SET THEORY

Texts:
Introduction to Set Theory, Karel Hrbacek and Thomas Jech, 3rd Edition,
Marcel Dekker.
Set Theory, Charles C. Pinter, reprinted in Korea by KyungMoon.
References:
Naive Set Theory, Paul R. Halmos, UTM, Springer.
Elements of Set Theory, Herbert B. Enderton, Academic Press.
The Joy of Sets, Keith Devlin, UTM, Springer.
Set Theory, You-Feng Lin and Shwu-Yeng Lin, reprinted in Korea by KyungMoon.
0.1 A Brief History of Mathematical Logic
Cantors Set Theory

Russells Paradox

Hilberts Formalism and Godels Work

ZFC (Zermelo-Fraenkel + Choice) Axioms for Set Theory

Big sets like {x|x = x}, {x| x
/ x} are called (proper) classes. The assumption of the
existence of those proper classes are the main causes of paradoxes such as Russells. Hence
in formal set theory (see Appendix where we summarize the essence of axiomatic method),
those classes do not exist. But this course mainly focuses on elementary treatments
of set theory rather than full axiomatic methods, although as the lectures proceed some ideas of the axiomatic methods will occasionally be examined.1

1In

the chapters of this note, those reviews will be stated after ZFC: mark.
1

0.2 A Review of Mathematical Logic

When we prove a theorem, we use common mathematical reasonings. Indeed mathematical
statements or reasonings are often expressed as a sequence of formal symbols for briefness
and notational simplicity. Here we point out some of typical logic we use.
Let p, q, r be mathematical statements. By combining symbols (not), (and), (or),
, to p, q, r properly, we can make further statements (called Boolean combinations of
p, q, r) such as
p (not p);
p q (p implies q, equivalently say, if p then q; p only if q; not q implies not p);
p p (p or not p);
(p q) p q (not (p and q) i(=if and only if) (not p) or (not q));
and so on. As we know some of such Boolean combinations are tautologies, i.e. the values
of their truth tables are all true. For example, p p, p p q, p r r, (p q)
(q p), (p q) p q, (p q) p q are all tautologies. We indeed freely
use all those tautologies when we prove theorems. Let us see the following.
Theorem 0.1. x A x B, (i.e. A B) i x A (x Ax B) (i.e. AB = A).
Proof. () Assume left. To prove right, first assume x A. We want to prove x A and
x B. By the first assumption, x A. By the assumption of left, we have x B too.
Now assume x A x B. Then we have x A. Hence right is proved.
() Assume right. Hence if x A, then x A and B holds. In particular, x B.

Now let us go over more complicated logic involving quantifiers (for all), and (there
exists; for some). Let P (x), Q(x, y) be certain properties on x and x, y respectively. Note
that the followings are logically valid (i.e. always true no matter what the properties P, Q
exactly be, hence of course all tautologies are logically valid):
xP x(= (xP x)) xP x(= x(P x)) (not everything satisfies P i there is something not satisfying P );
xP x(= x(P x)) xP x(= (xP x)) (everything does not satisfy P i there does
not exist one satisfying P );
xyQxy xyQxy (for some (fixed) x, Qxy holds for every y i it is not the case
that for each x there corresponds y such that Qxy fails to hold). For example, if Qxy means
x cuts ys hair, then saying there is the one who cuts everyones hair is the same amount of
saying that it is not the case that every person can find someone whose hair the person does
not cut;
xyQxy yxQxy (if there is x such that for every y, Qxy holds then for any y we
can find x such that Qxy holds).
But the converse yxQxy xyQxy is not logically valid, since for example if Qxy
means x is a biological father of y, then even if everyone has a father, there is no one who is
a biological father of everybody;
(P x yQxy) y(P x Qxy) is logically valid.
Here the same holds if or (or both) is (are) replaced by or (or both), respectively;
(xP x xQx) x(P x Qx) is logically valid.
Does the same hold if or (or both) is (are) replaced by or (or both), respectively?
2

There are many other logically valid sentences. Logically valid sentences are also freely
used in proving theorems.
Definition 0.2.
(1) By an indexed family {Ai | i I} of sets, we mean a collection of
sets
A
indexed
by i I.
i

(2) {Ai | i I} := {x| i I x Ai }.2

(3) {Ai | i I} := {x| i I, x Ai }.3
Theorem 0.3. Let {Ai | i I} be an indexed family of sets.

(1) IfAi B for all i

I, then {Ai | i I} B.
(2) ( {Ai | i I})c = {Aci | i I}.

Proof. (1) Suppose that Ai B for all i I. Now let x {Ai | i I}. We want to show
x B. By the definition, x Ai0 for some i0 I. Therefore by supposition, as desired
x B.

(2) x ( {Ai | i I})c i x

/ {Ai | i I} i i I, x
/ Ai i i I, x Aci i
c
x {Ai | i I}.

Throughout the course students will be trained to be capable to do these kinds

of logical reasonings fairly freely and comfortably.

2More
3More

precisely {x| i(i I x Ai )}.

precisely {x| i(i I x Ai )}.
3

1. Sets
Definition 1.1. A = B if x(x A x B).
A B (A is a subset of B) if x(x A x B). We say A is a proper subset of B if
A B and A = B (A B).
, called the empty set, is the set having no elements.
We can write a set in the form of {x| P (x)}.
E.g. = {x| x = x}, {x, y} = {z| z = x or z = y}.
A B := {x| x A or x B}. A B := {x| x A and x B}. P(A) := {B| B A}.
A B = {x| x A x
/ B}. A B := (A B) (B A). Ac := {x U | x
/ A}.
We say sets A, B are disjoint if A B = .
For a set S,

S = {X| X S} = XS X := {x| x A for some A S} = {x| A S s.t. x A}.

For a nonempty set S,

S = {X| X S} = XS X := {x| x A for all A S} = {x| A S, x A}.

