Professional Documents
Culture Documents
Set theory is the mathematics of infinity and part of the core curriculum for
mathematics majors. This book blends theory and connections with other parts of
mathematics so that readers can understand the place of set theory within the wider
context. Beginning with the theoretical fundamentals, the author proceeds to illustrate
applications to topology, analysis and combinatorics, as well as to pure set theory.
Concepts such as Boolean algebras, trees, games, dense linear orderings, ideals, filters
and club and stationary sets are also developed.
Pitched specifically at undergraduate students, the approach is neither esoteric nor
encyclopedic. The author, an experienced instructor, includes motivating examples and
over 100 exercises designed for homework assignments, reviews and exams. It is
appropriate for undergraduates as a course textbook or for self-study. Graduate
students and researchers will also find it useful as a refresher or to solidify their
understanding of basic set theory.
ERNEST SCHIMMERLING
Carnegie Mellon University, Pennsylvania
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, Sao Paulo, Delhi, Tokyo, Mexico City
Cambridge University Press
The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9781107008175
C E. Schimmerling 2011
A catalogue record for this publication is available from the British Library
1. The domain of R is
dom(R) = {x A | there exists y such that xRy}.
3. The range of R is
ran(R) = {y B | there exists x such that xRy}.
Exercises
Exercise 1.1 If R is a relation, then we dene
R1 = {(y, x) | xRy}.
Give an example where R is a function but R1 is not.
Exercise 1.2 How many functions whose domain is the empty
set are there? In other words, given a set B, how many functions
f : B are there?
Exercise 1.3 Explain why (x, y, z) = (x, (y, z)) is a reasonable
denition of an ordered triple.
1 Later in the book we will dene transitive set, which is dierent from transitive
relation. Unfortunately, it will be important to pay attention to this subtle
dierence in terminology.
2
ZFC
and
member = element.
x belongs to A,
x is an element of A and
x is a member of A,
Extensionality Axiom
This axiom says that two sets are equal if they have the same
members. Formally, it is written
A B [ x (x A x B) = A = B ] .
Because we dened
A B x (x A = x B) ,
another way to write the Extensionality Axiom is
A B [ (A B and B A) = A = B ] .
In other words, two sets are equal if each is a subset of the other.
By logic alone, if A = B, then A and B have the same members.
Combining this fact with the Extensionality Axiom, we have that
A B [ x (x A x B) A = B ] .
Equivalently,
A B [ (A B and B A) A = B ] .
Pairing Axiom
This axiom allows us to form singletons and unordered pairs. Its
formal statement is
x y !A z [z A (z = x or z = y)] .
If x = y, then we write {x, y} for the unique set whose only
members are x and y and call it an unordered pair. We always
ZFC 9
write {x} instead of {x, x} and call it a singleton. At this point,
it makes sense to dene the rst three natural numbers 0 = ,
1 = {0} and 2 = {0, 1}. We can also justify dening ordered pairs
by setting
(x, y) = {{x} , {x, y}}
whenever we are given x and y as we did in Denition 1.1. As a
reminder, when x = y, what we really have is
(x, x) = {{x}} .
Notice that, based on this denition, when we write (x, y), we can
tell that x is the rst coordinate and y is the second coordinate.
Formally, this means we can prove that for all x, y, x and y ,
(x, y) = (x , y ) (x = x and y = y ).
Union Axiom
This axiom allows us to form unions. Its formal statement is
F !A x [x A Y F (x Y )] .
We write F for the unique set whose members are exactly the
members of the members of F. In other words,
F = {x | there exists Y F such that x Y }.
It is important to note that, in the Union Axiom, the family F is
allowed to be innite. We often use dierent notation when F is
nite. For example, we dene
A B = {A, B}
and
ABC = {A, B, C}.
At this point, we can dene the remaining natural numbers
3 = 2 {2} = {0, 1, 2},
5 = 4 {4} = {0, 1, 2, 3, 4}
10 ZFC
and, in general,
n + 1 = n {n} = {0, . . . , n}.
V1 = P(V0 ) = {},
V2 = P(V1 ) = {, {}},
Comprehension Scheme
This axiom scheme gives us a way to form specic subsets of a
given set. It says the following.
For each property P (x), the following is an axiom:
A !B x [x B (x A and P (x))] .
Notice that the word property appears in quotes. There are
innitely many properties, which is why ZFC has innitely many
axioms. We will not give a formal denition of property because
it involves rst-order logic, which is not a prerequisite. It is enough
ZFC 11
for students in this course to depend on their intuition about the
meaning of property. Given a property P (x), we write
{x A | P (x)}
for the set of elements x of A for which P (x) is true. For example,
{x 10 | x is even} = {0, 2, 4, 6, 8}
and
{x 10 | x is odd} = {1, 3, 5, 7, 9}.
It is important to note that the Comprehension Scheme does not,
in general, permit us to dene sets by writing {x | P (x)}. In fact,
for some P (x), what we think of as {x | P (x)} is not a set. For
example {x | x = x} is not a set by the result in Exercise 2.6.
At this point, we are justied in making many familiar deni-
tions. For example, Denition 1.2 says that, given sets A and B,
we have the Cartesian product
A B = {(x, y) | x A and y B} .
In order to see that this is a legitimate denition, note that if
x A and y B, then
(x, y) = {{x} , {x, y}} P (P (A B)) .
Thus A B =
{z P (P (A B)) | z = (x, y) for some x A and y B} ,
which means the denition of A B is justied by a combination
of the axioms we have listed so far. See Exercise 2.4.
Innity Axiom
In an indirect way, this axiom tells us that the set of natural
numbers exists. Its formal statement is
I [ I and x (x I = x {x} I) ].
We say that a set I is inductive if I is a witness to the Innity
Axiom. In other words,
I is inductive [ I and x (x I = x {x} I) ].
12 ZFC
Proposition 2.1 There is a unique inductive set I such that
I J for every inductive set J.
