pdf version of the entry

Set Theory
http://plato.stanford.edu/archives/win2014/entries/set-theory/ Set Theory
from the Winter 2014 Edition of the First published Wed Oct 8, 2014

Stanford Encyclopedia Set theory is the mathematical theory of well-determined collections,

called sets, of objects that are called members, or elements, of the set. Pure
of Philosophy set theory deals exclusively with sets, so the only sets under consideration
are those whose members are also sets. The theory of the hereditarily-
finite sets, namely those finite sets whose elements are also finite sets, the
elements of which are also finite, and so on, is formally equivalent to
arithmetic. So, the essence of set theory is the study of infinite sets, and
therefore it can be defined as the mathematical theory of the actual—as
Edward N. Zalta Uri Nodelman Colin Allen R. Lanier Anderson
opposed to potential—infinite.
Principal Editor Senior Editor Associate Editor Faculty Sponsor
Editorial Board
The notion of set is so simple that it is usually introduced informally, and
http://plato.stanford.edu/board.html
regarded as self-evident. In set theory, however, as is usual in
Library of Congress Catalog Data mathematics, sets are given axiomatically, so their existence and basic
ISSN: 1095-5054
properties are postulated by the appropriate formal axioms. The axioms of
Notice: This PDF version was distributed by request to mem- set theory imply the existence of a set-theoretic universe so rich that all
bers of the Friends of the SEP Society and by courtesy to SEP mathematical objects can be construed as sets. Also, the formal language
content contributors. It is solely for their fair use. Unauthorized of pure set theory allows to formalize all mathematical notions and
distribution is prohibited. To learn how to join the Friends of the arguments. Thus, set theory has become the standard foundation for
SEP Society and obtain authorized PDF versions of SEP entries, mathematics, as every mathematical object can be viewed as a set, and
please visit https://leibniz.stanford.edu/friends/ . every theorem of mathematics can be logically deduced in the Predicate
Calculus from the axioms of set theory.
Stanford Encyclopedia of Philosophy
Copyright c 2011 by the publisher
The Metaphysics Research Lab Both aspects of set theory, namely, as the mathematical science of the
Center for the Study of Language and Information infinite, and as the foundation of mathematics, are of philosophical
Stanford University, Stanford, CA 94305
importance.
Set Theory
Copyright c 2014 by the author 1. The origins
Joan Bagaria
2. The axioms of set theory
All rights reserved.
Copyright policy: https://leibniz.stanford.edu/friends/info/copyright/

1
 Set Theory Joan Bagaria

2.1 The axioms of ZFC real numbers are both infinite, there are more real numbers than there are
3. The theory of transfinite ordinals and cardinals natural numbers, which opened the door to the investigation of the
3.1 Cardinals different sizes of infinity. See the entry on the early development of set
4. The universe V of all sets theory for a discussion of the origin of set-theoretic ideas and their use by
5. Set theory as the foundation of mathematics different mathematicians and philosophers before and around Cantor's
5.1 Metamathematics time.
5.2 The incompleteness phenomenon
6. The set theory of the continuum According to Cantor, two sets A and B have the same size, or cardinality,
6.1 Descriptive Set Theory if they are bijectable, i.e., the elements of A can be put in a one-to-one
6.2 Determinacy correspondence with the elements of B. Thus, the set ℕ of natural numbers
6.3 The Continuum Hypothesis and the set ℝ of real numbers have different cardinalities. In 1878 Cantor
7. Gödel's constructible universe formulated the famous Continuum Hypothesis (CH), which asserts that
8. Forcing every infinite set of real numbers is either countable, i.e., it has the same
8.1 Other applications of forcing cardinality as ℕ, or has the same cardinality as ℝ. In other words, there
9. The search for new axioms are only two possible sizes of infinite sets of real numbers. The CH is the
10. Large cardinals most famous problem of set theory. Cantor himself devoted much effort to
10.1 Inner models of large cardinals it, and so did many other leading mathematicians of the first half of the
10.2 Consequences of large cardinals twentieth century, such as Hilbert, who listed the CH as the first problem
11. Forcing axioms in his celebrated list of 23 unsolved mathematical problems presented in
Bibliography 1900 at the Second International Congress of Mathematicians, in Paris.
Academic Tools The attempts to prove the CH led to major discoveries in set theory, such
Other Internet Resources as the theory of constructible sets, and the forcing technique, which
Related Entries showed that the CH can neither be proved nor disproved from the usual
axioms of set theory. To this day, the CH remains open.

1. The origins Early on, some inconsistencies, or paradoxes, arose from a naive use of
the notion of set; in particular, from the deceivingly natural assumption
Set theory, as a separate mathematical discipline, begins in the work of that every property determines a set, namely the set of objects that have
Georg Cantor. One might say that set theory was born in late 1873, when the property. One example is Russell's Paradox, also known to Zermelo:
he made the amazing discovery that the linear continuum, that is, the real
consider the property of sets of not being members of themselves.
line, is not countable, meaning that its points cannot be counted using the
If the property determines a set, call it A, then A is a member of
natural numbers. So, even though the set of natural numbers and the set of

2 Stanford Encyclopedia of Philosophy Winter 2014 Edition 3

 Set Theory Joan Bagaria

itself if and only if A is not a member of itself. 2. The axioms of set theory
Thus, some collections, like the collection of all sets, the collection of all ZFC is an axiom system formulated in first-order logic with equality and
ordinals numbers, or the collection of all cardinal numbers, are not sets. with only one binary relation symbol ∈ for membership. Thus, we write
Such collections are called proper classes. A ∈ B to express that A is a member of the set B. See the supplement
Basic Set Theory for further details. See also the supplement Zermelo-
In order to avoid the paradoxes and put it on a firm footing, set theory had
Fraenkel Set Theory for a formalized version of the axioms and further
to be axiomatized. The first axiomatization was due to Zermelo (1908) and
comments. We state below the axioms of ZFC informally.
it came as a result of the need to spell out the basic set-theoretic principles
underlying his proof of Cantor's Well-Ordering Principle. Zermelo's 2.1 The axioms of ZFC
axiomatization avoids Russell's Paradox by means of the Separation
axiom, which is formulated as quantifying over properties of sets, and thus Extensionality: If two sets A and B have the same elements, then they
it is a second-order statement. Further work by Skolem and Fraenkel led to are equal.
the formalization of the Separation axiom in terms of formulas of first-
order, instead of the informal notion of property, as well as to the Null Set: There exists a set, denoted by ∅ and called the empty set,
introduction of the axiom of Replacement, which is also formulated as an which has no elements.
axiom schema for first-order formulas (see next section). The axiom of
Pair: Given any sets A and B, there exists a set, denoted by {A, B},
Replacement is needed for a proper development of the theory of
which contains A and B as its only elements. In particular, there
transfinite ordinals and cardinals, using transfinite recursion (see Section
exists the set {A} which has A as its only element.
3). It is also needed to prove the existence of such simple sets as the set of
hereditarily finite sets, i.e., those finite sets whose elements are finite, the Power Set: For every set A there exists a set, denoted by (A) and
elements of which are also finite, and so on; or to prove basic set-theoretic called the power set of A, whose elements are all the subsets of A.
facts such as that every set is contained in a transitive set, i.e., a set that
contains all elements of its elements (see Mathias 2001 for the weaknesses Union: For every set A, there exists a set, denoted by ⋃ A and called
of Zermelo set theory). A further addition, by von Neumann, of the axiom the union of A, whose elements are all the elements of the elements
of Foundation, led to the standard axiom system of set theory, known as of A.
the Zermelo-Fraenkel axioms plus the Axiom of Choice, or ZFC.
Infinity: There exists an infinite set. In particular, there exists a set Z
Other axiomatizations of set theory, such as those of von Neumann- that contains ∅ and such that if A ∈ Z , then ⋃{A, {A}} ∈ Z.
Bernays-Gödel (NBG), or Morse-Kelley (MK), allow also for a formal
treatment of proper classes. Separation: For every set A and every given property, there is a set
containing exactly the elements of A that have that property. A

4 Stanford Encyclopedia of Philosophy Winter 2014 Edition 5

 Set Theory Joan Bagaria

property is given by a formula φ of the first-order language of set Principle, which asserts that every set can be well-ordered, i.e., it can be
theory. linearly ordered so that every non-empty subset has a minimal element.

Thus, Separation is not a single axiom but an axiom schema, that is, Although not formally necessary, besides the symbol ∈ one normally uses
an infinite list of axioms, one for each formula φ. for convenience other auxiliary defined symbols. For example, A ⊆ B
expresses that A is a subset of B, i.e., every member of A is a member of
Replacement: For every given definable function with domain a set B. Other symbols are used to denote sets obtained by performing basic
A, there is a set whose elements are all the values of the function. operations, such as A ∪ B, which denotes the union of A and B, i.e., the set
whose elements are those of A and B; or A ∩ B, which denotes the
Replacement is also an axiom schema, as definable functions are
intersection of A and B, i.e., the set whose elements are those common to
given by formulas.
A and B. The ordered pair (A, B) is defined as the set {{A}, {A, B}}.
Foundation: Every non-empty set A contains an ∈-minimal element, Thus, two ordered pairs (A, B) and (C, D) are equal if and only if A = C
that is, an element such that no element of A belongs to it. and B = D. And the Cartesian product A × B is defined as the set of all
ordered pairs (C, D) such that C ∈ A and D ∈ B. Given any formula
These are the axioms of Zermelo-Fraenkel set theory, or ZF. The axioms φ(x, y1 , … , yn ), and sets A, B1 , … , Bn, one can form the set of all those
of Null Set and Pair follow from the other ZF axioms, so they may be elements of A that satisfy the formula φ(x, B1 , … , Bn ). This set is denoted
omitted. Also, Replacement implies Separation. by {a ∈ A : φ(a, B1 , … , Bn )}. In ZF one can easily prove that all these
sets exist. See the supplement on Basic Set Theory for further discussion.
Finally, there is the Axiom of Choice (AC):