Laws of Set Operations
A = A, A = , A
A =
,
A B A B and B A (when A = ).
Commutativity A B = B A, A B = B A, A B = B A.
Associativity A (B C) = (A B) C, A (B C) = (A B) C, A (B C) =
(A B) C.
Distributivity

A (B C) = (A B)
(A C), A (B C) = (A B) (A C).
A S = {A X| X S}, A S = {A X| X S}.
DeMorgans laws U (A B) = (U A) (U B) ((A B)c = Ac B c ),
c
U (A
((A B)
= Ac B c ).
B) =
(U A) (U B)
c
U ( S) = {U X| X S} (( S) = {X c | X S}),
U ( S) = {U X| X S} (( S)c = {X c | X S}).

2. Relations, Functions, and Orderings

Relations
Definition 2.1. (a, b) := {{a}, {a, b}}, the ordered pair of a, b.
By iteration, we can define ordered tuples: (a, b, c) := ((a, b), c); (a, b, c, d) := ((a, b, c), d);
...
A B := {(a, b)| a A, b B}.
A B C := (A B) C = {(a, b, c)| a A, b B, c C}.
...
A2 := A A, A3 := A A A, ...
Theorem 2.2. (a, b) = (a , b ) i a = a and b = b .
Definition 2.3. A (binary) relation R is some set of ordered pairs. We may write xRy for
(x, y) R.
We say R is a unary relation in A, if R A.
R is a binary relation in A, if R A2 .
R is a ternary relation in A, if R A3 .
...
R is an n-ary relation in A, if R An .
Definition 2.4. Let R be a relation.
(1) dom R := {x| y(x, y) R}.
(2) ran R := {y| x xRy}.
(3) field R := dom R ran R.
(4) R1 := {(x, y)| (y, x) R}.
(5) R[A] := {y ran R| x A xRy}, the image of A under R. It can be seen that
R1 [B] = {x dom R| y B s.t. xRy}, the inverse image of B under R.
(6) RA := {(x, y) R| x A}, the restriction of R to A.
(7) Let S be a relation. Then S R := {(x, z)| y (xRy ySz)}.
Excercise 2.5. Let R, S, T be relations.
dom R = ran R1 , ran R = dom R1 , (R1 )1 = R,
R[A B] = R[A] R[B], R[A B] R[A] R[B].
R (S T ) = (R S) T , (R S)1 = S 1 R1 .

Functions
Definition 2.6. A relation F is said to be a function (or mapping) if for each a dom F ,
there is a unique b such that aF b holds.
If F is a function, then F (a) := the unique b such that of (a, b) F , the value of F at a.
We write f : A B if f is a function, A = dom f , and ran f B, and say f is a function
from (or on) A (in)to B.
5

Lemma 2.7. Let F, G be functions. Then F = G i dom F = dom G and x dom F ,

F (x) = G(x).
Definition 2.8. F : A B is given.
(1) We say F is 1-1 (or injective, an injection) if for a = b A, we have F (a) = F (b).
(2) F is onto (or surjective, a surjection) if B = ran F .
(3) F is bijective (or, a bijection) if F is 1-1 and onto.
Theorem 2.9. Let f, g be functions.
(1) g f is a function.
(2) dom(g f ) = {x dom f | f (x) dom g}.
(3) (g f )(x) = g(f (x)) for all x dom(g f ).
Theorem 2.10. A function f is invertible (i.e. the relation f 1 is a function too) i f is
injective.
f : A B is invertible and f 1 : B A i f : A B is bijective.
Notation 2.11. AB := {f | f : A B}.
Note that for all A, A = {} and if A = , then A = .
Now assume f : I S = {X| X S} is onto. Then S = ran f = {f (i)| i I}. We may
write f (i) = Xi , and write S = {Xi | i I} = {Xi }iI . We call S, an indexed family of sets.
For S = {Xi | i I},

Xi denotes
S,
Xi (with nonempty I) denotes
S, and
iI

S=

iI

Xi := {f | f : I

S s.t. i I, f (i) Xi }.

iI

Excercise 2.12. f is a function.

(1) Let f : A B. If f is bijective, then f 1 f = IdA , and f f 1 = IdB . Conversely
if g f = IdA and f g = IdB for some g : B A, then f is bijective and g = f 1 .
(2) Letf : A B,
h : B C be given. If both f, h are 1-1 (onto resp.), then so is h f .
(3) f [ iI
i ].
f [A1
Ai ] = iI
f
[Ai ].
A
]
=
f 1
[
iI i iI
f [ iI
Ai ] iI
f [Ai ].
f 1 [ iI Ai ] = iI f 1 [Ai ].

Equivalence Relations
Definition 2.13. Let R be a relation in A (i.e. R A2 ).
(1) R is reflexive if for all a A, aRa.
(2) R is symmetric if a, b A, aRb bRa.
(3) R is transitive if a, b, c A, aRb bRc aRc.
R is said to be an equivalence relation on A, if R is reflexive, symmetric, and transitive.
6

Definition 2.14. Let E be an equivalence relation on A. For a, b A with aEb, we say a

and b are equivalent (or, a is equivalent to b) modulo E.
[a]E = a/E := {x| xEa} is called the equivalence class of a modulo E (or, the Eequivalence class of a). A/E := {[a]E | a A}.
Lemma 2.15. E is an equivalence relation on A. Then for x, y A, xEy i [x]E = [y]E .
Definition 2.16. By a partition of a set A, we mean a family S of nonempty subsets of A
such that
(a) for C= D S, C and D are disjoint,
(b) A = S.
Theorem 2.17. Let F be an equivalence relation on A, and let S be a partition of A.
(1) {[a]F | a A} partitions A.
(2) If we define
ES := {(x, y) A2 | x, y C for some C S},
then it is an equivalence relation on A such that S = A/ES . Moreover, if S = A/F ,
then ES = F .
Definition 2.18.
(1) Let E, F be equivalence relations on A. We say E refines (is an
refinement of, or is finer than) F (equivalently say F is coarser than E), if E F .
(2) Let both E F be equivalence relations on A. The quotient of F by E (written as
F/E) is
{([x]E , [y]E )| xF y}.
Theorem 2.19. Let E F be equivalence relations on A. Then F/E is an equivalence
relation on A/E.
Theorem 2.20. Let f : A B. Then f induces an equivalence relation on A such that
a b i f (a) = f (b). Define : A A/ by mapping x A to [x] . Then there is a
unique f : A/ B such that f = f . Moreover f is 1-1, and if f surjective, so is f.
A