The proof of Proposition 2.1 is broken up into smaller steps in
Exercise 2.5. The set of natural numbers is dened to be the unique
inductive set that is a subset of every other inductive set. We write
(lower case Greek omega) for the set of natural numbers. In
other words,
= {0, 1, 2, 3, . . . }.
It is also common for mathematicians to write N instead of
although we will not. Not only does the Innity Axiom allow us
to dene , it implies that we can prove statements by induction
on n and make recursive denitions for n just as you have
done in your other mathematics courses.
To see the relationship with induction, suppose we wish to prove
that a given property P (n) holds for every natural number n. By
Proposition 2.1 and the denition of , it would be enough to
show that {n | P (n)} is an inductive set. In other words,
show that P (0) holds and if P (n) holds, then so does P (n + 1).
Replacement Scheme
This is a scheme for generating innitely more axioms.
For each property P (x, y), the following is an axiom:
Foundation Axiom
This axiom says that if S is a non-empty set, then there exists
x S such that, for every y S, y x. In symbols,
S (S = = x S y S (y x)).
For example, should S = 2, the only witness to the Foundation
Axiom would be x = 0 because 2 = {0, 1} and 0 {0} = 1 but
1 = 0. On the other hand, when S = {0, {1}}, both elements
14 ZFC
of S satisfy the requirement of the Foundation Axiom because
0 {1} and {1} 0.
The Foundation Axiom turns out to be equivalent to the state-
ment that there is no sequence xn | n such that
xn+1 xn x1 x0 .
In particular, it implies that no set is an element of itself. For
otherwise, if x x, then
x x x x,
which contradicts the Foundation Axiom. We will make additional
comments about this axiom later but it is not a focus of the book.
Axiom of Choice
This axiom says that every family of sets has a choice function.
F function c A F (A = = c(A) A) .
The function c is called a choice function for F because it chooses
an element c(A) out of every non-empty A that belongs to F. From
experience, given nitely many non-empty sets, you can pick one
element from each. The Axiom of Choice says this is possible even
if you start with innitely many non-empty sets.
The rst time the Axiom of Choice is used is in Chapter 4.
There, it is essential for being able to assign a numerical cardinal-
ity (size) to each set, which is one of the most important things
we do in this book. As we go along, we will point out where and
how the Axiom of Choice is used.
Exercises
In the following exercises, you only need to be attentive to the
axioms of ZFC when so instructed, namely, in Exercises 2.1, 2.4,
2.6, the second part of 2.8 and the rst part of 2.11. Otherwise,
use the same style of mathematical argumentation as in your other
proof-based mathematics courses without reference to ZFC.
ZFC 15
Exercise 2.1 Prove that the following theories are equivalent.
1. Empty Set Axiom + Extensionality Axiom.
2. A x (x A) + Extensionality Axiom.
Hint: Obviously 1 implies 2. In the other direction, uniqueness is
the issue.
Exercise 2.2 Recall that 0 = and n + 1 = {0, . . . , n} for every
n . Recall also that V0 = and Vn+1 = P(Vn ) for every n .
1. Prove by induction that if n , then P(n) has 2n elements.
2. List the elements of V4 .
3. Make a conjecture regarding the size of Vn . Then prove your
conjecture by induction on n .
Exercise 2.3 Dene z to be a transitive set i for all x and y,
if x y and y z, then x z. This is equivalent to saying that,
for every y, if y z, then y z.
1. Prove that Vn is a transitive set for every n .
2. Prove that Vn Vn+1 for every n .
3. Prove that Vn = n for every n .
Exercise 2.4 Give a detailed explanation of how the denitions
of A B, dom(R) and R[S] given in Denitions 1.2 and 1.4 are
justied by the axioms of ZFC.
Exercise 2.5 Prove Proposition 2.1 by showing the following.
1. If I and J are inductive sets, then I J is an inductive set.
2. Suppose that K is an inductive set. Let
F = {J P(K) | J is an inductive set}
and
I = {x K | J F ( x J )} .
Prove the following statements.
(a) I is an inductive set.
(b) For every J, if J is an inductive set, then I J.
(c) Suppose that I is an inductive set and, for every inductive
set J,
I J.
Then I = I.
16 ZFC
Exercise 2.6 Prove that there is no set V such that x V for
every set x. (This shows that {x | x = x} is not a set.) If you
use the Foundation Axiom in your proof, then nd a second proof
that does not use the Foundation Axiom.
Exercise 2.7 In general, dene the intersection of a non-empty
family G to be
G = {a | X G ( a X )}.
B0 = A
and
Bn+1 = Bn .
Let
C= {Bn | n < }.
f (X B Y ) = f (X) C f (Y ),
f (B X) = C f (X),
f (B ) = C
and
f (B ) = C .
Example You should work out the details of the following asser-
tions before attempting the exercises that follow. Let S = and
put
B(S) = (P(S), , , , , S)
where X means S X in this context. Then B(S) is a Boolean
algebra. It is called the Boolean algebra of subsets of S. If S is
nite and has n elements, then P(S) has 2n elements hence B(S)
is nite. For B(S), the Boolean algebra relation amounts to
X Y X = X Y X Y.
The atoms of B(S) are exactly the singletons {a} for a S. The
function X S X is an isomorphism from B(S) to the Boolean
algebra
(P(S), , , , S, ).
Notice that union and intersection are exchanged, as are and S.
20 ZFC
Exercise 2.12 Consider an arbitrary nite Boolean algebra
B = (B, , , , , ).
Let S be the set of atoms of B.
1. Prove that if X B and X = , then there exists A S such
that A X.
2. Let X B. Suppose that X = and
{A S | A X} = {A1 , . . . , Am }.
Let
Y = A1 Am .
Prove
X = Y.
Hint: Certain basic facts about sets generalize to Boolean al-
gebras. For example, for sets we know that
if X Y and Y X, then X = Y ,
and for Boolean algebras we have that
if X Y and Y X, then X = Y .