Choice: For every set A of pairwise-disjoint non-empty sets, there 3. The theory of transfinite ordinals and cardinals
exists a set that contains exactly one element from each set in A .
In ZFC one can develop the Cantorian theory of transfinite (i.e., infinite)
The AC was, for a long time, a controversial axiom. On the one hand, it is ordinal and cardinal numbers. Following the definition given by Von
very useful and of wide use in mathematics. On the other hand, it has Neumann in the early 1920s, the ordinal numbers, or ordinals, for short,
rather unintuitive consequences, such as the Banach-Tarski Paradox, are obtained by starting with the empty set and performing two operations:
which says that the unit ball can be partitioned into finitely-many pieces, taking the immediate successor, and passing to the limit. Thus, the first
which can then be rearranged to form two unit balls. The objections to the ordinal number is ∅. Given an ordinal α , its immediate successor, denoted
axiom arise from the fact that it asserts the existence of sets that cannot be by α + 1, is the set α ∪ {α} . And given a non-empty set X of ordinals
explicitly defined. But Gödel's 1938 proof of its consistency, relative to such that for every α ∈ X the successor α + 1 is also in X, one obtains the
the consistency of ZF, dispelled any suspicions left about it. limit ordinal ⋃ X. One shows that every ordinal is (strictly) well-ordered
by ∈, i.e., it is linearly ordered by ∈ and there is no infinite ∈-descending
The Axiom of Choice is equivalent, modulo ZF, to the Well-ordering sequence. Also, every well-ordered set is isomorphic to a unique ordinal,

6 Stanford Encyclopedia of Philosophy Winter 2014 Edition 7

 Set Theory Joan Bagaria

called its order-type. above are either finite or countable. But the set of all finite and countable
ordinals is also an ordinal, called ω1 , and is not countable. Similarly, the
Note that every ordinal is the set of its predecessors. However, the class set of all ordinals that are bijectable with some ordinal less than or equal to
ON of all ordinals is not a set. Otherwise, ON would be an ordinal greater ω1 is also an ordinal, called ω2 , and is not bijectable with ω1 , and so on.
than all the ordinals, which is impossible. The first infinite ordinal, which
is the set of all finite ordinals, is denoted by the Greek letter omega (ω). In 3.1 Cardinals
ZFC, one identifies the finite ordinals with the natural numbers. Thus,
∅ = 0, {∅} = 1, {∅, {∅}} = 2, etc., hence ω is just the set ℕ of natural A cardinal is an ordinal that is not bijectable with any smaller ordinal.
numbers. Thus, every finite ordinal is a cardinal, and ω, ω1 , ω2 , etc. are also
cardinals. The infinite cardinals are represented by the letter aleph (ℵ) of
One can extend the operations of addition and multiplication of natural the Hebrew alphabet, and their sequence is indexed by the ordinals. It
numbers to all the ordinals. For example, the ordinal α + β is the order- starts like this
type of the well-ordering obtained by concatenating a well-ordered set of
order-type α and a well-ordered set of order-type β . The sequence of ℵ0 , ℵ1 , ℵ2 , …, ℵω , ℵω+1 , …, ℵω+ω, …, ℵω2 , …, ℵωω , …, ℵω1 ,
ordinals, well-ordered by ∈, starts as follows …, ℵω2 , …

0, 1, 2,…, n,…, ω, ω + 1, ω + 2,…, ω + ω,…, n ⋅ ω, …, ω ⋅ ω,…, Thus, ω = ℵ0, ω1 = ℵ1 , ω2 = ℵ2 , etc. For every cardinal there is a
ωn , …, ωω , … bigger one, and the limit of an increasing sequence of cardinals is also a
cardinal. Thus, the class of all cardinals is not a set, but a proper class.
The ordinals satisfy the principle of transfinite induction: suppose that C
is a class of ordinals such that whenever C contains all ordinals β smaller An infinite cardinal κ is called regular if it is not the union of less than κ
than some ordinal α , then α is also in C. Then the class C contains all smaller cardinals. Thus, ℵ0 is regular, and so are all infinite successor
ordinals. Using transfinite induction one can prove in ZFC (and one needs cardinals, such as ℵ1 . Non-regular infinite cardinals are called singular.
the axiom of Replacement) the important principle of transfinite The first singular cardinal is ℵω , as it is the union of countably-many
recursion, which says that, given any definable class-function G : V → V , smaller cardinals, namely ℵω = ⋃n<ω ℵn .
one can define a class-function F : ON → V such that F(α) is the value of
The cofinality of a cardinal κ, denoted by cf (κ) is the smallest cardinal λ
the function G applied to the function F restricted to α. One uses
such that κ is the union of λ-many smaller ordinals. Thus, cf (ℵω ) = ℵ0.
transfinite recursion, for example, in order to define properly the
arithmetical operations of addition, product, and exponentiation on the By the AC (in the form of the Well-Ordering Principle), every set A can
ordinals. be well-ordered, hence it is bijectable with a unique cardinal, called the
cardinality of A. Given two cardinals κ and λ , the sum κ + λ is defined as
Recall that an infinite set is countable if it is bijectable, i.e., it can be put
the cardinality of the set consisting of the union of any two disjoint sets,
into a one-to-one correspondence, with ω. All the ordinals displayed

8 Stanford Encyclopedia of Philosophy Winter 2014 Edition 9

 Set Theory Joan Bagaria

one of cardinality κ and one of cardinality λ. And the product κ ⋅ λ is Vα = ⋃β<α Vβ , whenever α is a limit ordinal.
defined as the cardinality of the Cartesian product κ × λ. The operations of
sum and product of infinite cardinals are trivial, for if κ and λ are infinite The Power Set axiom is used to obtain Vα+1 from Vα . Replacement and
cardinals, then κ + λ = κ ⋅ λ = maximum{κ, λ}. Union allow one to form Vα for α a limit ordinal. Indeed, consider the
function that assigns to each β < α the set Vβ . By Replacement, the
In contrast, cardinal exponentiation is highly non-trivial, for even the collection of all the Vβ , for β < α , is a set, hence the Union axiom applied
value of the simplest non-trivial infinite exponential, namely 2ℵ0 , is not to that set yields Vα. The axiom of Infinity is needed to prove the existence
known and cannot be determined in ZFC (see below). The cardinal κ λ is of ω and hence of the transfinite sequence of ordinals. Finally, the axiom
defined as the cardinality of the Cartesian product of λ copies of κ; of Foundation is equivalent, assuming the other axioms, to the statement
equivalently, as the cardinality of the set of all functions from λ into κ. that every set belongs to some Vα, for some ordinal α. Thus, ZF proves
König's theorem asserts that κ cf(κ) > κ, which implies that the cofinality of that the set theoretic universe, denoted by V , is the union of all the Vα , α
the cardinal 2ℵ0 , whatever that cardinal is, must be uncountable. But this is an ordinal.
essentially all that ZFC can prove about the value of the exponential 2ℵ0 .
The proper class V , together with the ∈ relation, satisfies all the ZFC
In the case of exponentiation of singular cardinals, ZFC has a lot more to axioms, and is thus a model of ZFC. It is the intended model of ZFC, and
say. In 1989, Shelah proved the remarkable result that if ℵω is a strong one may think of ZFC as providing a description of V , a description
limit, that is, 2ℵn < ℵω , for every n < ω, then 2ℵω < ℵω4 (see Shelah however that is highly incomplete, as we shall see below.
(1994)). The technique developed by Shelah to prove this and similar
theorems, in ZFC, is called pcf theory (for possible cofinalities), and has One ZFC-provable important property of V is the so-called Reflection
found many applications in other areas of mathematics. Principle. Namely, for each formula φ(x1 , … , xn ), ZFC proves that there
exists an ordinal α such that Vα reflects it, that is, for every
4. The universe V of all sets a1 , … , an ∈ Vα,

A posteriori, the ZF axioms other than Extensionality—which needs no φ(a1 , … , an ) holds in V if and only if φ(a1 , … , an ) holds in Vα.
justification because it just states a defining property of sets—may be
Thus, V cannot be characterized by any sentence, as any sentence that is
justified by their use in building the cumulative hierarchy of sets. Namely,
true in V must be also true in some initial segment Vα. In particular, ZFC
in ZF we define using transfinite recursion the class-function that assigns
is not finitely axiomatizable, for otherwise ZFC would prove that, for
to each ordinal α the set Vα , given as follows:
unboundedly many ordinals α , Vα is a model of ZFC, contradicting
V0 = ∅ Gödel's second incompleteness theorem (see Section 5.2).

Vα+1 = (Vα ) The Reflection Principle encapsulates the essence of ZF set theory, for it is
equivalent to the axioms of Infinity plus Replacement, modulo the other

10 Stanford Encyclopedia of Philosophy Winter 2014 Edition 11

 Set Theory Joan Bagaria

ZF axioms. Algebraic structures can also be viewed as sets, as any n -ary relation on
the elements of a set A can be viewed as a set of n-tuples (a1 , … , an ) of
5. Set theory as the foundation of mathematics elements of A. And any n-ary function f on A, with values on some set B,
can be seen as the set of n + 1 -tuples ((a1 , … , an ), b) such that b is the
Every mathematical object may be viewed as a set. For example, the value of f on (a1 , … , am ). Thus, for example, a group is just a triple
natural numbers are identified with the finite ordinals, so ℕ = ω. The set (A, +, 0), where A is a non-empty set, + is a binary function on A that is
of integers ℤ may be defined as the set of equivalence classes of pairs of associative, 0 is an element of A such that a + 0 = 0 + a = a, for all
natural numbers under the equivalence relation (n, m) ≡ (n′ , m′) if and a ∈ A, and for every a ∈ A there is an element of A, denoted by −a, such
only if n + m′ = m + n′. By identifying every natural number n with the that a + (−a) = (−a) + a = 0. Also, a topological space is just a set X
equivalence class of the pair (n, 0), one may extend naturally the together with a topology τ on it, i.e., τ is a subset of (X) containing X
operations of sum and product of natural numbers to ℤ (see Enderton and ∅, and closed under arbitrary unions and finite intersections. Any
(1977) for details, and Levy (1979) for a different construction). Further, mathematical object whatsoever can always be viewed as a set, or a proper
one may define the rationals ℚ as the set of equivalence classes of pairs class. The properties of the object can then be expressed in the language of
(n, m) of integers, where m ≠ 0, under the equivalence relation set theory. Any mathematical statement can be formalized into the
(n, m) ≡ (n′ , m′) if and only if n ⋅ m′ = m ⋅ n′ . Again, the operations + language of set theory, and any mathematical theorem can be derived,
and ⋅ on ℤ may be extended naturally to ℚ. Moreover, the ordering ≤ℚ on using the calculus of first-order logic, from the axioms of ZFC, or from
the rationals is given by: r ≤ℚ s if and only if there exists t ∈ ℚ such that some extension of ZFC. It is in this sense that set theory provides a
s = r + t. The real numbers may be defined as Dedekind cuts of ℚ, foundation for mathematics.
namely, a real number is given by a pair (A, B) of non-empty disjoint sets
such that A ∪ B = ℚ, and a ≤ℚ b for every a ∈ A and b ∈ B. One can The foundational role of set theory for mathematics, while significant, is
then extend again the operations of + and ⋅ on ℚ, as well as the ordering by no means the only justification for its study. The ideas and techniques
≤ℚ, to the set of real numbers ℝ. developed within set theory, such as infinite combinatorics, forcing, or the
theory of large cardinals, have turned it into a deep and fascinating
Let us emphasize that it is not claimed that, e.g., real numbers are mathematical theory, worthy of study by itself, and with important
Dedekind cuts of rationals, as they could also be defined using Cauchy applications to practically all areas of mathematics.
sequences, or in other different ways. What is important, from a
foundational point of view, is that the set-theoretic version of ℝ, together 5.1 Metamathematics
with the usual algebraic operations, satisfies the categorical axioms that
the real numbers satisfy, namely those of a complete ordered field. The The remarkable fact that virtually all of mathematics can be formalized
metaphysical question of what the real numbers really are is irrelevant within ZFC, makes possible a mathematical study of mathematics itself.
here. Thus, any questions about the existence of some mathematical object, or
the provability of a conjecture or hypothesis can be given a