B
f

A/
Corollary 2.21. Let E F be equivalence relations on A. Then the canonical map f :
A/E A/F sending [x]E to [x]F is well-defined and onto. Moreover F/E is an equivalence
relation on A/E induced by f . Hence f : (A/E)/(F/E) A/F is a bijection.
f

A/F
f

A/E

y
(A/E)/(F/E)

Orderings
Definition 2.22. Let R be a relation in A.
(1) R is said to be antisymmetric if a, b A, aRb bRa a = b.
(2) R is asymmetric if a, b A, aRb bRa.
(3) We say R is a (partial) ordering (or an order relation) of A if R is reflexive, antisymmetric, and transitive. The pair (A, R) is called an ordered set (or, a poset). If R
is asymmetric and transitive, then we call R a strict ordering of A.
(4) Let (A, ) be a poset. Clearly partially orders any subset of A (i.e. for B( A),
(B, B ) is a poset where B := {(a, b) | a, b B}.) We say a, b A are
comparable if a b or b a. Otherwise a, b are incomparable. C( A) is called a
chain in A if any two elements of C are comparable. We say is a linear (or total)
ordering of A if any two elements of A are comparable, i.e. A itself is a chain in A.
As the reader may notice, an order relation is an abstraction of the notion less than or
equal to, while a strict relation is an abstraction of the notion strictly less than.
Theorem 2.23. Let both , be relations in A.
(1) is a strict ordering i is irreflexive (i.e. x A, x x) and transitive.
(2) (a) If is an ordering, then the relation < defined by x y x = y is a strict
ordering.
(b) Similarly, if is a strict ordering, then the relation defined by x y x = y
is an ordering.
(3) (A, ) is linearly ordered i is reflexive, transitive, and x = y A, exactly one
of x y or y x holds.
Definition 2.24. Let A be ordered by , and B A.
(1) b B is the least element of B if b x for all x B. b is a minimal element of B
if x B s.t. x < b.
(2) b B is the greatest element of B if x b for all x B. b is a maximal element of
B if x B s.t. b < x.
(3) a A is a lower bound of B if a x for all x B. We say B is bounded below (in A)
if there is a lower bound of B.
a A is an upper bound of B if x a for all x B. We say B is bounded above
(in A) if there is an upper bound of B.
(4) a A is the infimum (or g.l.b.) of B (in A) if a is the greatest element of the set of
all lower bounds of B. We write inf A B = glbA B = a.
a A is the supremum of B in A if a is the least element of the set of all upper
bounds of B. We write supA B = lubA B = a.
(5) Assume linearly orders A. For a, b A, we say b is a successor of a (a is a
predecessor of b) if a < b and there is no c such that a < c < b.
Excercise 2.25. Let (A, ) be an ordered set, and B A. Then b A is the least (greatest,
resp.) element of B i b = inf B (= sup B, resp.) and b B.

Definition 2.26. Let (A, A ), (B, B ) be posets. We say A and B are order isomorphic
(write A
= B) if there is a bijective f : A B, called an order isomorphism, such that for
all a, b A, a <A b i f (a) <B f (b).
Definition 2.27. An ordering of a set A is called a well-ordering if for any nonempty subset
B of A has a least element. Every well-ordering is a linear ordering. If well-orders a set
A, then clearly it well-orders any subset of A, too.
Let (W, ) be well-ordered. We say S( W ) is an initial segment (of a W ) if S = {x
W | x < a}. We write S = W [a] or = seg a.
Lemma 2.28. Any well-ordered set A is not order-isomorphic with a subset of an initial
segment of A.
Theorem 2.29. (Well-Ordering Isomorphism Theorem) Let (A, A ), (B, B ) be wellordered. Then exactly one of the following holds.
(1) A
= B.
(2) A
= B[b] for unique b B.
(3) B
= A[a] for unique a B.
In each case, the isomorphism is unique.
Corollary 2.30. Let (A, A ) be well-ordered. Then for B A, it is order-isomorphic with
A or with an initial segment of A.

Definition 2.31. Let (A, <A ), (B, <B ) be linear orderings.

Then A l B, called the lexicographic product of A and B, is the set A B ordered by <l ,
called the lexicographic ordering of A and B, as follows:
(x1 , y1 ) <l (x2 , y2 ) i x1 <A x2 or (x1 = x2 and y1 <B y2 ).
Similarly, A al B, called the antilexicographic product of A and B, is again the set A B
ordered by <al , called the antilexicographic ordering of A and B, as follows:
(x1 , y1 ) <al (x2 , y2 ) i y1 <B y2 or (y1 = y2 and x1 <A x2 ).
If A, B are disjoint, then A B, called the sum of A and B, is the set A B ordered by
< as follows:
x < y i x, y A and x <A y or
x, y B and x <B y or
x A and y B.
Lemma 2.32. Assume that (A, <A ), (B, <B ) are linear orderings. Then so are A l B,
A al B, and A B (when A B = ). Similarly, if (A, <A ), (B, <B ) are well-orderings,
then so are the products and the sum.

3. Natural Numbers
ZFC: The hidden intension of this chapter is to figure out how to formalize , the set of
natural numbers, and its arithmetic system in ZFC. To show the existence of , we need
Axiom of Infinity, and to define arithmetic, we need the Recursion Theorem which we shall
prove later.
Definition 3.1. For a set a, its successor S(a) = a+ = a + 1 := a {a}. Note that a a+
and a a+ . 0 := , 1 := 0+ , 2 := 1+ , 3 := 2+ , ...
A set A is said to be inductive, if A, and is closed under successor (i.e. x A x+
A).
N = := {x| I(I is an inductive set x I)}. A natural number is an element in .
Theorem 3.2. itself is inductive.
Induction Principle Any inductive subset of is equal to . Namely the following holds.
Let P (x) be a property.4 Assume that
(a) P (0) holds, and
(b) for all n , P (n) implies P (n+ ).
Then P holds for all n .