The reason is that if X Y and Y X, then
X =X Y =Y X =Y
by the commutativity law for and the denition of . The
moral is that you should base your intuition about Boolean
algebras on what you already know about Boolean algebras of
sets. Of course, ultimately, you need to prove your intuition
is correct using just the laws of Boolean algebras. It should
also be said that the solution to this exercise is relatively long
so organizing your answer into a well-chosen series of lemmas
would be very helpful.
3. Let f : B P(S) be dened by
f (X) = {A S | A X}.
Prove that f is an isomorphism from B to B(S). Remark: This
explains why intuition coming from Boolean algebras of sets
really is valuable intuition about nite Boolean algebras.
ZFC 21
Exercise 2.13 As in Exercise 2.10, let E be the equivalence
relation on P() dened by
x E y x y is nite.
1. Prove that the following table of equations determines a Boolean
algebra B = (B, , , , , ).
B = P()/E
[x]E [y]E = [x y]E
[x]E [y]E = [x y]E
[x]E = [ x]E
= []E
= []E
Before proving the laws of Boolean algebras, you must show
that the operations , and are well-dened by the equations
listed above. So part of what you must show is that if x E x
and y E y , then
(x y) E (x y ),
(x y) E (x y )
and
( x) E ( x ).
Remark: This is an example of a quotient Boolean algebra. In
the literature, it is referred to as P()/Finite.
2. Prove that the Boolean algebra B that was dened in part 1
has no atoms.
3
Order
0th President
3.1 Wellorderings
Denition 3.1 Let A be a set and be a relation on A.
1. (A, ) is transitive i for all x, y, z A, if x y and y z,
then x z.
2. (A, ) is irreexive i for every x A, x x.
3. (A, ) is total i for all x, y A, either x y or x = y or
y x.
24 Order
4. (A, ) is a strict linear ordering i it is transitive, irreexive
and total.
V+1 = P (V )
and
V = {V | < }
if is a limit ordinal. If we really want to match up this denition
of V with Theorem 3.24, then we end up with G : V starting
with
if = 0
F (, g) = P(g()) if = + 1
{g() | < } if is a limit ordinal.
Notice that the denition V does not depend on so really what
we have is a way of assigning a set V to each ordinal . The
assignment V is not a function because what should be its
domain is not a set by the following lemma.
Lemma 3.25 There is no set of all ordinals.
Proof Suppose for contradiction that
= { | is an ordinal}
is a set. It is easy to see that is transitive and (, ) is a
wellordering. But then is an ordinal. In other words, .
But then
is an innite descending sequence of members of . This contra-
dicts that (, ) is a wellordering.
Notice that, once we realized that , we could have used the
Foundation Axiom to nish the proof of Lemma 3.25 slightly more
quickly. The Foundation Axiom automatically holds for ordinals,
which is essentially how we got away without using it above.
We have been discussing recursion; now let us see an example
of induction.
Lemma 3.26 Let be an ordinal and < . Then the following
hold.
3.2 Ordinal numbers 37
(1) V is a transitive set.
(2) For every < , V V .
+ 0 = ,
+ ( + 1) = ( + ) + 1 and
if is a limit ordinal, then + = sup ({ + | < }).
0 = 1,
for every ordinal ,
+1 =
and
for every limit ordinal ,
= sup | 0 < < .
Exercises
Exercise 3.1 Let (A, A ) be a wellordering such that A = .
For each y A, dene
pred(A,A ) (y) = {x A | x A y}.
Suppose that S A and, for all x, y A, if y S and x A y,
then x S. Prove that there exists y A such that
S = pred(A,A ) (y).
Exercise 3.2 Let (A, A ) and (B, B ) be wellorderings. Use
Theorem 3.28 to prove that exactly one of the following conditions
holds:
(A, A ) (B, B ).
There exists y B,
(A, A ) pred(B,B ) (y) ordered by B .
There exists y A,
(B, B ) pred(A,A ) (y) ordered by A .
Granted, it is strange to say that two sets have the same cardi-
nality without having said what cardinality means. But we need
a lemma before giving that denition.
Lemma 4.2 For every set A, there exists an ordinal such that
A.
F = {X A | X = } .
c(X) X.
54 Cardinality
Dene f () by recursion on ordinals as follows. If
A f [] F,
then let
f () = c (A f []) .
On the other hand, if
A f [] F,
then leave f () undened for every .
Observe that if < and both f () and f () are dened,
then f () f [] and f () A f [], hence f () = f (). This
calculation almost shows that f is an injection; what is missing is
a proof that f has a set domain.
First suppose there is an ordinal such that f () is undened.
Let be the least such ordinal. Then f is an injection with =
dom(f ) and ran(f ) A. Also,
A f [] F,
which means exactly that A f [] = . Equivalently, it means
that f [] = A. Therefore, f is a bijection from to A as desired.
It remains to see that there is an ordinal such that f () is
undened. Suppose otherwise. Then, for every ordinal , f () is
dened. Let
S = {a A | there exists such that f () = a}.
Then, for every a S, there exists a unique such that f () = a.
Apply the Replacement Scheme to this property to conclude that
there is a set and a function g : S such that, for every
a S, f (g(a)) = a. Because f () is dened for every ordinal ,
and because f () = f () whenever = , it must be that is
the set of all ordinals. But we proved earlier, in Lemma 3.25, that
there is no set of all ordinals.
We remark that the use of the Axiom of Choice cannot be elim-
inated from the proof of Lemma 4.2. Elaborating on the meaning
of this remark: If you remove the Axiom of Choice from ZFC the
result is known as ZF. It turns out that ZF + Lemma 4.2 implies
the Axiom of Choice. Prove this as a practice problem!
At last, we dene cardinal numbers and the cardinality of sets.
4.1 Cardinal numbers 55
Denition 4.3 is a cardinal i is an ordinal and, for every
< ,
Denition 4.4 |A| is the least ordinal such that A .
You should convince yourself of the following facts, whose proofs
amount to composing bijections.
Lemma 4.5 |A| is a cardinal.
Lemma 4.6 A B |A| = |B|.