12 Stanford Encyclopedia of Philosophy Winter 2014 Edition 13

 Set Theory Joan Bagaria

mathematically precise formulation. This makes metamathematics consistency or of undecidability of a given sentence φ is always a relative
possible, namely the mathematical study of mathematics itself. So, the consistency proof. That is, one assumes that ZFC is consistent, hence it
question about the provability or unprovability of any given mathematical has a model, and then one constructs another model of ZFC where the
statement becomes a sensible mathematical question. When faced with an sentence φ is true. We shall see several examples in the next sections.
open mathematical problem or conjecture, it makes sense to ask for its
provability or unprovability in the ZFC formal system. Unfortunately, the 6. The set theory of the continuum
answer may be neither, because ZFC, if consistent, is incomplete.
From Cantor and until about 1940, set theory developed mostly around the
5.2 The incompleteness phenomenon study of the continuum, that is, the real line ℝ. The main topic was the
study of the so-called regularity properties, as well as other structural
Gödel's completeness theorem for first-order logic implies that ZFC is properties, of simply-definable sets of real numbers, an area of
consistent—i.e., no contradiction can be derived from it—if and only if it mathematics that is known as Descriptive Set Theory.
has a model. A model of ZFC is a pair (M, E), where M is a non-empty set
and E is a binary relation on M such that all the axioms of ZFC are true 6.1 Descriptive Set Theory
when interpreted in (M, E), i.e., when the variables that appear in the
axioms range over elements of M, and ∈ is interpreted as E. Thus, if φ is a Descriptive Set Theory is the study of the properties and structure of
sentence of the language of set theory and one can find a model of ZFC in definable sets of real numbers and, more generally, of definable subsets of
which φ holds, then its negation ¬φ cannot be proved in ZFC. Hence, if ℝn and other Polish spaces (i.e., separable, metric, and complete), such as
one can find a model of φ and also a model of ¬φ, then φ is neither the Baire space  of all functions f : ℕ → ℕ , the space of complex
provable nor disprovable in ZFC, in which case we say that φ is numbers, Hilbert space, and separable Banach spaces. The simplest sets of
undecidable in, or independent of, ZFC. real numbers are the basic open sets (i.e., the open intervals with rational
endpoints), and their complements. The sets that are obtained in a
In 1931, Gödel announced his striking incompleteness theorems, which countable number of steps by starting from the basic open sets and
assert that any reasonable formal system for mathematics is necessarily applying the operations of taking the complement and forming a countable
incomplete. In particular, if ZFC is consistent, then there are undecidable union of previously obtained sets are the Borel sets. All Borel sets are
propositions in ZFC. Moreover, Gödel's second incompleteness theorem regular, that is, they enjoy all the classical regularity properties. One
implies that the formal (arithmetical) statement CON(ZFC), which asserts example of a regularity property is the Lebesgue measurability: a set of
that ZFC is consistent, while true, cannot be proved in ZFC. And neither reals is Lebesgue measurable if it differs from a Borel set by a null set,
can its negation. Thus, CON(ZFC) is undecidable in ZFC. namely, a set that can be covered by sets of basic open intervals of
arbitrarily-small total length. Thus, trivially, every Borel set is Lebesgue
If ZFC is consistent, then it cannot prove the existence of a model of ZFC,
measurable, but sets more complicated than the Borel ones may not be.
for otherwise ZFC would prove its own consistency. Thus, a proof of
Other classical regularity properties are the Baire property (a set of reals

14 Stanford Encyclopedia of Philosophy Winter 2014 Edition 15

 Set Theory Joan Bagaria

has the Baire property if it differs from an open set by a meager set, inaccessible cardinal (see Section 10), which implies that one cannot
namely, a set that is a countable union of sets that are not dense in any prove in ZFC that every co-analytic set has the perfect set property.
interval), and the perfect set property (a set of reals has the perfect set
property if it is either countable or contains a perfect set, namely, a closed The theory of projective sets of complexity greater than co-analytic is
set with no isolated points). In ZFC one can prove that there exist non- completely undetermined by ZFC. For example, in L there is a Σ12 set that
regular sets of reals, but the AC is necessary for this (Solovay 1970). is not Lebesgue measurable and does not have the Baire property, whereas
if Martin's axiom holds (see Section 11), every such set has those
The analytic sets, also called Σ11 , are the continuous images of Borel sets. regularity properties. There is, however, an axiom, called the axiom of
And the co-analytic, or Π11 , sets are the complements of analytic sets. Projective Determinacy, or PD, that is consistent with ZFC, modulo the
consistency of some large cardinals, and implies that all projective sets are
Starting from the analytic (or the co-analytic) sets and applying the regular. Moreover, PD settles essentially all questions about the projective
operations of projection (from the product space ℝ ×  to ℝ) and sets. See the entry on large cardinals and determinacy for further details.
complementation, one obtains the projective sets. The projective sets form
a hierarchy of increasing complexity. For example, if A ⊆ ℝ ×  is co- 6.2 Determinacy
analytic, then the projection {x ∈ ℝ : ∃y ∈ ((x, y) ∈ A)} is a projective
set in the next level of complexity above the co-analytic sets. Those sets A regularity property of sets that subsumes all other classical regularity
are called Σ12 , and their complements are called Π12 . properties is that of being determined. For simplicity, we shall work with
the Baire space  . Recall that the elements of  are functions
The projective sets come up very naturally in mathematical practice, for it f : ℕ → ℕ, that is, sequences of natural numbers of length ω. The space
turns out that a set A of reals is projective if and only if it is definable in  is topologically equivalent (i.e., homeomorphic) to the set of irrational
the structure points of ℝ. So, since we are interested in the regularity properties of
subsets of ℝ , and since countable sets, such as the set of rationals, are
 = (ℝ, +, ⋅, ℤ).
negligible in terms of those properties, we may as well work with  ,
That is, there is a first-order formula φ(x, y1 , … , yn ) in the language for instead of ℝ .
the structure such that for some r1 , … , rn ∈ ℝ,
Given A ⊆  , the game associated to A, denoted by A, has two players, I
A = {x ∈ ℝ :  ⊨ φ(x, r1 , … , rn )}. and II, who play alternatively ni ∈ ℕ: I plays n0 , then II plays n1 , then I
plays n2 , and so on. So, at stage 2k, player I plays n2k and at stage 2k + 1,
ZFC proves that every analytic set, and therefore every co-analytic set, is
player II plays n2k+1 . We may visualize a run of the game as follows:
Lebesgue measurable and has the Baire property. It also proves that every
analytic set has the perfect set property. But the perfect set property for co- I n0 n2 n4 ⋯ n2k ⋯
analytic sets implies that the first uncountable cardinal, ℵ1 , is a large II n1 n3 ⋯ ⋯ n2k+1 ⋯
cardinal in the constructible universe L (see Section 7), namely a so-called

16 Stanford Encyclopedia of Philosophy Winter 2014 Edition 17

 Set Theory Joan Bagaria

After infinitely many moves, the two players produce an infinite sequence that every infinite set of real numbers has cardinality either ℵ0 or the same
n0 , n1 , n2 , … of natural numbers. Player I wins the game if the sequence cardinality as ℝ. Thus, the CH is equivalent to 2ℵ0 = ℵ1 .
belongs to A. Otherwise, player II wins.
Cantor proved in 1883 that closed sets of real numbers have the perfect set
The game A is determined if there is a winning strategy for one of the property, from which it follows that every uncountable closed set of real
players. A winning strategy for one of the players, say for player II, is a numbers has the same cardinality as ℝ. Thus, the CH holds for closed sets.
function σ from the set of finite sequences of natural numbers into ℕ , such More than thirty years later, Pavel Aleksandrov extended the result to all
that if the player plays according to this function, i.e., she plays Borel sets, and then Mikhail Suslin to all analytic sets. Thus, all analytic
σ(n0 , … , n2k ) at the k-th turn, she will always win the game, no matter sets satisfy the CH. However, the efforts to prove that co-analytic sets
what the other player does. satisfy the CH would not succeed, as this is not provable in ZFC.