Definition 3.3. A set A is said to be transitive if every member of a member of A is a

member of A, i.e. x a A x A.
Theorem 3.4. Every natural number is transitive. is transitive as well.
Definition 3.5. Define the relation < on by m < n i m n.
Then m n i (m < n or m = n) i m n+ i m < n+ .
Theorem 3.6. < (= ) is a well-ordering of .
Corollary 3.7. Any natural number is well-ordered. Any subset of is order isomorphic
with either a natural number or . Any X( n ) is isomorphic with some m n.
Corollary 3.8.
(1) is a linear ordering on .
(2) For any m, n , exactly one of m < n, m = n, n < m holds.
(3) For any m, n , we have m < n i m n i m n, and m n i m n.

4ZFC:

Indeed a formula (x).

10

Induction Principle (2nd Version) Let P (x) be a property. Assume that for every
n ,
if P (k) holds for all k < n, then P (n) holds.
Then P (n) holds for all n .

The Recursion Theorem Let A be a set and let c A. Suppose that f : A A. Then
there is a unique function h : A such that h(0) = c and h(k + ) = f (h(k)) for all k .5
+
Proof. Let A = {R|
R A (0, c) R k ((k, x) R (k , f (x)) R)}. Since
AA=
, A exists. We let h = A.

Claim 1) h is the desired function:

Claim 2) If h : A satisfies the two conditions then h = h :

Arithmetic of Natural Numbers We will use the Recursion Theorem to define + and
on .
Definition 3.9.
(1) Fix k . We define +k : as follows.
+k (0) = k,
+k (n+ ) = (+k (n))+ for each n .
By the Recursion Theorem, there is a unique such +k . Now define
+ : 2 by k + m = +k (m).
(2) Fix k . We define k : as follows.
k (0) = 0,
k (n+ ) = (k (n)) + k for each n .
By the Recursion Theorem, there is a unique such k . Now define
: 2 by k m = k (m).
Theorem 3.10. +, satisfy the usual arithmetic laws such as associativity, commutativity,
distributive law, cancelation laws, and 0 is an identity for +, and 1 is an identity for .
From , we can consecutively build Z = the set of integers; Q = the set of rational
numbers; R = the set of real numbers; C = the set of complex numbers. In particular, when
we construct R from Q, well-known notions such as Dedekind cuts or Cauchy sequences are
used.

5ZFC:

Formally,

Acf [c A f is a function !h[h : A h(0) = c k(k h(k + ) = f (h(k)))]]

is provable.
11

4. Axiom of Choice
ZFC: One of historically important issues is whether Axiom of Choice (AC) is provable
from other axioms (ZF). Kurt Godel and Paul Cohen proved that AC is independent from
other axioms. It means even if we deny AC, we can still do consistent mathematics. Indeed
there are a few of mathematicians who do not accept AC mainly because AC supplies non
constructible existence proofs mostly in the form of Zorns Lemma, and also Banach-Tarski
phenomena. However the absolute majority of contemporary mathematicians accept AC
since it is so natural and again a great deal of abstract existence theorems in basic algebra
and analysis are relying on AC.
Now in this chapter, formally we only assume ZF. We then prove the equivalence of several
statements to AC. From chapter 5, we freely assume AC (so formally then we work under
ZFC).6
Definition 4.1. Axiom of Choice (Version 1): Let A be a family of mutually disjoint
nonempty sets. Then there is a set C containing exactly one element from each set in A.
Axiom of Choice
(Version 2): Let {Ai }iI be an indexed family of nonempty sets. If

I = , then iI Ai is nonempty.
Axiom of Choice (Version 3): For any nonempty set A, there is a function F : P(A) A
such that F (X) X for any nonempty subset X of A. (Such a function is called a choice
function for P(A).)
Theorem 4.2. AC 1st, 2nd, and 3rd versions are all equivalent.
Theorem 4.3. The following are equivalent.
(1) AC.
(2) (Well-Ordering Theorem) Every set can be well ordered.
(3) (Zorns Lemma) Given a poset (A, ), if every chain B of A has an upper bound,
then A has a maximal element.

(3) (Zorns Lemma) Given a poset (A, ), if every chain B of A has the supremum,
then A has a maximal element.
(4) (Hausdor s Maximal Principle) Every poset has a maximal chain.

Proof. (1)(3) Assume (1). Let (A, ) be a poset such that

each chain has the supremum.

(*)

In particular there is the least element p = sup in A.

To lead a contradiction, suppose that A has no maximal element. Then for each x A,
S(x) := {y A| x < y} is a nonempty subset of A. Let F be a choice function for P(A) so
that F (S(x)) S(x). Define f : A A by f (x) = F (S(x)). Then since f (x) S(x),
x < f (x), for each x A.
Now let H = {B A|
6In

(i) p B, (ii) x B f (x) B,

(iii) if C( B) is a chain then supA C B}.

some set theory books, authors are sensitive to discern between proofs using AC and not.
12

(**)