Every natural number is both an ordinal and a cardinal. Also,
is a cardinal. However, the ordinals
+ 1, + 2, + 3, . . .
are all countably innite, which is to say that their cardinality is .
(For us, countable means nite or countably innite. Uncountable
means not countable.) In particular, the ordinals displayed above
are not cardinals. Moreover, the ordinals
2, 3, 4, . . .
are not cardinals. Nor are the ordinals
2 , 3 , 4 , . . . .
The ordinals
, , ...
are not cardinals either. All of these are ordinals are countable, as
you should try to verify. We need another idea to reach uncount-
able sets.
Theorem 4.7 (Cantor) There is no surjection from A to P(A).
Proof Consider an arbitrary function f : A P(A). Let
C = {x A | x f (x)}.
For every x A,
x C x f (x)
hence
C = f (x).
56 Cardinality
In particular,
C f [A].
Therefore f : A P(A) is not a surjection.
The proof we just gave is an example of a diagonal argument.
This is an imprecise term that you will see used in more and more
general ways throughout the book. Here, the idea is that if you
visualize the graph of the relation
{(x, y) A A | x f (y)},
then what we call the diagonal is
D = {x | x f (x)}.
To come up with a set C missing from the range of f , we take the
complement of the diagonal,
C = A D.
The reason C = f (x) is that one of C and f (x) has x as an element,
and the other does not. We can express the previous sentence in
terms of the symmetric dierence:
x C f (x)
hence
C f (x) =
thus
C = f (x).
Corollary 4.8 P() is uncountable.
Proof Clearly P() is innite. By Theorem 4.7, there is no sur-
jection from to P(). Hence there is no bijection.
Bijections are used in the denition of cardinality but sometimes
only surjections or injections are easily available. Here are some
basic facts that relate these notions.
Lemma 4.9 If < and there is a surjection f : , then
is not a cardinal.
4.1 Cardinal numbers 57
Proof Suppose that < and f : is a surjection. Let
S = { < | f () = f () for every < }.
Let g = f S. Then g is a bijection from S to . Let = type(S)
and : (S, <) be the Mostowski collapse of (S, <). By
Lemma 3.31, since S , . So < . Let
h = g 1 .
Then h : is a bijection. In other words . Thus is not
a cardinal.
Lemma 4.10 Let and be cardinals. Then
there is an injection f : .
Proof If , then the identity function is an injection from
to . For the reverse direction, suppose that f : is an
injection. Let
S = f [] = {f () | < }
and : (S, <) be the Mostowski collapse. Then the composi-
tion
f :
is a bijection. Thus || = ||. Since is a cardinal, || = . Because
S , by Lemma 3.31, . Putting these facts together we
have that
= || = || .
but keep in mind that what we have inside min() is not a set.
Corollary 4.16 We have that
0 = ,
+1 = ( )+ for every ordinal , and
= sup ({ | < }) for every limit ordinal .
Proof The rst two clauses are obvious. The last clause follows
from Lemma 4.14.
..
.
(0, + 1) _ _ _ _ _ _ _ _ _ _ _ _/ ( + 1, + 1)
(3 )
O
(0, ) _ _ _ _ _ _ _ _/ (, )
(1 )
O
(4 )
.. (2 )
.
(0, 0) (, 0) ( + 1, 0)
and
= .
There is a bijection
F :
dened by
F (g)(, ) = g()().
Putting together these observations, we are done.
Denition 4.29 (Strange notation) = for ordinals 1.
We write when we want to emphasize that it is an ordinal.
We write when we want to emphasize that it is a cardinal.
However, it is dicult to keep the notation consistent with such
intentions when we simultaneously consider cardinal and ordinal
properties of = . Ultimately, whether we write or
reduces to a matter of style.
Note that 0 is not dened; we always write either or 0 .
4.3 Conality
Recall that is the least cardinal greater than every n for n < .
In this sense, feels rather large. On the other hand, the function
n n
maps to and has range unbounded in , i.e.,
= sup n .
n<
cf ( + ) = 0
cf ( ) = 0
cf ( ) = 0 (ordinal exponentiation)
cf (1 + ) = 0
cf (1 + 1 ) = 1
cf (1 + 1 + ) = 0
we see that
cf( ) = 0 .
Example Using the fact that
+ = sup +n ,
n<
we see that
cf(+ ) = 0 .
Example Using the fact that
1 = sup ,
< 1
we see that
cf(1 ) 1 .
We claim that
cf(1 ) = 0 .
Consider an arbitrary function
f : 1 .
Dene g : 1 by letting
g(n) = the least < 1 such that f (n) < .
Then, for every n < ,
|f (n)| f (n) < g(n) < 1 .
Since 1 is a regular cardinal, there exists < 1 such that, for
every n < , g(n) < . Hence f (n) < for every n < . So
sup f (n) < 1 .
n<
Exercises
Exercise 4.1 Let
< n
2= 2.
n<
and
0 <0 = < .
1. 0 0 = 20
2. 1 0 = 20
3. 0 1 = 21
4. 1 1 = 21
74 Cardinality
Exercise 4.4 Let
<
=
<
and
< = <
+)
whenever and are innite cardinals. Show that (< = 2 .
Exercise 4.5 Prove that if are innite cardinals, then
|{X | |X| = }| = .
Then explain why
|{X 2 | |X| = 0 }| = max(2 , 20 ),
|{X 2 | |X| = 1 }| = 21
and
|{X 2 | |X| = 2 }| = 22 .
Exercise 4.6 Prove that, for every ordinal , there is a cardinal
> such that
cf() =
and
= .
Hint: First recall that for every ordinal by Lemma 4.19.
Now consider the sequence of cardinals n | n < dened by
induction according to 0 = +1 and n+1 = n .
Exercise 4.7 Express the cardinality of the sets below in the
form
, 2 , 22 , ...
and explain your calculations. For your solutions, you may use
facts about Q, R, C and continuous functions from calculus courses.
