need the functor law regarding identities, since every group has
an inverse, i.e., in this case the functor is between two groupoids.
(c) If G is a group, a functor G → Set is a permutation representation
of G, while G → M atrK is a matrix representation of G
Proof
Since G is a one object category, the object function will map it to some
set S in Set. The arrow function will:
• send the identity e of the group to the identity function on S
• since every element in a group has an inverse, every function on S
should have an inverse too. If a · b = e, in G, then since T (a ◦ b) =
T (e) = idS = T (a) ◦ T (b). So every arrow T (a) will have an inverse.
So the elements of the group map to bijections from S to itself.
From the previous two points, we have by definition that a functor from
G to Set is a permutation representation of G.
A functor from G to M atrK :
• The object function will map the single object to some positive integer
n.
• The arrow function will map elements of the group to some n × n ma-
trix, such that functor laws are satisfied. This means that it will only
map to non-singular matrices, which obey the group homomorphism
laws with respect to matrix multiplication.
Since, matrices are nothing but linear transformations over vector spaces,
functors from G to M atrK will correspond to matrix representations of G.
Exercise 4
Prove that there is no functor Grp → Ab sending each group G to its
center.
Proof
Consider the symmetric groups S2 and S3 . Let:

5
 • f : S2 → S3 be the inclusion homomorphism. Since in S3 if we keep 3
fixed, we’ll find that subgroup of S3 to be isomorphic to S2 .

• g : S3 → S2 , such that the kernel of g is A3 .

• The center of S2 is itself.

• The center of S3 is the trivial subgroup.

Then g ◦ f = idS2 . Let To be any object function which sends each group
to its center. Then, Tf (f ) must be the zero map, as the center of S3 is the
trivial subgroup. Therefore, for all maps g from Z(S3 ) to Z(S2 ), Tf (g) must
be the zero map. Hence, Tf (g ◦ f ) will be the zero homomorphism. But
functors should take identities to identities. So, we cannot construct any
arrow function which will obey the functor rules.
Exercise 5
Find two different functors T : Grp → Grp with object function T (G) =
G the identity for every group G.
Proof
One is the identity functor. For the other functor pick a group G and
f g
an automorphism φ on G. G0 − →G→ − G00 . Replace all such f ’s with φ ◦ f
−1
and all such g’s with g ◦ φ , and leave all other arrows in the category
unchanged. This will form another functor.

6
1.2 Natural Transformations
Exercise 1
Let S be a fixed set, and X S the set of all functions h : S → X. Show
that X 7→ X S is the object function of a functor Set → Set, and that
.
evaluation ex : X S × S → − X, defined by e(h, s) = h(s), the value of the
function h at s ∈ S, is a natural transformation.
Proof
ex forms a natural transformation among the identity functor and the
functor constructed below:

• object function: X 7→ X S × S

• arrow function: f 7→ (f ◦ g) × id for all g ∈ X S

Exercise 2
If H is a fixed group, show that G 7→ H × G defines a functor H × − :
Grp → Grp, and that each morphism f : H → K of groups defines a natural
.
transformation H × − → − K × −.
Proof
Let f : H → K be a morphism. Let φ : G → G0 be a morphism
between two groups. Then, (h, g) 7→ (h, φ(g)) in the H× — functor; and
(k, g) 7→ (k, φ(g)) in the K× — functor.
Now if f is any morphism from H to K, then
(h, g) 7→ (h, φ(g)) 7→ (f (h), φ(g)) and (h, g) 7→ (f (h), g) 7→ (f (h), φ(g))
map to the same element, which implies that the diagram is commutative.
Hence, it is a natural transformation.
Exercise 3
If B and C are groups (regarded as categories with one object each) and
S, T : B → C are functors (homomorphisms or groups), show that there is
.
a natural transformation S → − T if and only if S and T are conjugate; i.e.,
if and only if there is an element h ∈ C with T g = h(Sg)h−1 for all g ∈ B.
Proof
.
Let’s assume there is a natural transformation from S → − T . Since, there
is only one object in each of the categories, there will only be one arrow in
the natural transformation that we have to think about – let’s say h ∈ C.
From the natural transformation commutative condition we have: h ◦
Sg = T g ◦ h for all arrows g ∈ B.