Due to (*), A H =
. Hence we let P := H. It is easy to check that P H and
p P A. We shall show P is a chain. For this end we claim the following.
Claim 1) Let P := {a P | x P, x < a f (x) a}. Let a P . Then
Ba := {x P | x a or f (a) x}
is equal to P . Indeed, a is comparable with any element in P , and P is a chain: It is not
hard to show that Ba ( P ) satisfies (i)(ii)(iii) (Exercise). Hence P Ba H, and P = Ba .
Now given b P = Ba , either b a or a < f (a) b (by (**)), so a, b are comparable, and
P is a chain. We have proved Claim 1.
Claim 2) P satisfies (i)(ii)(iii). Hence P = P :
Firstly, that p P is vacuously true.
Secondly, suppose that a P . We want to prove f (a) P , i.e. given x( P ) < f (a),
show f (x) f (a). Now due to Claim 1, we have P = Ba , and thus x a. Then since
a P , either f (x) a(< f (a)) or (x = a and f (x) = f (a)). Hence always f (x) f (a).
Thirdly, suppose that C( P P ) is a chain. Let c = supA C. Then since P ( H)
satisfies (iii), c P . It remains to show c P . Let x( P ) < c. We want to see f (x) c.
Now there is y0 C s.t. x < y0 , since otherwise x < y for each y C, so due to Claim 1
(y is comparable with x), we have y x and c = supA C x, a contradiction. Then since
y0 P , f (x) y0 c. Therefore Claim 2 is proved.
Now by Claim 1, 2, we conclude that P is a chain. Then due to (*), there is m := supA P .
Since P H, by (iii)(ii) we have m P and f (m) P . But by (**), m < f (m) and
m = supA P , a desired contradiction is obtained. Hence A must have a maximal element.

We have proved (1)(3) .

(3) (4)
(4)(3)
(3)(2) Let A be a set. Let
A = {(B, )| B A, B 2 , and well-orders B}.
Given (B1 , 1 ), (B2 , 2 ) A, define
(B, 1 ) (B2 , 2 ) i 1 2 ( B1 B2 ), and x B1 , y B2 B1 , x <2 y.
Then (A, ) forms an ordered set (Exercise).

Claim)
Assume
a
chain
C
=
{(B
,

)}

A
is
given.
Let
C
:=
{Bi | i I} and let
i
i
iI

C := {i |i I}. Then (C, C ) A, and is an an upper bound of C (i.e. (Bi , i )

(C, C ) for all (Bi , i ) C): Firstly, it is easy to see that C is a relation in C (Exercise).
Secondly, we shall show that C well-orders C. It can be seen that C is reflexive and
antisymmetric (Exercise). To show transitivity, assume x C y and y C z. Then x i y
and y j z for some i, j I. Since C is a chain we have say i j . Then x j y and by
the transitivity of j , we have x j
z and x C z. It remains to show C well-orders C.
Let D be a nonempty subset of C = iI Bi . For some i0 , D := D Bi0 = . Hence D has
the least element b in (Bi0 , i0 ). We shall prove in fact b is the desired least element of D
in (C, C ). Let x D. Hence x Bj for some j I. Now if x Bi0 , then (b, x) i0 C
and we are done. If x Bj Bi0 , then since C is a chain, we must have Bi0 Bj and
13

(Bi0 , i0 ) (Bj , j ). Hence by the definition of , we have b <j x and b <C x. We are
done.
Thirdly, it can be seen (C, C ) is an upper bound of C (Exercise). We have proved Claim.
By Claim and Zorns Lemma, there is a maximal element (D, D ) A. It remains
to show A = D. Note that D A. So if D = A, then there is d A D. Now let
D := D {(y, d)|y D}. Then it is easy to see that (D, D ) (D {d}, D ) A
(Exercise). It contradicts the maximality of (D, D ). Therefore A = D, and the proof is
done.

(2)(1)
Applications of Zorns Lemma

Typical steps of the existence proofs using Zorns Lemma (e.g. above proof (3)(2)).
1st, consider a family F of all the candidate sets. F is suitably ordered (often ordered by
inclusion).

2nd, show that every chain C in F has an upper bound (often showing C F).
3rd, thus by Zorns Lemma, there is a maximal element M in F. To show M is the desired
object, often, arguing that if not, then the maximality of M fails.
Theorem 4.4. Every vector space V has a basis.
Proof. Let F be the family of linearly
independent subsets of V . F is orderedby inclusion.

Let C be
a
chain
in
F.
Then
C(
V
) is also linearly independent. Hence C F and

indeed C = sup C. Hence by Zorns Lemma, F has a maximal element, say B F. We

shall show B is a basis for V . If not, then there is v0 (= 0) V such that
v0 is not a linear combination of any finite subset of B.

(*)

In particular, v0
/ B, and B := B {v0 } is a proper superset of B.
Claim) B is linearly independent as well: Assume 0 v0 + 1 v1 ... + n vn = 0 where
v1 , ..., vn B and scalars i . We want to show 0 = ... = n = 0. Note that 0 = 0 since
otherwise v0 is a linear combination of {v1 , ..., vn } contradicting (*). Now then since B is
linearly independent, 1 = ... = n = 0 and we are done.
The claim contradicts then the maximality of B. Hence B must be a basis for V .

Theorem 4.5. Every ring with identity has a proper maximal ideal.
Theorem 4.6. Let G be a group and let A G with 1G A. Among the subgroups of G
contained in A, there is a maximal one.
Proof. Let F = {H A| H is a subgroup of G}. F is ordered by inclusion. Let C( F ) be
a chain. We want to show C has an upper bound. If C = , then {1G }( F ) is an upper
bound. Hence we can assume C =
. Now let C := C. Then clearly 1G C A. We claim
that C is a subgroup of G (and thus C F): If x, y C, then x H1 C, y H2 C.
Since C is a chain, say H1 H2 , and since H2 is a group, we have x y H2 C. Moreover
x1 H1 C. Hence C is a subgroup.
14

Therefore either case, C has an upper bound. Hence by Zorns Lemma, there is a maximal
element in F, which is the desired one.