We say that a subset A of  is determined if and only if the game A is In 1938 Gödel proved the consistency of the CH with ZFC. Assuming that
determined. ZF is consistent, he built a model of ZFC, known as the constructible
universe, in which the CH holds. Thus, the proof shows that if ZF is
One can prove in ZFC—and the use of the AC is necessary—that there are consistent, then so is ZF together with the AC and the CH. Hence,
non-determined sets. Thus, the Axiom of Determinacy (AD), which asserts assuming ZF is consistent, the AC cannot be disproved in ZF and the CH
that all subsets of  are determined, is incompatible with the AC. But cannot be disproved in ZFC.
Donald Martin proved, in ZFC, that every Borel set is determined. Further,
he showed that if there exists a large cardinal called measurable (see See the entry on the continuum hypothesis for the current status of the
Section 10), then even the analytic sets are determined. The axiom of problem, including the latest results by Woodin.
Projective Determinacy (PD) asserts that every projective set is
determined. It turns out that PD implies that all projective sets of reals are 7. Gödel's constructible universe
regular, and Woodin has shown that, in a certain sense, PD settles
essentially all questions about the projective sets. Moreover, PD seems to Gödel's constructible universe, denoted by L, is defined by transfinite
be necessary for this. Another axiom, ADL(ℝ) , asserts that the AD holds in recursion on the ordinals, similarly as V , but at successor steps, instead of
L(ℝ) , which is the least transitive class that contains all the ordinals and taking the power set of Vα to obtain Vα+1 , one only takes those subsets of
all the real numbers, and satisfies the ZF axioms (see Section 7). So, Lα that are definable in Lα , using elements of Lα as parameters. Thus,
ADL(ℝ) implies that every set of reals that belongs to L(ℝ) is regular. letting  Def (X) to denote the set of all the subsets of X that are definable
Also, since L(ℝ) contains all projective sets, ADL(ℝ) implies PD. in the structure (X, ∈) by a formula of the language of set theory, using
elements of X as parameters of the definition, we let
6.3 The Continuum Hypothesis
L0 = ∅
The Continuum Hypothesis (CH), formulated by Cantor in 1878, asserts
Def
α+1 = ( α)
18 Stanford Encyclopedia of Philosophy Winter 2014 Edition 19
 Set Theory Joan Bagaria

Lα+1 =  Def (Lα ) the analysis of bigger L-like models, such as L(ℝ) or the inner models for
large cardinals (see Section 10.1).
Lλ = ⋃α<λ Lα , whenever λ is a limit ordinal.

Then L is the union of all the Lα , for α an ordinal, i.e., L = ⋃α∈ON Lα .

8. Forcing

Gödel showed that L satisfies all the ZFC axioms, and also the CH. In In 1963, twenty-five years after Gödel's proof of the consistency of the
fact, it satisfies the Generalized Continuum Hypothesis (GCH), namely CH, relative to the consistency of ZFC, and the consistency of the AC
2ℵα = ℵα+1 , for every ordinal α . from the consistency of ZF, Paul Cohen (1966) proved the consistency of
the negation of the CH, relative to the consistency of ZFC. Thus, if ZFC is
The statement V = L, called the axiom of constructibility, asserts that consistent, then the CH is undecidable in ZFC. He also proved the
every set belongs to L . It holds in L, hence it is consistent with ZFC, and consistency of the negation of the AC, relative to the consistency of ZF.
implies both the AC and the GCH. To achieve this, Cohen devised a new and extremely powerful technique,
called forcing, for expanding transitive models of ZFC.
The proper class L, together with the ∈ relation restricted to L, is an inner
model of ZFC, that is, a transitive (i.e., it contains all elements of its Since the axiom V = L implies the AC and the CH, any model of the
elements) class that contains all ordinals and satisfies all the ZFC axioms. negation of the AC or the CH must violate V = L. So, let's illustrate the
It is in fact the smallest inner model of ZFC, as any other inner model idea of forcing in the case of building a model for the negation of V = L.
contains it. We start with a transitive model M of ZFC, which we may assume,
without loss of generality, to be a model of V = L. To violate V = L we
More generally, given any set A, one can build the smallest transitive need to expand M by adding a new set r so that, in the expanded model, r
model of ZF that contains A and all the ordinals in a similar manner as L, will be non-constructible. Since all hereditarily-finite sets are
but now starting with the transitive closure of {A}, i.e., the smallest constructible, we aim to add an infinite set of natural numbers. The first
transitive set that contains A, instead of ∅. The resulting model, L(A), problem we face is that M may contain already all subsets of ω.
need not be however a model of the AC. One very important such model Fortunately, by the Löwenheim-Skolem theorem for first-order logic, M
is L(ℝ) , the smallest transitive model of ZF that contains all the ordinals has a countable elementary submodel N . So, since we are only interested
and all the real numbers. in the statements that hold in M, and not in M itself, we may as well work
with N instead of M, and so we may assume that M itself is countable.
Then, since (ω) is uncountable, there are plenty of subsets of ω that do
The theory of constructible sets owes much to the work of Ronald Jensen.
He developed the so-called fine structure theory of L and isolated some
not belong to M. But, unfortunately, we cannot just pick any infinite
combinatorial principles, such as the diamond (♢) and square (◻), which
subset r of ω that does not belong to M and add it to M. The reason is that
can be used to carry out complicated constructions of uncountable
r may encode a lot of information, so that when added to M, M is no
mathematical objects. Fine structure theory plays also an important role in
longer a model of ZFC, or it is still a model of V = L. To avoid this, one

20 Stanford Encyclopedia of Philosophy Winter 2014 Edition 21

 Set Theory Joan Bagaria

needs to pick r with great care. The idea is to pick r generic over M, satisfy the sentence that says that there are ℵ2 real numbers.
meaning that r is built from its finite approximations in such a way that it
does not have any property that is definable in M and can be avoided. For 8.1 Other applications of forcing
example, by viewing r as an infinite sequence of natural numbers in the
Besides the CH, many other mathematical conjectures and problems about
increasing order, the property of r containing only finitely-many even
the continuum, and other infinite mathematical objects, have been shown
numbers can be avoided, because given any finite approximation to r—
undecidable in ZFC using the forcing technique.
i.e., any finite increasing sequence of natural numbers—one can always
extend it by adding more even numbers, so that at the end of the One important example is Suslin's Hypothesis (SH). Cantor had shown
construction r will contain infinitely-many even numbers; while the that every linearly ordered set S without endpoints that is dense (i.e.,
property of containing the number 7 cannot be avoided, because when a between any two different elements of S there is another one), complete
finite approximation to r contains the number 7, then it stays there no (i.e., every subset of S that is bounded above has a supremum), and with a
matter how the construction of r proceeds. Since M is countable, there are countable dense subset is isomorphic to the real line. Suslin conjectured
such generic r. Then the expanded model M[r], which includes M and that this is still true if one relaxes the requirement of containing a
contains the new set r, is called a generic extension of M. Since we countable dense subset to being ccc, i.e., every collection of pairwise-
assumed M is a transitive model of V = L, the model M[r] is just Lα (r), disjoint intervals is countable. In the early 1970s, Thomas Jech produced a
where α is the supremum of the ordinals of M. Then one can show, using consistent counterexample using forcing, and Ronald Jensen showed that a
the forcing relation between finite approximations to r and formulas in the counterexample exists in L. About the same time, Robert Solovay and
language of set theory expanded with so-called names for sets in the Stanley Tennenbaum (1971) developed and used for the first time the
generic extension, that M[r] is a model of ZFC and r is not constructible in iterated forcing technique to produce a model where the SH holds, thus
M[r], hence the axiom of constructibility V = L fails. showing its independence from ZFC. In order to make sure that the SH
holds in the generic extension, one needs to destroy all counterexamples,
In general, a forcing extension of a model M is obtained by adding to M a
but by destroying one particular counterexample one may inadvertently
generic subset G of some partially ordered set ℙ that belongs to M. In the
create new ones, and so one needs to force again and again; in fact one
above example, ℙ would be the set of all finite increasing sequences of
needs to go on for at least ω2 -many steps. This is why a forcing iteration is
natural numbers, seen as finite approximations to the infinite sequence r,
needed.
ordered by ⊆; and G would be the set of all finite initial segments of r.
Among other famous mathematical problems that have been shown
In the case of the consistency proof of the negation of the CH, one starts
undecidable in ZFC thanks to the forcing technique, especially using
from a model M and adds ℵ2 new subsets of ω, so that in the generic
iterated forcing and sometimes combined with large cardinals, we may
extension the CH fails. In this case one needs to use an appropriate partial
mention the Measure Problem and the Borel Conjecture in measure
ordering ℙ so that the ℵ2 of M is not collapsed, i.e., it is the same as the
theory, Kaplansky's Conjecture on Banach algebras, and Whitehead's
ℵ2 of the generic extension, and thus the generic extension M[G] will

22 Stanford Encyclopedia of Philosophy Winter 2014 Edition 23

 Set Theory Joan Bagaria

Problem in group theory. One cannot prove in ZFC that there exists a regular limit cardinal κ, for if
κ is such a cardinal, then Lκ is a model of ZFC, and so ZFC would prove
9. The search for new axioms its own consistency, contradicting Gödel's second incompleteness
theorem. Thus, the existence of a regular limit cardinal must be postulated
As a result of 50 years of development of the forcing technique, and its as a new axiom. Such a cardinal is called weakly inaccessible. If, in
applications to many open problems in mathematics, there are now addition κ is a strong limit, i.e., 2λ < κ, for every cardinal λ < κ, then κ is
literally thousands of questions, in practically all areas of mathematics, called strongly inaccessible. A cardinal κ is strongly inaccessible if and
that have been shown independent of ZFC. These include almost all only if it is regular and Vκ is a model of ZFC. If the GCH holds, then
questions about the structure of uncountable sets. One might say that the every weakly inaccessible cardinal is strongly inaccessible.
undecidability phenomenon is pervasive, to the point that the investigation
of the uncountable has been rendered nearly impossible in ZFC alone (see Large cardinals are uncountable cardinals satisfying some properties that
however Shelah (1994) for remarkable exceptions). make them very large, and whose existence cannot be proved in ZFC. The
first weakly inaccessible cardinal is just the smallest of all large cardinals.
This prompts the question about the truth-value of the statements that are Beyond inaccessible cardinals there is a rich and complex variety of large
undecided by ZFC. Should one be content with them being undecidable? cardinals, which form a linear hierarchy in terms of consistency strength,
Does it make sense at all to ask for their truth-value? There are several and in many cases also in terms of outright implication. See the entry on
possible reactions to this. One is the skeptic's position: the statements that independence and large cardinals for more details.
are undecidable in ZFC have no definite answer; and they may even be
inherently vague. Another, the common one among mathematicians, is To formulate the next stronger large-cardinal notion, let us say that a
Gödel's position: the undecidability only shows that the ZFC system is too subset C of an infinite cardinal κ is closed if every limit of elements of C
weak to answer those questions, and therefore one should search for new is also in C; and is unbounded if for every α < κ there exists β ∈ C
axioms that once added to ZFC would answer them. The search for new greater than α. For example, the set of limit ordinals less than κ is closed
axioms has been known as Gödel's Program. See Hauser (2006) for a and unbounded. Also, a subset S of κ is called stationary if it intersects
thorough philosophical discussion of the Program, and also the entry on every closed unbounded subset of κ. If κ is regular and uncountable, then
large cardinals and determinacy for philosophical considerations on the the set of all ordinals less than κ of cofinality ω is an example of a
justification of new axioms for set theory. stationary set. A regular cardinal κ is called Mahlo if the set of strongly
inaccessible cardinals smaller than κ is stationary. Thus, the first Mahlo
A central theme of set theory is thus the search and classification of new cardinal is much larger than the first strongly inaccessible cardinal, as
axioms. These fall currently into two main types: the axioms of large there are κ-many strongly inaccessible cardinals smaller than κ.
cardinals and the forcing axioms.
Much stronger large cardinal notions arise from considering strong
10. Large cardinals reflection properties. Recall that the Reflection Principle (Section 4),