7
 =⇒ T g = h(Sg)h−1 for each g ∈ B, we can do this since the category
is actually a groupoid.
=⇒ h and g are conjugate.
S and T are conjuagte =⇒ ∃h.T g = h(Sg)h−1
Since, h ∈ C, h is an arrow of C, so we can pick this to be the arrow of
the natural transformation (we can also pick h−1 ).
Exercise 4
For functors S, T : C → P where C is a category and P a preorder, show
.
that there is a natural transformation S →
− T (which is then unique) if and
only if Sc ≤ T c for every object c ∈ C.
Proof
For all c ∈ C, there is the identity arrow. So, if we draw the commutative
diagram with c0 = c, then for every object we have Sc ≤ T c if there is a
natural transformation from S to T .
Conversely, if Sc ≤ T c for every c, then there will be exactly one arrow
from Sc → T c for all c. Since in a preorder between any two objects there
can be at most one arrow, the diagram will commute, and the arrow defined
above will be the natural transformation.
Exercise 5
.
Show that every natural transformation τ : S → − T defines a function
0
(also called τ ) which sends each arrow f : c → c of C to an arrow τ f : Sc →
T c0 of B in such a way that T g ◦ τ f = τ (gf ) = τ g ◦ Sf for each composable
pair < g, f >. Conversely, show that every such function τ comes from a
unique natural transformation with τc = τ (1c ) (This gives an ”arrows only”
description of a natural transformation.)
Proof
Let τ (f ) = T (f ) ◦ τc = τc0 ◦ S(f ), then the relations are satisfied because
the diagram is commutative.
1 f 1 0
Conversely, let there be such a τ . Then, consider c −→c
c−→ c0 −−→
c
c0 .
τ (f ◦ 1c ) = τ (1c0 ◦ f )
=⇒ T (f ) ◦ τc = τc0 ◦ S(f ), which is the definition of a natural trans-
formation. The transformation is unique because τ is a function and τ (1c )
and τ (10c ) have only one value each.
Exercise 6
Let F be a field. Show that the category of all finite-dimensional vector
spaces over F (with morphisms all linear transformations) is equivalent to

8
the category M atrF described in 2.
Proof

• object function: V n 7→ n

• arrow function: takes L : V n → V m 7→ Am×n , where the matrix is the

transformation matrix.

We can pick the natural transformation to be V n 7→ n. Since this is

a bijection, there is a natural equivalence between the identity functor on
the category of vector spaces and M atrF . Hence, both these categories are
equivalent.

9
1.3 Monics, Epis, and Zeros
Exercise 1
Find a category with an arrow which is both epi and monic, but not
invertible.
Proof
Let R be a one object category with arrows all differentiable functions
from R to itself. Then, consider x 7→ x3 , which is a bijective function, so it is
both monic and epi, but is not invertible since x 7→ x1/3 is not differentiable
at 0.
Another example can be constructed from the inclusion function i : Q →
R. i is clearly not invertible. We know:

• i is continuous

• Each point in R is the limit point of some sequence of points in Q,

from the sequence lemma (since the closure of Q is R).

• And we know that xn 7→ x =⇒ f (xn ) 7→ f (x) if f is continuous.

So, g1 ◦ i = g2 ◦ i =⇒ g1 = g2 for all possible g1 and g2 (as Hausdorff spaces

have unique limits).
Exercise 2
Prove that the composite of monics is monic, and likewise for epis.
Proof
g ◦ f is monic if g, f are monic. Let p and q be two functions whose
codomain is the same as the domain of g ◦ f . Then,
(g ◦ f ) ◦ p = (g ◦ f ) ◦ q
=⇒ g ◦ (f ◦ p) = g ◦ (f ◦ q)
=⇒ f ◦ p = g ◦ q (since g is monic)
=⇒ p = q (since f is monic)
=⇒ g ◦ f is monic.
g ◦ f is epi if g, f are epi. Let p and q be 2 functions whose domain is
the same as the codomain of g ◦ f . Then,
p ◦ (g ◦ f ) = q ◦ (g ◦ f )
=⇒ (p ◦ g) ◦ f = (q ◦ g) ◦ f
=⇒ p ◦ g = q ◦ g (since f is epi)
=⇒ p = q (since g is epi)
=⇒ g ◦ f is epi.

10
Exercise 3
If a composite g ◦ f is monic, so is f . Is this true of g?
Proof
Let’s say f is not monic. Then there are functions x and x0 not equal,
such that:
f ◦ x = f ◦ x0
=⇒ (g ◦ f ) ◦ x = (g ◦ f ) ◦ x0
=⇒ x = x0 (g ◦ f is monic)
which is a contradiction. So f is monic. It’s not necessarily true for g.
For example, in Set, g has to be injective only in the image of f , and outside
this set it may not be injective.
Exercise 4
Show that the inclusion Z → Q is epi in the category Rng.
Proof
i φ,ψ
Let Z → − Q −−→ R
In such a case, we must prove that if φ ◦ i = ψ ◦ i =⇒ φ = ψ. Let
p
q ∈ Q, then:
φ( pq ) 6= ψ( pq )
=⇒ φ( 1q ) 6= ψ( 1q ) (since φ and ψ will be the same on Z)
=⇒ φ(1) 6= ψ(1) (multiplying both sides by φ(q) = ψ(q)), which is a
contradiction.
Hence, φ = ψ, and the inclusion is epi in the category Rng.
Exercise 5
In Grp prove that every epi is surjective.
Proof
Let φ : G → H be an epi in Grp. Consider the order k of H/Im(φ):

• If k = 2, then we know that Im(φ) is a normal subgroup. Let π :

H → H/Im(φ) be the canonical map, i.e., h 7→ h + Im(φ); and let
z : H → H/Im(φ) be the zero map. Then, π ◦ φ = z ◦ φ. But, since φ
is epi, π = z =⇒ z is surjective =⇒ Im(φ) = H =⇒ k = 1, which
is a contradiction.