1. Q = {x | x is a rational number}
2. R = {x | x is a real number}
3. R Q = {x R | x is irrational}
4. {x R | 0 < x < 1}
5. {x C | x is a root of a polynomial with rational coecients}
4.3 Conality 75
6. R R = {f | f is a function from R to R}
7. Q R = {f | f is a function from Q to R}
8. R Q = {f | f is a function from R to Q}
9. {f | f is a continuous function from R to R}
Exercise 4.8 Let <R be the usual ordering of R. For every
A R, we may abuse notation by writing
(A, <R )
when we really mean
(A, {(x, y) A A | x <R y}).
You may use facts from calculus courses in your solutions to:
1. Prove that, for every A R, if (A, <R ) is a wellordering, then
type(A, <R ) < 1 .
2. Prove that, for every < 1 , there exists A R such that
type(A , <R ) = .
Exercise 4.9 Let f : B B be a function and X B. Prove
that there exists A B such that X A, f [A] A and
|A| |X| 0 .
Exercise 4.10 If Sn | n < is a sequence of sets, then dene
f is a function with dom(f ) =
Sn = f .
and f (n) Sn for all n <
n<
1. Show that
|P( )| P(n ) .
n<
2. Show that
|P( )| P(n ) .
n<
and
A = An .
n<
Dene
f (x) if x Aeven
1
h(x) = g (x) if x Aodd
f (x) if x A .
Prove that h is well-dened and h is a bijection from A to B.
Tn = (Tn , n , n , n , 0, 1)
as follows.
and
(n f )(a) = 1 f (a) = 0.
0 0 f (0, 0, . . . )
0 1 f (0, 1, . . . )
1 0 f (1, 0, . . . )
1 1 f (1, 1, . . . )
0 f (0, . . . )
1 f (1, . . . )
and
= n .
n<
80 Cardinality
To make sure you understand the denition of T , convince
yourself that if f Tm and g Tn where m < n < , then
f Tn and f g = f n g.
(a) It is a fact that T is a Boolean algebra. Pick any three of
the ten dening equations for Boolean algebras and show
that they hold for T . You may use the fact that Tn is a
Boolean algebra for n < .
(b) Prove that T has no atoms.
(c) Explain why T is countable.
(d) Give a specic example of a function f : 2 that does
not belong to T .
0 0 1
0 1 1
1 0 1
1 1 1
0 0 0 0 0 1 0 0 1 0 0 1
0 1 1 0 1 0 0 1 1 0 1 1
1 0 1 1 0 1 1 0 0 1 0 1
1 1 1 l5 1 1 1 1 1 1 1 1 0
x< lllll
x ll
xx lll
xxx lllll
x lll
0 0 0 ll 0 0 0 0 0 1 0 0 0 0 0 1 0 0 1
0 1 0 0 1 1 0 1 0 0 1 1 0 1 0 0 1 1
1 0 1 1 0 0 1 0 0 1 0 1 1 0 1 1 0 0
1 1 1 iRRR 1 1 1 1 1 1 1 1 0 5 1 1 0 1 1 0
RRR x< lll
RRR x l llll
RRR xx lll
RRR
RRR xx lll
RRR xx lllll
0 0 0 0 0 0 ll 0 0 0 0 0 1
0 1 0 0 1 0 0 1 1 0 1 0
1 0 0 1 0 1 1 0 0 1 0 0
1 1 1 1 1 0 1 1 0 1 1 0
0 0 0
0 1 0
1 0 0
1 1 0
Figure 4.2 What is the signicance of this picture? See Exercise 4.17.
5
Trees
R {0} = (, 0) (0, )
and
RZ= {(n, n + 1) | n Z}.
The following fact is left as an exercise; we will give a similar proof
in the next section.
Lemma 5.2 The family of open subsets of R is a topology on R.
Topological spaces are related to metric spaces, which we dene
next.
Denition 5.3 A metric space is a pair (S, d) where
d : S S [0, ) = {x R | 0 x}
is a function from S S to the set of non-negative real numbers
such that, for all x, y, z S,
d(x, y) = d(y, x),
d(x, y) = 0 x = y
84 Trees
and
d(x, z) d(x, y) + d(y, z).
S=R
and
d = |x y|.
|x y| 0,
|x y| = |y x|,
|x y| = 0 x = y
and
|x z| |x y| + |y z|.
{x R | |x c| < r}
>> x
>>
>>
>>
>>
>>
>>
>>
s
if s t, then Nt Ns ,
if t s, then Ns Nt , and
otherwise, Ns Nt = .
Denition 5.7 U is an open subset of i there is a family F
of basic open subsets of such that
U= F
86 Trees
Note that, because < is countable, U is an open subset of
i there is a sequence sn | n < from < such that
U= Ns n .
n<
and
B= G,
where F and G are families of basic open sets. We must show that
A B is an open set. Let H be the collection of basic open sets
Nt for which there are r, s < such that
Nr F and Ns G,
r s or s r, and
t = r s.
The last two clauses say that either
r = s dom(r)
or
s = r dom(s)
(we say r and s are comparable in this case), and t is the longer
of the two (which is the union because they are comparable). We
will be done when we show that
AB = H.
5.2 The Baire space 87
First suppose that x A B. Then there are r, s < such that
x Nr F
and
x Ns G.
Then the nite sequences r and s are comparable because they
are both restrictions of the same innite sequence x. That is,
r = x dom(r)
and
s = x dom(s).
Let t = r s be the longer of the two. Then
x Nt H,
so
x H.
This shows that
AB H.
We leave the easier reverse inclusion to the reader.
Denition 5.9 C is a closed subset of i C is an open
subset of .
It is time for some examples. Consider an arbitrary x . The
singleton {x} is closed since its complement is open:
{x} = {Ns | s < but s x}.
Exercises
Exercise 5.1 Let
I = {x | x is an injection from to }
and
S = {x | x is a surjection from to }.
Answer the following questions and prove your answer is correct.
