3 Classification of sets

3.1 Maps between sets

A recurrent theme in mathematics is the classification of spaces by means of structure-
preserving maps between them.
A space is usually meant to be some set equipped with some structure, which is usually
some other set. We will define each instance of space precisely when we will need them. In
the case of sets considered themselves as spaces, there is no extra structure beyond the set
and hence, the structure may be taken to be the empty set.

Definition. Let A, B be sets. A map : A B is a relation such that for each a A

there exists exactly one b B such that (a, b).

The standard notation for a map is:

: A B
a 7 (a)

which is technically an abuse of notation since , being a relation of two variables, should
have two arguments and produce a truth value. However, once we agree that for each a A
there exists exactly one b B such that (a, b) is true, then for each a we can define (a)
to be precisely that unique b. It is sometimes useful to keep in mind that is actually a
relation.
Example 3.1. Let M be a set. The simplest example of a map is the identity map on M :

idM : M M
m 7 m.

The following is standard terminology for a map : A B:

the set A is called the domain of ;

the set B is called the target of ;

the set (A) im (A) := {(a) | a A} is called the image of A under .

Definition. A map : A B is said to be:

injective if a1 , a2 A : (a1 ) = (a2 ) a1 = a2 ;

surjective if im (A) = B;

bijective if it is both injective and surjective.

Definition. Two sets A and B are called (set-theoretic) isomorphic if there exists a bijection
: A B. In this case, we write A
=set B.

Remark 3.2. If there is any bijection A B then generally there are many.

1
 Bijections are the structure-preserving maps for sets. Intuitively, they pair up the
elements of A and B and a bijection between A and B exists only if A and B have the same
size. This is clear for finite sets, but it can also be extended to infinite sets.

Definition (Classification of sets). A set A is:

infinite if there exists a proper subset B A such that B

=set A. In particular, if A
is infinite, we further define A to be:

countably infinite if A
=set N;
uncountably infinite otherwise.

finite if it is not infinite. In this case, we have A

=set {1, 2, . . . , N } for some N N
and we say that the cardinality of A, denoted by |A|, is N .

Given two maps : A B and : B C, we can construct a third map, called the
composition of and , denoted by (read psi after phi), defined by:

: A C
a 7 ((a)).

This is often represented by drawing the following diagram

A
C

and by saying that the diagram commutes (although sometimes this is assumed even if
it is not explicitly stated). What this means is that every path in the diagram gives the
same result. This might seem notational overkill at this point, but later we will encounter
situations where we will have many maps, going from many places to many other places
and these diagrams greatly simplify the exposition.

Proposition 3.3. Composition of maps is associative.

Proof. Indeed, let : A B, : B C and : C D be maps. Then we have:

( ) : A D
a 7 (((a)))

and:

( ) : A D
a 7 (((a))).

Thus ( ) = ( ) .

2
 The operation of composition is necessary in order to defined inverses of maps.
Definition. Let : A B be a bijection. Then the inverse of , denoted 1 , is defined
(uniquely) by:
1 = idA
1 = idB .
Equivalently, we require the following diagram to commute:

idA A B idB

The inverse map is only defined for bijections. However, the following notion, which we will
often meet in topology, is defined for any map.
Definition. Let : A B be a map and let V B. Then we define the set:
preim (V ) := {a A | (a) V }
called the pre-image of V under .
Proposition 3.4. Let : A B be a map, let U, V B and C = {Cj | j J} P(B).
Then:
i) preim () = and preim (B) = A;

ii) preim (U \ V ) = preim (U ) \ preim (V );

S S T T
iii) preim ( C) = jJ preim (Cj ) and preim ( C) = jJ preim (Cj ).
Proof. i) By definition, we have:
preim (B) = {a A : (a) B} = A
and:
preim () = {a A : (a) } = .

ii) We have:
a preim (U \ V ) (a) U \ V
(a) U (a)
/V
a preim (U ) a
/ preim (V )
a preim (U ) \ preim (V )

iii) We have:
S S
a preim ( C) (a) C
W
jJ ((a) Cj )
W
jJ (a preim (Cj ))
S
a jJ preim (Cj )
Similarly, we get preim ( C) = jJ preim (Cj ).
T T

3
3.2 Equivalence relations
Definition. Let M be a set and let be a relation such that the following conditions are
satisfied:

i) reflexivity: m M : m m;

ii) symmetry: m, n M : m n n n;

iii) transitivity: m, n, p M : (m n n p) m p.