24 Stanford Encyclopedia of Philosophy Winter 2014 Edition 25

 Set Theory Joan Bagaria

which is provable in ZFC, asserts that every true sentence (i.e., every exists an elementary embedding j : V → M, with M transitive, so that κ is
sentence that holds in V ) is true in some Vα. A strengthening of this the first ordinal moved by j, i.e., the first ordinal such that j(κ) ≠ κ. We
principle to second-order sentences yields some large cardinals. For say that κ is the critical point of j, and write crit(j) = κ. The embedding j
example, κ is strongly inaccessible if and only if every Σ11 sentence (i.e., is definable from a κ-complete non-principal measure on κ, using the so-
existential second-order sentence in the language of set theory, with one called ultrapower construction. Conversely, if j : V → M is an elementary
additional predicate symbol) true in the structure (Vκ , ∈, A), for any embedding, with M transitive and κ = crit(j), then the set
A ⊆ Vκ , is true in some (Vα , ∈, A ∩ Vα), α < κ. The same type of U = {X ⊆ κ : κ ∈ j(X)} is a κ-complete non-principal ultrafilter on κ. A
reflection, but now for Π11 sentences (i.e., universal second-order measure U obtained in this way from j is called normal.
sentences), yields a much stronger large cardinal property of κ, called
weak compactness. Every weakly compact cardinal κ is Mahlo, and the set Every measurable cardinal κ is weakly compact, and there are many
of Mahlo cardinals smaller than κ is stationary. By allowing reflection for weakly compact cardinals smaller than κ. In fact, below κ there are many
more complex second-order, or even higher-order, sentences one obtains cardinals that are totally indescribable, i.e., they reflect all sentences, of
large cardinal notions stronger than weak compactness. any complexity, and in any higher-order language.

The most famous large cardinals, called measurable, were discovered by If κ is measurable and j : V → M is given by the ultrapower construction,
Stanisław Ulam in 1930 as a result of his solution to the Measure Problem. then Vκ ⊆ M, and every sequence of length less than or equal to κ of
A (two-valued) measure, or ultrafilter, on a cardinal κ is a subset U of elements of M belongs to M. Thus, M is quite similar to V , but it cannot be
(κ) that has the following properties: (i) the intersection of two elements V itself. Indeed, a famous theorem of Kenneth Kunen shows that there
of U is in U; (ii) if X ∈ U and Y is a subset of κ such that X ⊆ Y , then cannot be any elementary embedding j : V → V , other than the trivial one,
Y ∈ U; and (iii) for every X ⊆ κ, either X ∈ U or κ − X ∈ U, but not both. i.e., the identity. All known proofs of this result use the Axiom of Choice,
A measure U is called κ-complete if every intersection of less than κ and it is an outstanding important question if the axiom is necessary.
elements of U is also in U. And a measure is called non-principal if there Kunen's Theorem opens the door to formulating large cardinal notions
is no α < κ that belongs to all elements of U. A cardinal κ is called stronger than measurability by requiring that M is closer to V .
measurable if there exists a measure on κ that is κ-complete and non-
For example, κ is called strong if for every ordinal α there exists an
principal.
elementary embedding j : V → M, for some M transitive, such that
Measurable cardinals can be characterized by elementary embeddings of κ = crit(j) and Vα ⊆ M.
the universe V into some transitive class M. That such an embedding
Another important, and much stronger large cardinal notion is
j : V → M is elementary means that j preserves truth, i.e., for every
supercompactness. A cardinal κ is supercompact if for every α there exists
formula φ(x1 , … , xn ) of the language of set theory, and every a1 , … , an,
an elementary embedding j : V → M, with M transitive and critical point
the sentence φ(a1 , … , an ) holds in V if and only if φ(j(a1 ), … , j(an ))
κ, so that j(κ) > α and every sequence of elements of M of length α
holds in M. It turns out that a cardinal κ is measurable if and only if there

26 Stanford Encyclopedia of Philosophy Winter 2014 Edition 27

 Set Theory Joan Bagaria

belongs to M. ZFC plus φ is equiconsistent with ZFC plus the existence of some
large cardinal.
Woodin cardinals fall between strong and supercompact. Every
supercompact cardinal is Woodin, and if δ is Woodin, then Vδ is a model Thus, large cardinals can be used to prove that a given sentence φ does not
of ZFC in which there is a proper class of strong cardinals. Thus, while a imply another sentence ψ , modulo ZFC, by showing that ZFC plus ψ
Woodin cardinal δ need not be itself very strong—the first one is not even implies the consistency of some large cardinal, whereas ZFC plus φ is
weakly compact—it implies the existence of many large cardinals in Vδ . consistent assuming the existence of a smaller large cardinal, or just
assuming the consistency of ZFC. In other words, ψ has higher
Beyond supercompact cardinals we find the extendible cardinals, the huge, consistency strength than φ, modulo ZFC. Then, by Gödel's second
the super huge, etc. incompleteness theorem, ZFC plus φ cannot prove ψ, assuming ZFC plus
φ is consistent.
Kunen's theorem about the non-existence of a non-trivial elementary
embedding j : V → V actually shows that there cannot be an elementary As we already pointed out, one cannot prove in ZFC that large cardinals
embedding j : Vλ+2 → Vλ+2 different from the identity, for any λ . exist. But everything indicates that their existence not only cannot be
disproved, but in fact the assumption of their existence is a very
The strongest large cardinal notions not known to be inconsistent, modulo
reasonable axiom of set theory. For one thing, there is a lot of evidence for
ZFC, are the following:
their consistency, especially for those large cardinals for which it is
There exists an elementary embedding j : Vλ+1 → Vλ+1 different possible to construct an inner model.
form the identity.
10.1 Inner models of large cardinals
There exists an elementary embedding j : L(Vλ+1 ) → L(Vλ+1 )
different form the identity. An inner model of ZFC is a transitive proper class that contains all the
ordinals and satisfies all ZFC axioms. Thus, L is the smallest inner model,
Large cardinals form a linear hierarchy of increasing consistency strength. while V is the largest. Some large cardinals, such as inaccessible, Mahlo,
In fact they are the stepping stones of the interpretability hierarchy of or weakly-compact, may exist in L. That is, if κ has one of these large
mathematical theories. See the entry on independence and large cardinals cardinal properties, then it also has the property in L. But some large
for more details. Given any sentence φ, exactly one the following three cardinals cannot exist in L. Indeed, Scott (1961) showed that if there exists
possibilities holds about the theory ZFC plus φ: a measurable cardinal κ, then V ≠ L. It is important to notice that κ does
belong to L, since L contains all ordinals, but it is not measurable in L
ZFC plus φ is inconsistent. because a κ-complete non-principal measure on κ cannot exist there.

ZFC plus φ is equiconsistent with ZFC. If κ is a measurable cardinal, then one can construct an L-like model in
which κ is measurable by taking a κ-complete non-principal and normal

28 Stanford Encyclopedia of Philosophy Winter 2014 Edition 29

 Set Theory Joan Bagaria

measure U on κ, and proceeding as in the definition of L , but now using U cardinals implies that the theory of L(ℝ) , even with real numbers as
as an additional predicate. The resulting model, called L[U], is an inner parameters, cannot be changed by forcing, which implies that all sets of
model of ZFC in which κ is measurable, and in fact κ is the only real numbers that belong to L(ℝ) are regular. Further, under a weaker
measurable cardinal. The model is canonical, in the sense that any other large-cardinal hypothesis, namely the existence of infinitely many Woodin
normal measure witnessing the measurability of κ would yield the same cardinals, Martin and Steel (1989) proved that every projective set of real
model, and has many of the properties of L. For instance, it has a numbers is determined, i.e., the axiom of PD holds, hence all projective
projective well ordering of the reals, and it satisfies the GCH. sets are regular. Moreover, Woodin showed that the existence of infinitely
many Woodin cardinals, plus a measurable cardinal above all of them,
Building similar L-like models for stronger large cardinals, such as strong, implies that every set of reals in L(ℝ) is determined, i.e., the axiom
or Woodin, is much harder. Those models are of the form L[E], where E ADL(ℝ) holds, hence all sets of real numbers that belong to L(ℝ) , and
is a sequence of extenders, each extender being a system of measures, that therefore all projective sets, are regular. He also showed that Woodin
encode the relevant elementary embeddings. cardinals provide the optimal large cardinal assumptions by proving that
the following two statements:
The largest L -like inner models for large cardinals that have been obtained
so far can contain Woodin limits of Woodin cardinals (Neeman 2002). 1. There are infinitely many Woodin cardinals.
However, building an L-like model for a supercompact cardinal is still a
challenge. The supercompact barrier seems to be the crucial one, for 2. ADL(ℝ).
Woodin has shown that for a kind of L-like inner model for a
supercompact cardinal, which he calls the Ultimate-L, all stronger large are equiconsistent, i.e., ZFC plus 1 is consistent if and only if ZFC plus 2
cardinals, such as extendible, huge, I1, etc. would also exist in the model. is consistent. See the entry on large cardinals and determinacy for more
The construction of Ultimate-L is still incomplete, and it is not clear yet details and related results.
that it will succeed, for it rests upon some technical hypotheses that need
Another area in which large cardinals play an important role is the
to be confirmed.
exponentiation of singular cardinals. The so-called Singular Cardinal
10.2 Consequences of large cardinals Hypothesis (SCH) completely determines the behavior of the
exponentiation for singular cardinals, modulo the exponentiation for
The existence of large cardinals has dramatic consequences, even for regular cardinals. The SCH follows from the GCH, and so it holds in L. A
simply-definable small sets, like the projective sets of real numbers. For consequence of the SCH is that if 2ℵn < ℵω, for all finite n , then
example, Solovay (1970) proved, assuming that there exists a measurable 2ℵω = ℵω+1 . Thus, if the GCH holds for cardinals smaller than ℵω , then it
cardinal, that all Σ12 sets of reals are Lebesgue measurable and have the also holds at ℵω . The SCH holds above the first supercompact cardinal
Baire property, which cannot be proved in ZFC alone. And Shelah and (Solovay). But Magidor (1977) showed that, remarkably, assuming the
Woodin (1990) showed that the existence of a proper class of Woodin existence of large cardinals it is possible to build a model of ZFC where