92 Trees
1. Is I open?
2. Is I closed?
3. Is S open?
4. Is S closed?
The following exercises are about the Baire space with the metric
dened by
d(x, y) = 1/2n (x n = y n but x(n) = y(n))
and
d(x, y) = 0 x = y
but we remark that parts 1 and 2 hold in every metric space.
1. Let C . Prove that C is closed i for every sequence
xi | i <
from C and every y S, if
lim xi = y,
i
then y C. This says that a set is closed i it has all its limit
points.
2. Prove that if limi xi = y, then xi | i < is a Cauchy
sequence.
3. Prove that the Baire space is complete.
Prove that
|E| = 20 .
Hint: By Exercise 5.9, it is enough to show that there is a perfect
tree T such that [T ] E. Construct T using ideas similar to the
solution to Exercise 5.8.
T +1 = (T )
whenever is an ordinal, and
T = T .
<
T | < 20
x | < 20
and
y | < 20
y [T ] {x | < }
and
x [T ] {y | }.
.
..
3, 0, 0, 0
2, 0, 0 3, 0, 0
x 2 x 1 x 0.
Theorem 5.17 says that an innite tree with nite levels is ill-
founded. We will characterize wellfounded trees in terms of rank
functions.
and
.
..
0 1
0 1 2
0? 1 2 o3
??? ooooo
?? o
?? ooo
oooo
The examples above beg the question: which ordinals are the
ranks of trees on ? The answer is exactly the countable ordinals
by Theorem 5.23 and Exercise 5.20.
102 Trees
.
..
0 1
0 1 2
0 EE 1 2 m3
EE yyy mmmmm
EE y m
EE yy mm
E yymymmmm
m
+1
Exercises
Exercise 5.14 Let T be a tree on . Assume that [T ] = . Prove
that there exists a unique x [T ] such that, for all y [T ] and
n < , if y n = x n, then x(n) y(n). We call x the left-most
branch of T .
Exercise 5.15 You should notice that the proof of Theorem 5.17
is similar to the solution to Exercise 5.4(2). This exercise explains
why.
1. Use Lemma 5.16 and Corollary 5.18 to derive the fact that the
Cantor space is compact.
2. Use Lemma 5.16 and the fact that the Cantor space is compact
to derive Corollary 5.18.
5.3 Illfounded and wellfounded trees 103
Exercise 5.16 Recall that clopen means closed and open.
1. Prove that, for every C 2, if C is a clopen subset of the
Cantor space, then C is a union of nitely many basic open
subsets of the Cantor space. In other words, there is a nite set
{s0 , . . . , sn1 } <
such that
C= Nsi 2.
i<n
2. Let T be a tree on .
(a) Prove that if the restriction of <KB to T is a wellordering,
then T is wellfounded.
(b) Prove that if T is wellfounded, then the restriction of <KB
to T is a wellordering.
Exercise 5.19 Prove Theorem 5.23 in the following two steps.
1. Explain why the properties of T listed in the statement of
Theorem 5.23 form a legitimate denition by recursion on the
restriction of <KB to T .
2. Explain why the range of T is a countable ordinal.
104 Trees
Exercise 5.20 Prove by induction on < 1 that there exists
a wellfounded tree T on with rank(T ) = .
Exercise 5.21 Let T be a non-empty wellfounded tree on 2.
Prove that rank(T ) < .
X B Y = X Y,
X B Y = X Y,
B X = 2 X,
B =
and
B = 2.
Prove B is a countable atomless Boolean algebra.
Exercise 5.23 Let B be the Boolean algebra of clopen subsets
of the Baire space. That is,
B = {X | X is a clopen subset of },
X B Y = X Y,
X B Y = X Y,
B X = X,
B =
and
B = .
Prove B is an atomless Boolean algebra of cardinality 20 .
5.4 Innite games 105
5.4 Innite games
Let A . We describe a game, which is called GA . The game
has two players, I and II, who take turns playing natural numbers
x0 , x1 , etc. A run of the game GA looks as follows.
I x0 x2 x4
II x1 x3 x5
If x = xn | n < is a run of GA , then player I wins the run i
x A. Otherwise, player II wins the run.
This is a very general sort of game. Notice that nite games also
t this scheme because we may ignore moves after a winner has
been declared. (There are really two kinds of nite length games.
Either the length is xed in advance or else the length depends
on exactly how the players move. Both kinds of nite games can
be modeled with our innite games.) One dierence between our
games and some familiar games like chess is that we do not allow a
run of the game to end in a draw. This is because either x A or
x A. An arbitrary way to get around this objection is to declare
that draws go to player II. See Exercise 5.24 for more about chess.
Many properties in mathematics can be expressed in terms of
games, so general theorems about games can be quite useful. Along
these lines, a series of exercises at the end of this section illustrate
one of many ways in which games and mathematical analysis are
related.
Naturally, we are more interested in winning strategies than we
are in the player who wins a particular run of a particular game.
So let us continue making denitions associated to the game GA .
A strategy is a function : < . If is a strategy and
b , then b is the run that results if I uses and II plays b.
Formally, b is dened by recursion according to the equations
( b)2n+1 = bn
and
( b)2n = (( b) 2n).
See Figure 5.5 for another way of depicting the run b. We call
a winning strategy for player I i for every b ,
b A.
I ( ) (( ), b0 ) (( ), b0 , (( ), b0 ), b1 )
II b0 b1 b2
b A.
Thus is a winning strategy for player I in GA . So A is determined.
Observe that is not the only winning strategy for player I in GA ;
there are innitely many others.
This section includes two theorems about the determinacy of
games. The rst, Theorem 5.26, says that some games are unde-
termined. It will be apparent from our proof that there are 20
many strategies and 22 0 many games. In particular, there are
strictly more games than strategies. By itself, this is not an argu-
ment that there are undetermined games because some strategies
win in more than one game. For example, if is a winning strategy
for player I in GA and A B, then is also a winning strategy
for player I in GB . So more work than mere counting is needed to
see that there are undetermined games.
5.4 Innite games 109
Theorem 5.26 (GaleStewart) There is an undetermined subset
of .