Then is called an equivalence relation on M .

Example 3.5. Consider the following wordy examples.

a) p q : p is of the same opinion as q. This relation is reflexive, symmetric and

transitive. Hence, it is an equivalence relation.

b) p q : p is a sibling of q. This relation is symmetric and transitive but not reflexive

and hence, it is not an equivalence relation.

c) p q : p is taller q. This relation is transitive, but neither reflexive nor symmetric

and hence, it is not an equivalence relation.

d) p q : p is in love with q. This relation is generally not reflexive. People dont like
themselves very much. It is certainly not normally symmetric, which is the basis of
much drama in literature. It is also not transitive, except in some French films.

Definition. Let be an equivalence relation on the set M . Then, for any m M , we

define the set:
[m] := {n M | m n}
called the equivalence class of m. Note that the condition m n is equivalent to n m
since is symmetric.

The following are two key properties of equivalence classes.

Proposition 3.6. Let be an equivalence relation on M . Then:

i) a [m] [a] = [m];

ii) either [m] = [n] or [m] [n] = .

Proof. i) Since a [m], we have a m. Let x [a]. Then x a and hence x m by

transitivity. Therefore x [m] and hence [a] [m]. Similarly, we have [m] [a] and
hence [a] = [m].

ii) Suppose that [m] [n] 6= . That is:

z : z [m] z [n].

Thus z m and z n and hence, by symmetry and transitivity, m n. This implies

that m [n] and hence that [m] = [n].

4
Definition. Let be an equivalence relation on M . Then we define the quotient set of M
by as:
M/ := {[m] | m M }.
This is indeed a set since [m] P(M ) and hence we can write more precisely:

M/ := {[m] P(M ) | m M }.

Then clearly M/ is a set by the power set axiom and the principle of restricted compre-
hension.
Remark 3.7. Due to the axiom of choice, there exists a complete system of representatives
for , i.e. a set R such that R
=set M/ .
Remark 3.8. Care must be taken when defining maps whose domain is a quotient set if one
uses representatives to define the map. In order for the map to be well-defined one needs
to show that the map is independent of the choice of representatives.
Example 3.9. Let M = Z and define by:

m n : n m 2Z.

It is easy to check that is indeed an equivalence relation. Moreover, we have:

[0] = [2] = [4] = = [2] = [4] =

and:
[1] = [3] = [5] = = [1] = [3] =
Thus we have: Z/ = {[0], [1]}. We wish to define an addition on Z/ by inheriting
the usual addition on Z. As a tentative definition we could have:

: Z/ Z/ Z/

being given by:

[a] [b] := [a + b].
However, we need to check that our definition does not depend on the choice of class
representatives, i.e. if [a] = [a0 ] and [b] = [b0 ], then we should have:

[a] [b] = [a0 ] [b0 ].

Indeed, [a] = [a0 ] and [b] = [b0 ] means a a0 2Z and b b0 2Z, i.e. a a0 = 2m and
b b0 = 2n for some m, n Z. We thus have:

[a0 + b0 ] = [a 2m + b 2n]
= [(a + b) 2(m + n)]
= [a + b],

where the last equality follows since:

(a + b) 2(m + n) (a + b) = 2(m + n) 2Z.

Therefore [a0 ] [b0 ] = [a] [b] and hence the operation is well-defined.

5
Example 3.10. As a counterexample, with the same set-up as in the previous example, let
us define an operation ? by:
a
[a] ? [b] := .
b
This is easily seen to be ill-defined since [1] = [3] and [2] = [4] but:
1 3
[1] ? [2] = 6= = [3] ? [4].
2 4
3.3 Construction of N, Z, Q and R
Recall that, invoking the axiom of infinity, we defined:

N := {0, 1, 2, 3, . . .},

where:
0 := , 1 := {}, 2 := {{}}, 3 := {{{}}}, ...
We would now like to define an addition operation on N by using the axioms of set theory.
We will need some preliminary definitions.

Definition. The successor map S on N is defined by:

S: N N
n 7 {n}.

Example 3.11. Consider S(2). Since 2 := {{}}, we have S(2) = {{{}}} =: 3. Therefore,
we have S(2) = 3 as we would have expected.
To make progress, we also need to define the predecessor map, which is only defined
on the set N := N \ {}.

Definition. The predecessor map P on N is defined by:

P : N N
n 7 m such that m n.