30 Stanford Encyclopedia of Philosophy Winter 2014 Edition 31

 Set Theory Joan Bagaria

the GCH first fails at ℵω, hence the SCH fails. Large cardinals stronger iterated forcing with the ccc property. At first sight, MA may not look like
than measurable are actually needed for this. In contrast, however, ZFC an axiom, namely an obvious, or at least reasonable, assertion about sets,
alone suffices to prove that if the SCH holds for all cardinals smaller than but rather like a technical statement about ccc partial orderings. It does
ℵω1 , then it also holds for ℵω1 . Moreover, if the SCH holds for all singular look more natural, however, when expressed in topological terms, for it is
cardinals of countable cofinality, then it holds for all singular cardinals simply a generalization of the well-known Baire Category Theorem,
(Silver). which asserts that in every compact Hausdorff topological space the
intersection of countably-many dense open sets is non-empty. Indeed, MA
11. Forcing axioms is equivalent to:

Forcing axioms are axioms of set theory that assert that certain existential In every compact Hausdorff ccc topological space, the intersection
statements are absolute between the universe V of all sets and its (ideal) of ℵ1 -many dense open sets is non-empty.
forcing extensions, i.e., some existential statements that hold in some
MA has many different equivalent formulations and has been used very
forcing extensions of V are already true in V . The first forcing axiom was
successfully to settle a large number of open problems in other areas of
formulated by Donald Martin in the wake of the Solovay-Tennenbaum
mathematics. For example, it implies Suslin's Hypothesis and that every
proof of the consistency of Suslin's Hypothesis, and is now known as
Σ12 set of reals is Lebesgue measurable and has the Baire property. It also
Martin's Axiom (MA). Before we state it, let us say that a partial ordering
implies the negation of the CH and that 2ℵ0 is a regular cardinal, but it
is a non-empty set P together with a binary relation ≤ on P that is
does not decide what cardinal it is. See Fremlin (1984) for many more
reflexive and transitive. Two elements, p and q, of P are called compatible
consequences of MA and other equivalent formulations. In spite of this,
if there exists r ∈ P such that r ≤ p and r ≤ q. An antichain of P is a
the status of MA as an axiom of set theory is still unclear. Perhaps the
subset of P whose elements are pairwise-incompatible. A partial ordering
most natural formulation of MA, from a foundational point of view, is in
P is called ccc if every antichain of P is countable. A non-empty subset G
terms of reflection. Writing HC for the set of hereditarily-countable sets
of P is called a filter if (i) every two elements of G are compatible, and (ii)
(i.e., countable sets whose elements are countable, the elements of which
if p ∈ G and p ≤ q , then also q ∈ G. Finally, a subset D of P is called
are also countable, and so on), MA is equivalent to:
dense if for every p ∈ P there is q ∈ D such that q ≤ p .
For every ccc partial ordering P, if an existential statement about
MA asserts the following:
HC holds in an (ideal) generic extension of V obtained by forcing
For every ccc partial ordering P and every set {Dα : α < ω1 } of with P, then the statement is true, i.e., it holds in V . In other words,
dense subsets of P, there exists a filter G ⊆ P that is generic for if a set having a property that depends only on sets in HC exists in
the set, i.e., G ∩ Dα ≠ ∅, for all α < ω1 . some (ideal) generic extension of V obtained by forcing with a ccc
partial ordering, then a set with that property already exists in V .
Martin and Solovay (1970) proved that MA is consistent with ZFC, using

32 Stanford Encyclopedia of Philosophy Winter 2014 Edition 33

 Set Theory Joan Bagaria

The notion of ideal generic extension of V can be made precise in terms of saturated ideals and non-regular ultrafilters”, Part I, Annals of
so-called Boolean-valued models, which provide an alternative version of Mathematics, 127: 1–47.
forcing. Fremlin, D.H., 1984, “Consequences of Martin's Axiom”, Cambridge
tracts in Mathematics #84. Cambridge: Cambridge University Press.
Much stronger forcing axioms than MA were introduced in the 1980s, Gödel, K., 1931, “Über formal unentscheidbare Sätze der Principia
such as J. Baumgartner's Proper Forcing Axiom (PFA), and the stronger Mathematica und verwandter Systeme I,” Monatshefte für
Martin's Maximum (MM) of Foreman, Magidor, and Shelah (1988), which Mathematik Physik, 38: 173–198. English translation in Gödel 1986,
is essentially the strongest possible forcing axiom. Both the PFA and MM 144–195.
are consistent relative to the existence of a supercompact cardinal. The –––, 1938, “The consistency of the axiom of choice and of the generalized
PFA asserts the same as MA, but for partial orderings that have a property continuum hypothesis”, Proceedings of the National Academy of
weaker than the ccc, called properness, introduced by Shelah. And MM Sciences, U.S.A. 24: 556–557.
asserts the same for the wider class of partial orderings that, when forcing –––, 1986, Collected Works I. Publications 1929–1936, S. Feferman et al.
with them, do not destroy stationary subsets of ω1 . (eds.), Oxford: Oxford University Press.
Hauser, K., 2006, “Gödel's program revisited, Part I: The turn to
Strong forcing axioms, such as the PFA and MM imply that all projective
phenomenology”, Bulletin of Symbolic Logic, 12(4): 529–590.
sets of reals are determined (PD), and have many other strong
Jech, T., 2003, Set theory, 3d Edition, New York: Springer.
consequences in infinite combinatorics. Notably, they imply that the
Jensen, R.B., 1972, “The fine structure of the constructible hierarchy”,
cardinality of the continuum is ℵ2 .
Annals of Mathematical Logic, 4(3): 229–308.
Kanamori, A., 2003, The Higher Infinite, Second Edition. Springer
Bibliography
Monographs in Mathematics, New York: Springer.
Bagaria, J., 2008, “Set Theory”, in The Princeton Companion to Kechris, A.S., 1995, Classical Descriptive Set Theory, Graduate Texts in
Mathematics, edited by Timothy Gowers; June Barrow-Green and Mathematics, New York: Springer Verlag.
Imre Leader, associate editors. Princeton: Princeton University Press. Kunen, K., 1980, Set Theory, An Introduction to Independence Proofs,
Cohen, P.J., 1966, Set Theory and the Continuum Hypothesis, New York: Amsterdam: North-Holland.
W. A. Benjamin, Inc. Levy, A., 1979, Basic Set Theory, New York: Springer.
Enderton, H.B., 1977, Elements of Set Theory, New York: Academic Magidor, M., 1977, “On the singular cardinals problem, II”, Annals of
Press. Mathematics, 106: 514–547.
Ferreirós, J., 2007, Labyrinth of Thought: A History of Set Theory and its Martin, D.A. and R. Solovay, 1970, “Internal Cohen Extensions”, Annals
Role in Modern Mathematics, Second revised edition, Basel: of Mathematical Logic, 2: 143–178.
Birkhäuser. Martin, D.A. and J.R. Steel, 1989, “A proof of projective determinacy”,
Foreman, M., M. Magidor, and S. Shelah, 1988, “Martin's maximum, Journal of the American Mathematical Society, 2(1): 71–125.

34 Stanford Encyclopedia of Philosophy Winter 2014 Edition 35

 Set Theory Joan Bagaria

Mathias, A.R.D., 2001, “Slim models of Zermelo Set Theory”, Journal of Mengenlehre, I”, Mathematische Annalen 65: 261–281. Reprinted in
Symbolic Logic, 66: 487–496. Zermelo 2010: 189–228, with a facing-page English translation, and
Neeman, I., 2002, “Inner models in the region of a Woodin limit of an Introduction by Ulrich Felgner (2010). English translation also in
Woodin cardinals”, Annals of Pure and Applied Logic, 116: 67–155. van Heijenoort 1967: 201–215.
Scott, D., 1961, “Measurable cardinals and constructible sets”, Bulletin de
l'Académie Polonaise des Sciences. Série des Sciences Academic Tools
Mathématiques, Astronomiques et Physiques, 9: 521–524.
Shelah, S., 1994, “Cardinal Arithmetic”, Oxford Logic Guides, 29, New How to cite this entry.
York: The Clarendon Press, Oxford University Press. Preview the PDF version of this entry at the Friends of the SEP
–––, 1998, Proper and improper forcing, 2nd Edition, New York: Society.
Springer-Verlag. Look up this entry topic at the Indiana Philosophy Ontology
Shelah, S. and W.H. Woodin, 1990, “Large cardinals imply that every Project (InPhO).
reasonably definable set of reals is Lebesgue measurable”, Israel
Enhanced bibliography for this entry at PhilPapers, with links
Journal of Mathematics, 70(3): 381–394.
to its database.
Solovay, R., 1970, “A model of set theory in which every set of reals is
Lebesgue measurable”, Annals of Mathematics, 92: 1–56.
Other Internet Resources
Solovay, R. and S. Tennenbaum, 1971, “Iterated Cohen extensions and
Souslin's problem”, Annals of Mathematics (2), 94: 201–245. Jech, Thomas, “Set Theory”, The Stanford Encyclopedia of
Todorcevic, S., 1989, “Partition Problems in Topology”, Contemporary Philosophy (Fall 2014 Edition), Edward N. Zalta (ed.), URL =
Mathematics, Volume 84. American Mathematical Society. <http://plato.stanford.edu/archives/fall2014/entries/set-theory/>.
Ulam, S., 1930, ‘Zur Masstheorie in der allgemeinen Mengenlehre’, [This was the previous entry on set theory in the Stanford
Fundamenta Mathematicae, 16: 140–150. Encyclopedia of Philosophy — see the version history.]
Woodin, W.H., 1999, The Axiom of Determinacy, Forcing Axioms, and Cantor's Attic, a website built by by Joel David Hamkins and
the Nonstationary Ideal, De Gruyter Series in Logic and Its Victoria Gitman, but community maintained.
Applications 1, Berlin-New York: Walter de Gruyter.
–––, 2001, “The Continuum Hypothesis, Part I”, Notices of the AMS, Related Entries
48(6): 567–576, and “The Continuum Hypothesis, Part II”, Notices of
the AMS 48(7): 681–690. set theory: continuum hypothesis | set theory: early development | set
Zeman, M., 2001, Inner Models and Large Cardinals, De Gruyter Series theory: independence and large cardinals | set theory: large cardinals and
in Logic and Its Applications 5, Berlin-New York: Walter de Gruyter. determinacy
Zermelo, E., 1908, “Untersuchungen über die Grundlagen der