Proof First we claim that if is a strategy, then the two functions
b b
and
a a
are injections from to itself. This claim is clear because b is
the sequence of odd values of b and a is the sequence of even
values of a .
Next observe that
| | = 0 0 = 20
and
< <
( 0 )
|{ | is a strategy}| = ( ) = 0 0 = 0 0 = 20 .
Say
{ | is a strategy} = | < 20 .
d(xn , y) 1/22n .
Hence
lim xn = y.
n
Since C is closed,
y C = U.
A more direct way to argue this last part is to note that, for every
n < ,
y 2n = xn 2n T
hence
y [T ] = C = U.
Exercises
Exercise 5.24 Use Theorem 5.27 to explain why, in chess, either
White has a winning strategy, or Black has a strategy to avoid
losing.
n1 = (s0 , n0 , s1 ),
n1 = (s0 , n0 , s1 , n1 , s2 ),
and so on for every ni with i < j. Now dene
p = s0
n0
. . .
sj1
nj1 .
Notice that p < 2 and the domain of p is its nite cardinality
|p | = j + |si |.
i<j
Exercise 5.33 Use Exercises 5.29, 5.31 and 5.32 to prove that
if C is a closed subset of 2, then either C is countable or C has
a perfect subset. Notice this also follows from the Cantor perfect
set theorem, which was the subject of Exercise 5.11, but the two
proofs are very dierent.
Exercise 5.36 Let 0 < m < . Show that Theorem 5.28 remains
true if 2 is replaced by m in its statement. Hint: Use induction
on m. The case m = 1 follows from the pigeonhole principle. (If
you partition an innite set into nitely many pieces, then one of
the pieces must be innite.) Think of the proof of Theorem 5.28 as
showing that the case m = 1 implies the case m = 2. Generalize
this to see that case m implies case m + 1.
5.6 Trees of uncountable height 119
Exercise 5.37 Prove that Theorem 5.28 becomes false if
is replaced by 1 and innite is replaced by uncountable
using the following example. Fix an injection
g : 1 R.
In order to avoid possible confusion, we write <R for the usual
order of R here. Dene
F : [1 ]2 2
by
0 if < and g() >R g()
F ({, }) =
1 if < and g() <R g().
such that
0 < |T | < 1
for every < 1 but T has no 1 -branch.
Proof Let
I = {s <1 | s is an injection}.
The good news is that I is obviously a subtree of <1 and I has
no 1 -branch because there is no injection from 1 to . The bad
news is that, whenever < 1 ,
|I | = |{s | s is an injection}| = 20 1 ,
which is too large. We will nd an appropriate subtree T I. For
< 1 and s, t , dene
s t |{ < | s() = t()}| < 0 .
It is easy to see that is an equivalence relation on .
Claim 5.30.1 There is a sequence s | < 1 so that, for
every < 1 ,
s I
and, for every < ,
s s .
Assuming Claim 5.30.1, if we dene
T = {t I | t s }
< 1
The good news is that t meets the rst and second requirements
for s of Claim 5.30.1, namely that
t I
and, for every < ,
s t .
The problem is that it might not satisfy the extra property that
| ran(t)| = 0 .
To x up this potential problem, put
t(2i ) if = i for some i <
s () =
t() if {i | i < }.
Then all three requirements on s are met.
Exercises
Exercise 5.38 Let T be the tree constructed in the proof of
Theorem 5.30. Suppose that < < 1 and u T . Explain
why there exists v T such that u = v . We say that u
has extensions to every level of T .
124 Trees
Exercise 5.39 Let T be the tree constructed in the proof of
Theorem 5.30. Suppose that < 1 and u T . Explain why
there exist > and v, w T such that
u=v=w
but
v = w.
We say that u splits in T .
This chapter is mainly about two theorems, one due to Cantor, the
other to Dedekind, which are characterizations of the rationals, Q,
and the reals, R, in terms of their respective orderings.
(f (z ), )
o f (z) /
aCC
CC
CC
CC
o / _o _ _ _ _ _ _ _ z _ _ _ _ _ _ _ _ _/
(0 , ) Q
Figure 6.1 R
Q R
Here, R
Q is the concatenation of R followed by Q. It is easy
to see that R
Q is a dense linear ordering without endpoints.
Suppose for contradiction that
f : R
Q R.
(0, ) Q (f (z), ).
The set on the left has cardinality 0 and the set on the right has
cardinality 20 , so there is no bijection between them. Figure 6.1
illustrates the point.
This brings up an interesting question: what about Q versus
Q
Q? In the next section, we will state and prove Theorem 6.5,
which implies that
Q Q
Q.
(/2, /2) Q Q,
f : (B, B ) (B , B )
f (x) = x.
and setting
L A,
L = ,
L = A,
for every x A and y L, if x A y, then x L, and
for every x L, there exists y L such that x A y.
The next to last condition says that left-cuts are closed to the left.
The last condition says that left-cuts do not have a right endpoint.
isomorphism
(B, B ) o_ _ _ _ _ _ g _ _ _ _ _ _ (D, D )
O O
countable dense subordering countable dense subordering
xL x
(A, A ) _ _ _ _ _ _ _ _ _ _ _ _/ (C, C )
isomorphism
Exercises
Exercise 6.1 The point of this exercise is to dene Q. Therefore,
you may not assume anything about Q, not even that Q exists,
in your solution. However, you may use standard properties of
with its order, addition and multiplication. For example, writing
3 6
2 = 4 is unacceptable at this point because we have not yet dened
fractions but writing 3 4 = 12 = 2 6 is acceptable because it only
refers to natural numbers and multiplication.
Let
S = ( 1).
F.
X F.
For all A, B X, if A F and A B, then B F.
For all A, B X, if A, B F, then A B F.
G is a lter,
A G or X A G , and
G G .
Base case = 0.
Dene G0 = F.
Successor case = + 1.
We break up this case into three subcases. The rst subcase is,
for every B G ,
B A = .