15

5. Finite, Countable, and Uncountable Sets

Definition 5.1. Two sets A, B are equipotent (or equinumerous) (write A B) if there is a
bijection f : A B.
Assumption There are sets called cardinal numbers (or cardinals) such that for any set A,
there is a unique cardinal |A| (or written Card(A), the cardinality of A) equipotent to A.7
Moreover for n , we have |n| = n, and || = . We use lowercase Greek letters , , ..
to denote cardinals.
Remark 5.2.
(1) A A.
(2) A B i B A.
(3) If A B and B C, then A C.
Hence A B i Card(A) = Card(B), and we equivalently say A, B have the same
cardinality when A, B are equipotent.
Definition 5.3. Write A B (or |A| |B|) if there is an injection f : A B.
Remark 5.4.
(1) A B A B. In particular, A A.
(2) A B and A C implies C B.
(3) If A B and B C, then A C.
(4) A B and B C implies A C.
Theorem 5.5. (AC) If f : B A is onto, then A B. The converse holds if A = .
Theorem 5.6. (Cantor-Bernstein) If |A| |B| and |B| |A|, then A B (i.e.
and implies = ).
Theorem 5.7. (AC) For any cardinals , , either or .
Theorem 5.8. (Cantor)
(1) |A| < |P(A)|.
(2) |P(A)| = |A 2|.
(3) < Card(R) = Card(]0, 1[) = Card(P()).
The Continuum Hypothesis There is no cardinal such that < < |R|.

Finite Sets, Countable Sets

Definition 5.9.
(1) We say a set is finite having n elements if it is equinumerous to
n( ). A set is infinite if it is not finite.
(2) A set A is called countable if |A| = . It is called at most countable if |A| ,
uncountable if not at most countable.
Due to 3.7, any subset of a finite set is finite, so any set having an infinite subset is infinite.
We shall see that any countable set (in particular ) is infinite.
7This

is a temporary assumption. We shall see in Chapter 8 that this follows as a theorem.

16

Theorem 5.10.
(1) Any infinite subset of a countable set is countable. Hence, any
subset of a countable set is either finite or countable.
(2) A set B is at most countable i B is either finite or countable. If B = , then B is
at most countable i there is surjective f : B.
Lemma 5.11. (Pigeonhole Principle) If X n( ), then X n.
Corollary 5.12.
(1) Let m, n . If m n, then m = n. Hence if A m and A n,
then m = n. Moreover, m n as natural numbers i m n as finite cardinals.
(2) If A is finite, then for any proper subset B of A, B A.
Theorem 5.13. (AC) The following are equivalent.
(1) A is infinite.
(2) A has a countable subset.
(3) There is B A such that A B.
Theorem 5.14.
(1) If A, B are countable, then so is A B. Hence for n = 0,
n
n
= | | = | |.

(2) Let {An | n } be a family of (at most, resp.) countable sets. Then n An is
(at most, resp.) countable. Hence if A, B are (at most, resp.) countable, then so is
A B.

(3) Moreover if A is countable, then both FinSeq(A) = A< := n n A, and

{B A| B is finite} are countable.
(4) Both Z, Q are countable.

17

6. Cardinal Numbers
Definition 6.1. Let , be cardinals.
(1) + := |A B| where |A| = , |B| = and A, B are disjoint.
(2) := |A B| where |A| = , |B| = .
(3) := |B A| where |A| = , |B| = .
To check the definitions are independent from the choice of sets A, B, we need the following
lemma.
Lemma 6.2. Let A A and B B . Then
(1) A B A B if A B = A B = .
(2) A B A B .

(3) B A B A .
Corollary 6.3. + = , = , |A 2| = |P(A)| = 2|A| .
Theorem 6.4.
(1) + = + ; = .
(2) ( + ) + = + ( + ); ( ) = ( ).
(3) ( + ) = + .
(4) + = ; ( ) = ; ( ) = .
For finite cardinals (i.e. natural numbers) the addition, multiplication, and exponentiation
defined in 6.1 are equal to those in 3.9.
Theorem 6.5. Assume 1 2 and 1 2 . Then 1 + 1 2 + 2 ; 1 1 2 2 ;
and if 1 = 0, 1 1 2 2 .
Theorem 6.6. (AC) For infinite cardinal , = .
Corollary 6.7.
(1) For infinite cardinals , , + = = max{, }.
(2) For 2 with infinite , we have 2 = = .

18

7. Ordinal Numbers
Definition 7.1. We say a set is an ordinal (number) if
(1) it is transitive and
(2) well-ordered by (i.e. := {(x, y)| x , y , x y} well-orders ).
We use lower-case Greek letters , , to denote ordinals.
Fact 7.2. Each natural number and are ordinals. Any element of an ordinal is an ordinal.
Lemma 7.3. Let , be ordinals.
(1)
/ .
(2) + = {} is an ordinal.
(3) is an ordinal too.
(4) i .
For ordinals , we write < i . Therefore i ( or = ) i .
Theorem 7.4. The class OR := {| ordinal } is well-ordered by <.8 Namely, for ordinals
, , ,
(1) if < and < , then < ;
(2) if < , then < ;
(3) exactly one of < , or = , or < holds;
(4) if A is a nonempty set of ordinals, then A is the <-least ordinal in A.
Moreover,
(5) < i + (hence, + isa successor of );
(6) if A is a set of ordinals, then A is an ordinal sup A.
Note that for an ordinal ,
= {x| x } = { ordinal | < } = OR[] = seg[].
Moreover for < , it follows [] = .
Lemma 7.5. Any two order isomorphic ordinals are equal.
Theorem 7.6. Every well-ordered set is order isomorphic with exactly one ordinal number.
ZFC: If Well-ordering Isomorphism Theorem holds for proper classes such as OR, then
above Theorem 7.6 is a direct consequence of it. Indeed the proof of Theorem 7.6 can
be completed by mimicking that of W.I.T., but a careful investigation says Replacement
Axioms should be used at a step. For more details see the textbook.
Transfinite Induction (First Version) Let P (x) be a property. Assume that for all
ordinal ,
if P () holds for all ordinal < , then P () holds.
Then P () holds for all .
8Formally,

for example, for (1)

xyz(OR(x) OR(y) OR(z) x y y z x z)

is provable, where OR(x) is a formula saying x is an ordinal.