36 Stanford Encyclopedia of Philosophy Winter 2014 Edition 37

 Set Theory Joan Bagaria

Basic Set Theory Commutativity:

A∪B=B∪A
Sets are well-determined collections that are completely characterized by
their elements. Thus, two sets are equal if and only if they have exactly the A∩B=B∩A
same elements. The basic relation in set theory is that of elementhood, or Distributivity:
membership. We write a ∈ A to indicate that the object a is an element, or
a member, of the set A. We also say that a belongs to A. Thus, a set A is A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C)
equal to a set B if and only if for every a, a ∈ A if and only if a ∈ B . In A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C)
particular, there is only one set with no elements at all. This set is called,
naturally, the empty set, and is represented by the symbol ∅. Idempotency:

A∪A=A
We say that A is a subset of B, written A ⊆ B, if every element of A is an
element of B. Thus, A = B if and only if A ⊆ B and B ⊆ A. Notice that A∩A=A
∅ ⊆ A, for every set A. A∪∅=A
Given sets A and B, one can perform some basic operations with them A∩∅=∅
yielding the following sets:
A−A=∅
The set A ∪ B , called the union of A and B, whose elements are the If A ⊆ B, then
elements of A and the elements of B.
A ∪ B = A ∪ (B − A) = B
The set A ∩ B , called the intersection of A and B , whose elements are
A∩B=A
the elements common to A and B.
Given an object a we can form the set that has a as its only element. This
The set A − B, called the difference of A and B, whose elements are set is denoted by {a}. More generally, given a, b, c, …, we can form the
those elements of A that are not members of B. set having a, b, c, … as its elements, which we denote by {a, b, c, …}. Of
course, we can actually write down all the elements of the set when there
It is routine to check that those operations satisfy the following properties:
are not too many of them. In the case of infinite sets this is clearly not
Associativity: possible.

A ∪ (B ∪ C) = (A ∪ B) ∪ C If a = b , then {a, b} = {a}. Also, for any a and b, the pair {a, b} is the
same as the pair {b, a} . So, if we wish to take into account the order in
A ∩ (B ∩ C) = (A ∩ B) ∩ C
which the two elements of a pair are given, we need to find another way of

38 Stanford Encyclopedia of Philosophy Winter 2014 Edition 39

 Set Theory Joan Bagaria

representing the pair. Thus, we define the ordered pair (a, b) as the set called transitive if (a, c) ∈ R whenever (a, b) ∈ R and (b, c) ∈ R. A
{{a}, {a, b}}. One can easily check that two ordered pairs (a, b) and (c, d) relation that is reflexive, symmetric, and transitive is called an equivalence
are equal if and only if a = c and b = d . The order is now important, for if relation. The identity relation on any set A is the paradigmatic example of
a ≠ b, then (a, b) ≠ (b, a) . an equivalence relation. Another example is the relation on the set of all
finite sets of natural numbers consisting of all the pairs (a, b) such that a
The Cartesian product A × B of two sets, A and B, is defined as the set of and b have the same number of elements.
all ordered pairs (a, b) such that a ∈ A and b ∈ B .
If R is an equivalence relation on a set A, and (a, b) ∈ R , then we say that
Having defined ordered pairs, one can now define ordered triples (a, b, c) that a and b are R-equivalent. For every a ∈ A , the equivalence class of a,
as (a, (b, c)), or in general ordered n-tuples (a1 , … , an ) as usually denoted by [a]R , is the set of all elements of A that are R-
(a1 , (a2 , … , an )). equivalent to a. The set of all R-equivalence classes is called the quotient
set and is denoted by A/R. One can easily check that A/R is a partition of
The Cartesian product A1 × … × An, of the sets A1 , … , An is the set of all
A, that is, no element of A/R is empty, any two elements of A/R are
n-tuples (a1 , … , an ) such that ai ∈ Ai , for every 1 ≤ i ≤ n. In particular,
disjoint, and every a ∈ A belongs to (exactly) one element of A/R, namely
for n ≥ 2 , the n-times Cartesian product of a set A, denoted by An , is the
the class [a]R .
set of all n-tuples of elements of A.
If R is a binary relation, then one usually writes aRb instead of (a, b) ∈ R.
1. Relations
2. Functions A binary relation R on a set A is called antisymmetric if a = b whenever
3. Sets and formulas aRb and bRa . A relation R on a set A that is reflexive, antisymmetric, and
4. Ordinals transitive, is called a (reflexive) partial order. If we remove from R all
5. Countable and uncountable sets pairs (a, a), for every a ∈ A, then we get a strict partial order. The ⊆
5.1 Cardinals relation on any set of sets is an example of a partial order. A partial order
Further Readings on a given set A is usually represented by the symbol ≤, and the
corresponding strict partial ordering by <. A partial order ≤ on a set A
1. Relations with the additional property that either a ≤ b or b ≤ a, for all elements a
and b of A, is called a total order, or a linear order. The usual orderings of
A binary relation on a set A is a set of ordered pairs of elements of A, that the set ℕ of natural numbers, the set ℤ of the integers, the set ℚ of the
is, a subset of A × A. In general, an n -ary relation on A is a subset of An . rational numbers, or the set ℝ of real numbers, are linear orders.

A binary relation R on a set A is called reflexive if (a, a) ∈ R for every Notice that if ≤ is a linear order on a set A, and B ⊆ A, then ≤ ∩ B2 is also
a ∈ A. It is called symmetric if (b, a) ∈ R whenever (a, b) ∈ R . And it is a linear order on B. If ≤ is a linear order on a set A, then we say that

40 Stanford Encyclopedia of Philosophy Winter 2014 Edition 41

 Set Theory Joan Bagaria

a ∈ A is the ≤-least element of A if there is no b ∈ A distinct from a such 3. Sets and formulas
that b ≤ a. The number 0 is the least element of ℕ, but ℤ has no least
element. The formal language of set theory is the first-order language whose only
non-logical symbol is the binary relation symbol ∈.
A linear order ≤ on a set A is a well-order if every non-empty subset of A
has a ≤-least element. Equivalently, if there is no infinite strictly Given any formula φ(x, y1 , … , yn ) of the language of set theory, and sets
descending sequence A, B1 , … , Bn, one can form the set of all those elements of A that satisfy
the formula φ(x, B1 , … , Bn ). This set is denoted by
… < a2 < a1 < a0 {a ∈ A : φ(a, B1 , … , Bn )}. The following are some examples
of elements of A. Thus, the usual ordering of ℕ is a well-order. But the
∅ = {a ∈ A : a ≠ a}
usual order on ℤ is not, because it has no least element.
A = {a ∈ A : a = a}
2. Functions
A − B = {a ∈ A : a ∉ B}.
A (1 -ary) function on a set A is a binary relation F on A such that for
every a ∈ A there is exactly one pair (a, b) ∈ F . The element b is called A ∩ B = {a ∈ A : a ∈ B}.
the value of F on a , and is denoted by F(a). And the set A is called the
domain of F. The notation F : A → B indicates that F is a function with And if B and C are subsets of A, then
domain A and values in the set B. For n ≥ 2, an n-ary function on A is a
B ∪ C = {a ∈ A : a ∈ B ∨ a ∈ C}.
function F : An → B, for some B.
Given a subset C ⊆ A × B, the projection of C (on the first coordinate) is
A function F : A → B is one-to-one if for all elements a and b of A, if
the set
a ≠ b, then F(a) ≠ F(b) . And is onto if for every b ∈ B there is some
a ∈ A such that F(a) = b. Finally, F is bijective if it is one-to-one and {a ∈ A : ∃b ∈ B((a, b) ∈ C)} .
onto. Thus, a bijection F : A → B establishes a one-to-one
correspondence between the elements of A and those of B, and A is It is not the case, however, that given any formula φ(x, y1 , … , yn ), and sets
bijectable with B if there is such a bijection. The identity function on a set B1 , … , Bn, one can form the set of all those sets that satisfy the formula
A, denoted by Id : A → A , and which consists of all the pairs (a, a) , with φ(x, B1 , … , Bn ). For let φ(x) be the formula x ∉ x. If A were the set of all
a ∈ A, is trivially a bijection. sets that satisfy the formula, then A ∈ A if and only if A ∉ A. A
contradiction! This contradiction is known as Russell's paradox, after
Given functions F : A → B and G : B → C, the composition of F and G, Bertrand Russell, who discovered it in 1901 (see the entry on Russell's
written G ∘ F, is the function G ∘ F : A → C whose elements are all pairs paradox).
(a, G(F(a))) , where a ∈ A. If F and G are bijections, then so is G ∘ F.