7.1 Motivation and denitions 145
In the rst subcase, let
G+1 = {C X | there exists B G such that B A C}.
Obvserve that G {A } G+1 because:
G G+1 since, if B G , then B A B, so B G+1 ,
and
A G+1 since X G , so X A G+1 .
Observe that G+1 is a lter because:
G+1 by the rst subcase hypothesis,
X G since X A X,
G+1 is clearly closed upward under , and
G+1 is closed under pairwise intersections since if B, C G ,
then
(B A ) (C A ) = (B C) A G+1 .
The second subcase is that the rst subcase fails and, for every
B G ,
B (X A ) = .
In the second subcase, let
G+1 = {C X | there exists B G such that B(XA ) C}.
Much like in the rst subcase, one shows that
G {X A } G+1
and G+1 is a lter. The reader should complete the verication.
The third subcase is that the rst and second subcases fail. We
will show this does not happen. For contradiction, suppose
B, C G ,
B A =
and
C (X A ) = .
Let D = B C. Then
D G ,
D A =
146 Filters and ideals
and
D (X A ) = .
Therefore D = . But G since G is a lter.
Limit case is a limit ordinal.
Dene
G = G .
<
/ (isomorphic to )
..
.
o / (isomorphic to Z)
..
.
o / (isomorphic to Z)
..
.
Clearly,
{n < | f (n) < b} = {n < | f (n) = a}.
a<b
{n < | f (n) = a} F.
[f ] = [n a].
E = { C | cf() = }
and
C E = { C | cf() = 1 }.
7.2 Club and stationary sets 157
1. Prove that C is club in 2 . (This is pretty obvious.)
2. Prove that E is stationary in 2 .
3. Prove that C E is stationary in 2 .
4. Use parts 1, 2 and 3 to prove that the club lter over 2 is not
an ultralter.
Now we come to one of the most fundamental tools for studying
club and stationary sets.
Theorem 7.11 (Fodor) Let be uncountable regular cardinal
and f : be a function such that
{ < | f () < }
is stationary in . Then there exists < such that
{ < | f () = }
is stationary in .
We will derive Theorem 7.11 from Lemma 7.12, which is inter-
esting in its own right. Exercise 7.9 tells us that the intersection
of many club sets might not be club. Lemma 7.12 says that the
diagonal intersection of many club sets is club.
Lemma 7.12 Let be an uncountable regular cardinal. Suppose
that
C | <
is a sequence of club subsets of . Let
D = { < | C for every < }.
Then D is club in .
We call D the diagonal intersection of C | < . Most
commonly, you will see it written < C .
Proof of Lemma 7.12 First we show that D is closed in . Let
< . Assume that D = and let = sup(D ). We
must show that D. For contradiction, suppose that D. It
follows easily that is a limit ordinal and D = D . Hence
= sup(D ).
By the denition of D, there exists < such that C . Since
C is closed, there are two cases:
158 Filters and ideals
1. C = .
2. C = and sup(C ) C .
In the second case, sup(C ) < because C . In either
case, we may pick D such that < < . In the second
case, we can also make sure that sup(C ) < . Then, in both
cases, D and < but C . This directly contradicts the
denition of D.
Now we show that D is unbounded in . By recursion on <
dene as follows. Let 0 = 1. Given < , pick +1 >
such that
+1 C .
<
1 Here and below, we cite only the most recent edition available.
Selected further reading 165
the former ties set theory to analysis more than logic, while for
the latter it is the other way around.
Yet another subject that is intertwined with set theory and logic
is model theory. In Chapter 6, we saw examples of classication up
to isomorphism. This idea is important throughout mathematics
but especially in model theory. We also touched on ultraproducts
in Section 7.1. This is a model-theoretic construction that has
applications in many elds, particularly in set theory. A classic
beginning graduate model theory text is Chang and Keisler (1990).
Set theory is a vast topic of current mathematical research.
The enormous Handbook of set theory, edited by Foreman and
Kanamori (2010), comes in three volumes with a total of twenty-
four chapters by various authors. It suces to give the reader an
accurate impression of the many directions the subject has taken
in recent decades.
Bibliography
Chang, C. C., and Keisler, H. J. 1990. Model theory. Third edn. Studies in
Logic and the Foundations of Mathematics, vol. 73. Amsterdam: North-
Holland.
Enderton, Herbert B. 1977. Elements of set theory. New York: Academic
Press [Harcourt Brace Jovanovich Publishers].
Enderton, Herbert B. 2001. A mathematical introduction to logic. Second edn.
Harcourt/Academic Press, Burlington, MA.
Foreman, Matthew, and Kanamori, Akihiro (eds). 2010. Handbook of set the-
ory. New York: Springer-Verlag. In three volumes.
Goldstern, Martin, and Judah, Haim. 1998. The incompleteness phenomenon.
Natick, MA: A. K. Peters Ltd. Reprint of the 1995 original.
Hrbacek, Karel, and Jech, Thomas. 1999. Introduction to set theory. Third
edn. Monographs and Textbooks in Pure and Applied Mathematics, vol.
220. New York: Marcel Dekker Inc.
Jech, Thomas. 2003. Set theory. Springer Monographs in Mathematics. Berlin:
Springer-Verlag. The third millennium edition, revised and expanded.
Kechris, Alexander S. 1995. Classical descriptive set theory. Graduate Texts
in Mathematics, vol. 156. New York: Springer-Verlag.
Kunen, Kenneth. 1983. Set theory. Studies in Logic and the Foundations of
Mathematics, vol. 102. Amsterdam: North-Holland. Reprint of the 1980
original.
Moschovakis, Yiannis N. 2009. Descriptive set theory. Second edn. Mathe-
matical Surveys and Monographs, vol. 155. Providence, RI: American
Mathematical Society.
Royden, H. L. 1988. Real analysis. Third edn. New York: Macmillan.
Rudin, Walter. 1976. Principles of mathematical analysis. Third edn. New
York: McGraw-Hill. International Series in Pure and Applied Mathe-
matics.
Index