Example 3.12. We have P (2) = P ({{}}) = {} = 1.

Definition. Let n N. The n-th power of S, denoted S n , is defined recursively by:

S n := S S P (n) if n N
S 0 := idN .

We are now ready to define addition.

Definition. The addition operation on N is defined as a map:

+: N N N
(m, n) 7 m + n := S n (m).

6
Example 3.13. We have:
2 + 1 = S 1 (2) = S(2) = 3
and:
1 + 2 = S 2 (1) = S(S 1 (1)) = S(S(1)) = S(2) = 3.
Using this definition, it is possible to show that + is commutative and associative. The
neutral element of + is 0 since:

m + 0 = S 0 (m) = idN (m) = m

and:
0 + m = S m (0) = S P (m) (1) = S P (P (m)) (2) = = S 0 (m) = m.
Clearly, there exist no inverses for + in N, i.e. given m N (non-zero), there exist no n N
such that m + n = 0. This motivates the extension of the natural numbers to the integers.
In order to rigorously define Z, we need to define the following relation on N N.

Definition. Let be the relation on N N defined by:

(m, n) (p, q) : m + q = p + n.

It is easy to check that this is an equivalence relation as:

i) (m, n) (m, n) since m + n = m + n;

ii) (m, n) (p, q) (p, q) (m, n) since m + q = p + n p + n = m + q;

iii) ((m, n) (p, q) (p, q) (r, s)) (m, n) (r, s) since we have:

m + q = p + n p + s = r + q,

hence m + q + p + s = p + n + r + q, and thus m + s = r + n.

Definition. We define the set of integers by:

Z := (N N)/ .

The intuition behind this definition is that the pair (m, n) stands for m n. In other
words, we represent each integer by a pair of natural numbers whose (yet to be defined)
difference is precisely that integer. There are, of course, many ways to represent the same
integer with a pair of natural numbers in this way. For instance, the integer 1 could be
represented by (1, 2), (2, 3), (112, 113), . . .
Notice however that (1, 2) (2, 3), (1, 2) (112, 113), etc. and indeed, taking the
quotient by takes care of this redundancy. Notice also that this definition relies entirely
on previously defined entities.

7
Remark 3.14. In a first introduction to set theory it is not unlikely to find the claim that the
natural numbers are part of the integers, i.e. N Z. However, according to our definition,
this is obviously nonsense since N and Z := (N N)/ contain entirely different elements.
What is true is that N can be embedded into Z, i.e. there exists an inclusion map , given
by:

: N , Z
n 7 [(n, 0)]

and it is in this sense that N is included in Z.

Definition. Let n := [(n, 0)] Z. Then we define the inverse of n to be n := [(0, n)].

We would now like to inherit the + operation from N.

Definition. We define the addition of integers +Z : Z Z Z by:

[(m, n)] +Z [(p, q)] := [(m + p, n + q)].

Since we used representatives to define +Z , we would need to check that +Z is well-

defined. It is an easy exercise.
Example 3.15. 2 +Z (3) := [(2, 0)] +Z [(0, 3)] = [(2, 3)] = [(0, 1)] =: 1. Hallelujah!
In a similar fashion, we define the set of rational numbers by:

Q := (Z Z )/ ,

where Z := Z \ {0} and is a relation on Z Z given by:

(p, q) (r, s) : ps = qr,

assuming that a multiplication operation on the integers has already been defined.
Example 3.16. We have (2, 3) (4, 6) since 2 6 = 12 = 3 4.
Similarly to what we did for the integers, here we are representing each rational number
by the collection of pairs of integers (the second one in each pair being non-zero) such that
their (yet to be defined) ratio is precisely that rational number. Thus, for example, we
have:
2
:= [(2, 3)] = [(4, 6)] = . . .
3
We also have the canonical embedding of Z into Q:

: Z , Q
p 7 [(p, 1)]

Definition. We define the addition of rational numbers +Q : Q Q Q by:

[(p, q)] +Q [(r, s)] := [(ps + rq, qs)]

8
and multiplication of rational numbers by:

[(p, q)] Q [(r, s)] := [(pr, qs)],

where the operations of addition and multiplication that appear on the right hand sides are
the ones defined on Z. It is again necessary (but easy) to check that these operations are
both well-defined.

There are many ways to construct the reals from the rationals. One is to define a set
A of almost homomorphisms on Z and hence define:

R := A / ,

where is a suitable equivalence relation on A .