19

Definition 7.7. We call a successor ordinal if = + for some . We call it a limit

ordinal if it is non-zero and not a successor ordinal.
Transfinite Induction (Second Version) Let P (x) be a property.
Assume that
P (0) holds;
if P () holds, then P (+ ) holds;
for a limit , if P () holds for all < , then P () holds.
Then P () holds for all .
Transfinite Recursion Let H(x, y) be a functional property and let A be a set. Then there
is a functional property F (x, y) on OR such that
F (0) = A,
F (+ ) =H(F ()),
F () = {F ()| < } for a limit .9
Ordinal Arithmetic We will use transfinite recursion to define + and for ordinals.
Definition 7.8.
(1) Fix an ordinal . We define + as follows.
+ (0) = ,
+ (+ ) = (+ ())+ for each ordinal ,
+ () = sup{+ ()| < } for limit .
By the Transfinite Recursion Theorem, such + is defined. Now define + such that,
for ordinals , ,
+ := + ().
Similarly define and exponentiation between ordinals.
(2) 0 := 0, + := + , := sup{ | < } for limit .
+
(3) 0 := 1, := , := sup{ | < } for limit .
Theorem 7.9. For ordinals , if
= A and
= B (A B = ), then

(1) + = A B,
(2)
= A al B,
(3)
= A l B.
Laws of Ordinal Arithmetic
+ = + 1; + ( + ) = ( + ) + ; but + 1 = 1 + .
1 2 1 + 2 + ; 1 + < 2 + 1 < 2 ; but even if 1 < 2, we have
1 + = = 2 + .
(Left Cancelation) + 1 < + 2 i 1 < 2 ; + 1 = + 2 i 1 = 2 .
If , then there is a unique such that + = .
9ZFC:

More formally, for any formula H(x, y), there corresponds a formula F (x, y; A) = FA (x, y) such
that
+
x!yH(x, y) A[[OR() [!zFA (, z) FA (0, A)
FA ( , H(FA ()))
( limit FA (, {FA ()| < }))]]]
is provable.
20

( ) = (); ( +) = +; but (1+1) = 2 = < 1 +1 = 2.

1 2 1 2 ; 1 < 2 1 < 2 ; but again 1 = = 2 .
(Left Cancelation) ( = 0) 1 < 2 i 1 < 2 ; 1 = 2 i 1 = 2 .
(Division Theorem) For any and nonzero , there are a unique < and a unique
such that = + .
( ) = ; = + ; but (2 2) = < 2 2 = 2 .
1 2 1 2 ; 1 < 2 1 < 2 ; but for 2 < 3, we have 2 = 3 = .
(1 < ) 1 < 2 i 1 < 2 ; 1 = 2 i 1 = 2 .
0,1,2,...,, + 1, + 2, ..., + = 2, 2 + 1..., 3, ..., = 2 , ..., 3 , ..., ,
2

+ 1, ..., 2, ..., = +1 , ..., , ..., , ..., , ...

where = sup{, , , , ...}.

Theorem 7.10. (The Normal Form Theorem) Every ordinal > 0 is uniquely expressed
as
= 1 k1 + ... + n kn
where 1 > ... > n , and 0 < ki (i = 1, ...n).

21

8. Alephs
Definition 8.1. By an initial ordinal we mean an ordinal not equinumerous to any ordinal
< .
Theorem 8.2. (AC) Any set is equinumerous to a unique initial ordinal.
Now as promised, we define cardinal numbers as initial ordinals, and for a set A, its
cardinality |A| is the unique cardinal as in Theorem 8.2. So now we can remove Assumption
in Chapter 5.10 Note also that cardinal and ordinal arithmetic are dierent even if
they use the same notation for addition, multiplication, and exponentiation. For example
as ordinal addition 2 + = = + 2, while as cardinal addition, 2 + = = + 2.
Similarly, as ordinals 2 = and < 2 while as cardinals < 2 and = 2 . But up to
the context, it should be clear whether what is stated is for ordinal, or cardinal
arithmetic.11
Both CD := {| is a cardinal }, IC := {| is an infinite cardinal } are proper classes.
Theorem 8.3. There is an order isomorphic function (functional property) : OR IC.12
0 = , ..., 2 , ..., , ..., , ..., 1 , ..., 2 , ..., , ...
The Continuum Hypothesis 20 = 1 .
Definition 8.4. : OR IC is defined as
0 :=
+1 := 2
:= sup{ | < } for limit .
The Generalized Continuum Hypothesis For all ordinal , 2 = +1 , equivalently,
= .

10In

most advanced book for Axiomatic Set Theory, ordinals are introduced first, then cardinals and
related topics are developed later.
11In other branches of mathematics such as Algebra and Analysis, cardinal arithmetic is mostly used.
12ZFC: Formally there is a formula (x, y) so that a sentence
!((, ) IC()) ( () ()) (IC() (, ))
is provable from ZFC. Here ..., ... stand for x(OR(x) ..., x(OR(x) ..., respectively.
22

9. Advanced Topics
, , ... denote ordinals and , , ... denote cardinals.
More on Cardinal Arithmetic
Definition 9.1. For a cardinal , we let + , a successor of , be a successor cardinal in the
ordering of CD (i.e. if = , then + = +1 ).
Hence we call , a successor cardinal if it is a successor of some cardinal. Call a limit
cardinal if < and is not a successor.
For a function f : (both infinite), we say f is cofinal if ran f has no upper bound
in .
cf() := the least cardinal in {| cofinal f : }.
For an infinite , we say is regular if cf() = , and say singular if not regular.
Clearly cf() , and cf( ) = .
Theorem 9.2. Any infinite successor cardinal is regular.