• If k > 2, then we can pick three unique cosets Im(φ), u + Im(φ), v +

Im(φ). Let P ermH be the group of all permutations of the set H.
Define a map λ ∈ P ermH, such that λ(xu) = xv and λ(xv) = xu for

11
 all x ∈ Im(f ) and equal to the identity function everywhere else. Let
ψ : H → P ermH send each h to left multiplication ψh by h, and let
ψh0 = λ−1 ψh λ. Then, ψφ = ψ 0 φ, but ψ 6= ψ 0 , contradicting φ is epi.

Hence, from the above two points k = 1 is the only valid value, which
implies that φ is surjective.
Exercise 6
In Set, show that all idempotents split.
Proof
Let f : b → b be idempotent. f 2 = f =⇒ f is the identity on Im(f ).
Define h be the identity on Im(f ) and g be f . Then f = hg and gh = 1.
Hence, all idempotents split in Set.
Exercise 7
An arrow f : a → b in a category C is regular when there exists an arrow
g : b → a such that f gf = f . Show that f is regular if it has either a left or
a right inverse, and prove that every arrow in Set with a 6= ∅ is regular.
Proof

• If it has a left inverse, then there is an r such that r ◦f = 1a . Choosing

g = r, we have f gf = f

• If it has a right inverse, then there is an r such that f ◦r = 1b . Choosing

g = r, we again have f gf = f

g:b→a

• g(b) = b if b ∈
/ Im(f )

• If b ∈ Im(f ), then pick any random preimage a, and let g(b) = a.

Then, f gf = f , and so every arrow in Set is regular when a 6= ∅.

Exercise 8
Consider the category with objects hX, e, ti, where X is a set, e ∈ X,
and t : X → X, and with arrows f : hX, e, ti → hX 0 , e0 , t0 i the functions f
on X to X 0 with f e = e0 and f t = t0 f . Prove that this category has an
initial object in which X is the set of natural numbers, e = 0, and t is the
successor function.
Proof

12
 An object s is initial in C if to each object a there is exactly one arrow
s → a. Let f : hN, 0, ++i → hX, e, ti and f ◦ succ = t ◦ f , where f : N → X.
We have to prove that there is exactly one such f .

• f (0) = e

• f (n + 1) = tn ◦ f for all n 6= 0

Therefore, f is defined for all of N from the last condition, and hence
f : hN, 0, ++i is an initial object.
Exercise 9
If the functor T : C → B is faithful and T f is monic. prove f monic.
Proof
A functor is faithful if T : Hom(a, b) → Hom(T (a), T (b)) is injective.
Let f ◦ g1 = f ◦ g2 and T (f ) is monic.
=⇒ T (f ) ◦ T (g1 ) = T (f ) ◦ T (g2 )
=⇒ T (g1 ) = T (g2 ) (from the monic condition)
=⇒ g1 = g2 , since T is faithful.

13
1.4 Foundations
Exercise 1
Given a universe U and a function f : I → b with domain I ∈ U and
Q value fi an element of U , for i ∈ I, prove that the usual cartesian
with every
product i fi is an element of U .
Proof Q
Refer to pg. 22 for properties of the universe. We have to show i fi ∈ U .
Q S
• i fi = {g : I → i fi | ∀i · g(i) ∈ fi }.

• Im(f ) is a set from the axiom of replacement.

• Im(f ) ⊂ U , since every fi ∈ U .

• Consider a function f 0 : I → Im(f ) formed by restricting the codomain

of f

• f 0 is surjective, and from (v), we know Im(f ) ∈ U .

S
• i fi ∈ U , from (iii).
S
• I × i fi ∈ U , from (ii)
S
• P(I × i fi ) ∈ U , from (iii)
Q S Q S
• i fi ⊂ I × i fi =⇒ i fi ∈ P(I × i fi )
Q
• Hence, i fi ∈ U , from (i).

Exercise 2
(a) Given a universe USand a function f : I → b with domain I ∈ U ,
show that the usual union i fi is a set of U .
Proof
We must assume b ⊂ U to prove this. We can employ the same reasoning
we used in the previous proof.
(b) Show that this one closure property of U may replace condition (v)
and the condition x ∈ U implies ∪x ∈ U in the definition of a universe.
Proof
I (a) implies (v)

14
 f : I → b, b ⊂ U and f is surjective. Then, from (a), we know that
∪b ∈ U .
=⇒ P(∪b) ∈ U
=⇒ P(P(∪b)) ∈ U
But, since b ∈ P(P(∪b)) ∈ U , we have b ∈ U , from (i).
II (a) implies (iii) b
Let b ∈ U . Consider the identity function i : b → U . Then x ∈ b ∈
U =⇒ x ∈ U , from S (i). So it is a valid function. Now Im(i) = b. From
(a), we know that Im(i) ∈ U , and so, ∪b ∈ U .

15
2 Constructions on Categories

16