Definition 9.3. Let i (i I) be cardinals. Then iI i := Card( A) where A = {Ai | i

I} is an indexed family
disjoint sets such that i = Card(Ai ).
of mutually

<
We define := < .
Theorem 9.4. (K
onigs Theorem) If i < i (i I), then

i < Card( i ).
iI

iI

Corollary 9.5. For an infinite , we have < cf() , and < cf(2 ). Hence (in ZFC) it
follows
20 = .
ZFC: Foundation Axiom
Theorem 9.6. There do not exist sets A1 , ..., An such that A1 A2 ... An A1 . In
particular B
/ B, for any set B.
Proof. Suppose not, and let A = {A1 , ..., An } = . Then and for each Ai A,
An A1 A =
if i = 1;
Ai1 Ai A = if 1 < i n.
Hence both cases contradict Axiom of Foundation.
By Transfinite Recursion, for each , we define a set V as follows:
V0 := ,
V+1 :=
P(V ),
V := {V | < } for limit .
Each set in V is said to be well-founded.
23

Indeed there is a formula WF(x, y) such that for , WF(, y) says y V .

y WF(, y) says every set is well-founded.

Thus

Theorem 9.7. (In ZF-{Axiom of Foundation}) Axiom of Foundation holds i every set is
well-founded.

24

Appendix: ZFC Axioms for Set Theory

Definition We recursively define formulas in set theory.
Variables in set theory are symbols x, y, z, ..., a, b, c, ..., X, Y, Z, ..., A, B, C, ....
Formulas:
(1) For variables u, v, we let both u = v, u v be formulas.
(2) If , are formulas, then each of (), ( ) is a formula.
(3) If is a formula and u is a variable, then u is a formula.
(4) Nothing else is a formula unless it can be obtained by finitely many applications of
(1)(2) and (3).
For example, := x((x a x b) a x a) is a formula. Here the variable
x is bounded by the quantifier , where as a, b have unbounded occurrence (even if a once
occurs boundedly). We call such variables having unbounded occurrence free variables in
the formula , and we may write as (a, b). It means the formula indeed is a property of
a and b. So in formal set theory, when we say a property P on x1 , ..., xn , it always is referred
to a formula P (x1 , ..., xn ) having free variables x1 , ..., xn .
If := a(b (a, b)) = a(bx((x a x b) a x a)), then now every
variable is bounded. We call such a formula having no free variable a sentence in set theory. Believe or not, every mathematical statement can be translated into a single set theory
sentence. This is one of the reasons why we take set theory as the foundation of mathematics.
Now we introduce axioms and the rules of inferences for set theory. Axioms consist of
logical axioms and nonlogical axioms (ZFC axioms).
Logical Axioms: Any formulas of the following forms.
(1) ( ).13
(2) ( ( )) (( ) ( )).
(3) ( ) ( ).
(4) u( ) ( u) where u is not free in .
(5) u(u, u1 , ..., un ) (v, u1 , ..., un ), as far as v remains free at each replacement of
free occurrence of u in (u, u1 , ..., un ) by v.
(6) x = y (x z y z).
ZFC Axioms: We will state these later.
Rules of inferences:
(1) is obtained from and .
(2) u is obtained from .
We say a formula is a theorem (or equivalently is provable in set theory) if there is a
sequence of formulas 0 , ..., n such that = n and for each i n, i is either an axiom
(logical or ZFC), or obtained from previous formulas by the rules of inferences (i.e. for some
j, k < i, we have k = j i or i = uj ). Again all the mathematical statements so
13For the rest,

, >, ).

we omit some of parentheses in the formulas by taking linkage strength convention (, , >
25

far mathematicians have proved indeed have these proof sequences.

ZFC (Zermelo-Fraenkel + Choice) Axioms

Notation
abbreviates ,
abbreviates ( ),
abbreviates ( ) ( ).
u abbreviates u.
!u(u, v1 , ..., vn ) abbreviates u((u, v1 , ..., vn ) v((v, v1 , ..., vn ) u = v)) (There is
a unique u satisfying (u, v1 , ..., vn )).
B A abbreviates x(x B x A).
u+ abbreviates u {u}.
Axiom of Extensionality AB(A = B x(x A x B)).
Two sets are equal i they have exactly the same elements.
Axiom of Pairing abA z(z A z = a z = b).
a, b, the set A = {a, b} exists.

Intended meaning:

Intended meaning: For any sets

Axiom of Union
AB x(x B y(x y y A)). Intended meaning: For any set
A, there exists A = {x|y(x y y A)}.
Axiom of Power Set AP B(B P B A).
there is the power set P(A) = {B| B A}.

Intended meaning: For any set A,

Axiom Schema14 of Replacement15 For each formula (x, y), the following formula.
x!y(x, y) AB z(z B w(w A(w, z))). Intended meaning: If (x, y) satisfies the functional property, then for any set A, there is a set B = {z| w(w A(w, z))},
the image of A under .
Existence of the empty set Above axioms imply A x(x A x = x), the existence
of = {x|x = x}.
Axiom of Infinity A( A x(x A x+ A)).16
inductive set.

14Not

Intended meaning: There is an

a single axiom but a set of axioms.

the dierence with the one in the text book p. 112. These forms of Replacement Axioms imply
Comprehension Axioms(=Subset Axioms) (text book p. 8): For each formula (x) having a free
variable x, the formula AB x(x B x A (x)). Intended meaning: For each set A, there is a
set B = {x A| (x)}.
16The reader can easily figure out the full form of this abbreviated formula.
15Notice

26

Axiom of Foundation A(A = x(x A x A = )).

nonempty set A has an element disjoint with A.

Intended meaning: Any

Choice Axiom F[B, C(B, C F B = (B = C BC = )) C A(A F

!z(z A C))]. Intended meaning: If F is a family of nonempty mutually disjoint sets
(i.e. F is a partition), then there is a set C collecting exactly one element from each set in F.
The reader should be convinced English mathematical statements can be translated into
set theoretic sentences, and vice versa. For example, in ZFC, the class {x|x
/ x} does not
exist. It simply means yx(x y x
/ x) is provable. Comprehension Axiom only
allows the existence of a set B = {x A| x
/ x} from a set A. Now if B B then B must
satisfy the property x
/ x, so B
/ B, a contradiction. Hence we have B
/ B. But B
/B
need not imply B B. Thus Russells paradox is avoidable. Another example is Theorem
0.1. Namely,
AB(x(x A x B) x(x A (x A x B)))
is provable.

27