42 Stanford Encyclopedia of Philosophy Winter 2014 Edition 43

 Set Theory Joan Bagaria

4. Ordinals the ordinals produced so far, as in the case of ω which yields a new limit
ordinal. Note, however, that one cannot take the set of all ordinals, for
The first ordinal number is ∅. Given an ordinal α, the next bigger ordinal, then this set would be a new limit ordinal, which is impossible, since we
called the (immediate) successor of α , is the set α ∪ {α} . Thus, the already had them all.
successor of α is just the set α together with one more element, namely, α
itself. The finite ordinal numbers are those obtained by starting with ∅ As with finite ordinals, every infinite ordinal is just the set of its
and repeatedly taking the successor. predecessors. One consequence of this is that the ∈ relation is a strict well-
order on any set of ordinals. Thus, for any ordinals α and β we define
In set theory the natural numbers are defined as the finite ordinals. Thus, α < β if and only if α ∈ β. Then the associated reflexive well-order is
0=∅ defined as α ≤ β if and only if α < β or α = β. Let us now observe that
α ⊆ β if and only if α ≤ β.
1 = ∅ ∪ {∅} = {∅}
2 = {∅} ∪ {{∅}} = {∅, {∅}} 5. Countable and uncountable sets
3 = {∅, {∅}} ∪ {{∅, {∅}}} = {∅, {∅}, {∅, {∅}}} If A is a finite set, there is a bijection F : n → A between a natural number
n and A. Any such bijection gives a counting of the elements of A,
⋮
namely, F(0) is the first element of A, F(1) is the second, and so on. Thus,
Notice that 1 = {0} , 2 = {0, 1} , 3 = {0, 1, 2} , and in general we have all finite sets are countable. An infinite set A is called countable if there is
n = {0, 1, 2, … , n − 1} . So, every natural number n is just the set of its a bijection F : ω → A between the set of natural numbers and A. The set
predecessors. ℕ of natural numbers is (trivially) countable. If A is an infinite subset of
ω, then A is also countable: for let F : ω → A be such that F(n) is the
A set A is finite if there is a one-to-one correspondence between some least element of A that is not in the set {F(m) ∈ A : m < n}. Then F is a
natural number n and the elements of A, i.e., a bijection F : n → A, in bijection.
which case we say that A has n elements. A set is infinite if it is not finite.
Every infinite subset of a countable set is also countable: for suppose
The set of all finite ordinals is denoted by the Greek letter omega (ω). F : ω → A is a bijection and B ⊆ A is infinite. Then the set
Thus, ω is just the set ℕ of natural numbers. ω is also an ordinal, the first {n ∈ ω : F(n) ∈ B} is an infinite subset of ω, hence countable, and so
infinite ordinal. Notice that ω is not the successor of any ordinal, and so it there is a bijection G : ω → {n ∈ ω : F(n) ∈ B}. Then the composition
is called a limit ordinal. Once we have ω we can continue generating more function F ∘ G : ω → B is a bijection.
ordinals by taking its successor ω ∪ {ω}, then its successor
(ω ∪ {ω}) ∪ {ω ∪ {ω}}, and so on. All ordinal numbers greater than 0 The union of a countable set and a finite set is also countable. For given
are produced in this way, namely, either by taking the successor of the last sets A and B, which without loss of generality we may assume they are
produced ordinal, or, if there is no such last ordinal, by taking the set of all disjoint, and given bijections F : ω → A and G : n → B, for some n < ω,

44 Stanford Encyclopedia of Philosophy Winter 2014 Edition 45

 Set Theory Joan Bagaria

let H : ω → A ∪ B be the bijection given by: H(m) = G(m), for every number different from F(n), all n, which is impossible because F is a
m < n, and H(m) = F(m − n), for every n ≤ m. bijection.

Moreover, the union of two countable sets is also countable: since we have The existence of uncountable sets follows from a much more general fact,
already shown that the union of a countable set and a finite set is also also discovered by Cantor. Namely, given any set A, the set of all its
countable, it is enough to see that the union of two disjoint countable sets subsets, called the power set of A, and denoted by (A), is not bijectable
is also countable. So, suppose A and B are countable sets and F : ω → A with A: for suppose that F : A → (A) is a bijection. Then the subset
and G : ω → B are bijections, then the function H : ω → A ∪ B {a ∈ A : a ∉ F(a)} of A is the value F(a) of some a ∈ A. But then
consisting of all pairs (2n, F(n)) , plus all pairs (2n + 1, G(n)) is a a ∈ F(a) if and only if a ∉ F(a). Hence, if A is any infinite set, then (A)
bijection. is uncountable.

Thus, the set ℤ, being the union of two countable sets, namely There are also uncountable ordinals. The set of all finite and countable
ordinals is also an ordinal, called ω1 , and is the first uncountable ordinal.
ℕ ∪ {−1, −2, −3, −4, …}
Similarly, the set of all ordinals that are bijectable with some ordinal less
is also countable. than or equal to ω1 is also an ordinal, called ω2 , and is not bijectable with
ω1 , and so on.
The Cartesian product of two infinite countable sets is also countable. For
suppose F : ω → A and G : ω → B are bijections. Then, using the fact 5.1 Cardinals
that the function J : ω × ω → ω given by J((m, n)) = 2m (2n + 1) − 1 is a
bijection, one has that the function H : ω → A × B given by The cardinality, or size, of a finite set A is the unique natural number n
H(2m (2n + 1) − 1) = (F(m), G(n)) is also a bijection. such that there is a bijection F : n → A.

Since any rational number is given by a pair of integers, i.e., a quotient mn , In the case of infinite sets, their cardinality is given, not by a natural
where m, n ∈ ℤ and n ≠ 0, the set ℚ of rational numbers is also number, but by an infinite ordinal. However, in contrast with the finite
countable. sets, an infinite set A is bijectable with many different ordinal numbers.
For example, the set ℕ is bijectable with ω, but also with it successor
However, Georg Cantor discovered that the set ℝ of real numbers is not ω ∪ {ω}: by assigning 0 to ω and n + 1 to n, for all n ∈ ω, we obtain a
countable. For suppose, aiming for a contradiction, that F : ω → ℝ is a bijection between ω ∪ {ω} and ω. But since the ordinals are well-ordered,
bijection. Let a0 = F(0). Choose the least k such that a0 < F(k) and put we may define the cardinality of an infinite set as the least ordinal that is
b0 = F(k). Given an and bn , choose the least l such that an < F(l) < bn , bijectable with it.
and put an+1 = F(l). And choose the least m such that an+1 < F(m) < bn ,
and put bn+1 = F(m). Thus, we have a0 < a1 < a2 < ⋯ In particular, the cardinality of an ordinal number α is the least ordinal κ
⋯ < b2 < b1 < b0. Now let a be the limit of the an . Then a is a real that is bijectable with it. Notice that κ is not bijectable with any smaller

46 Stanford Encyclopedia of Philosophy Winter 2014 Edition 47

 Set Theory Joan Bagaria

ordinal, for otherwise so would be α . The ordinal numbers that are not ∀x∀y[∀z(z ∈ x ↔ z ∈ y) → x = y]
bijectable with any smaller ordinal are called cardinal numbers. Thus, all
This axiom asserts that when sets x and y have the same members, they are
natural numbers are cardinals, and so are ω, ω1 , ω2 , and so on. In general,
the same set.
given any cardinal κ, the set of all ordinals that are bijectable with some
ordinal ≤ κ is also a cardinal; it is the smallest cardinal bigger than κ. The next axiom asserts the existence of the empty set:

The infinite cardinals are represented by the letter aleph (ℵ) of the Hebrew Null Set:
alphabet. Thus, the smallest infinite cardinal is ω = ℵ0 , the next one is ∃x¬∃y(y ∈ x)
ω1 = ℵ1 , which is the first uncountable cardinal, then comes ω2 = ℵ2 ,
Since it is provable from this axiom and the previous axiom that there is a
etc.
unique such set, we may introduce the notation ‘∅’ to denote it.
The cardinality of any set A, denoted by |A|, is the unique cardinal number
The next axiom asserts that if given any set x and y, there exists a pair set
that is bijectable with A. We saw already that |ℝ| is uncountable, hence
of x and y, i.e., a set which has only x and y as members:
greater than ℵ0 , but it is not known what cardinal number it is. The
conjecture that |ℝ| = ℵ1 , formulated by Cantor in 1878, is the famous Pairs:
Continuum Hypothesis. ∀x∀y∃z∀w(w ∈ z ↔ w = x ∨ w = y)

Further Readings Since it is provable that there is a unique pair set for each given x and y,
we introduce the notation ‘{x,y}’ to denote it.
Devlin, K., 1979, Fundamentals of Contemporary Set Theory,
Undergraduate Texts in Mathematics, Springer, Second edition 1993, The next axiom asserts that for any set x, there is a set y which contains as
The Joy of Sets: Fundamentals of Contemporary Set Theory. members all those sets whose members are also elements of x, i.e., y
Undergraduate Texts in Mathematics, New York: Springer. contains all of the subsets of x:
Enderton, H.B., 1977, Elements of Set Theory, New York: Academic
Power Set:
Press.
∀x∃y∀z[z ∈ y ↔ ∀w(w ∈ z → w ∈ x)]
Jech, T. and K. Hrbaček, 1978 [1999], Introduction to set theory, New
York: Marcel Dekker, 3rd edition 1999. Since every set provably has a unique ‘power set’, we introduce the
notation ‘𝒫(x)’ to denote it. Note also that we may define the notion x is a
Zermelo-Fraenkel Set Theory subset of y (‘x ⊆ y’) as: ∀z(z ∈ x → z ∈ y). Then we may simplify the
statement of the Power Set Axiom as follows:
Axioms of ZF
∀x∃y∀z(z ∈ y ↔ z ⊆ x)
Extensionality:
∀x∀y[∀z(z ∈ x ↔ z ∈ y) → x = y]

48 Stanford Encyclopedia of Philosophy Winter 2014 Edition 49

 Set Theory Joan Bagaria

The next axiom asserts that for any given set x, there is a set y which has in ψ(x, û ). Then the Separation Schema asserts:
as members all of the members of all of the members of x:
Separation Schema:
Unions: ∀u1 … ∀uk [∀w∃v∀r(r ∈ v ↔ r ∈ w ∧ ψx,û [r, û])]
∀x∃y∀z[z ∈ y ↔ ∃w(w ∈ x ∧ z ∈ w)]
In other words, if given a formula ψ and a set w, there exists a set v which
Since it is provable that there is a unique ‘union’ of any set x, we introduce has as members precisely the members of w which satisfy the formula ψ .
the notation ‘⋃ x’ to denote it.
The next axiom of ZF is the Replacement Schema. Suppose that ϕ(x, y, û )
The next axiom asserts the existence of an infinite set, i.e., a set with an is a formula with x and y free, and let û represent the variables u1 , … uk ,
infinite number of members: which may or may not be free in ϕ. Furthermore, let ϕx,y,û [s, r, û ] be the
result of substituting s and r for x and y, respectively, in ϕ(x, y, û ). Then
Infinity:
every instance of the following schema is an axiom:
∃x[∅ ∈ x ∧ ∀y(y ∈ x → ⋃{y, {y}} ∈ x)]
Replacement Schema:
We may think of this as follows. Let us define the union of x and y (‘
∀u1 … ∀uk [∀x∃!yϕ(x, y, û ) → ∀w∃v∀r(r ∈ v ↔ ∃s(s ∈ w ∧ ϕx,y,û [s, r
x ∪ y’) as the union of the pair set of x and y, i.e., as ⋃{x, y}. Then the
Axiom of Infinity asserts that there is a set x which contains ∅ as a In other words, if we know that ϕ is a functional formula (which relates
member and which is such that whenever a set y is a member of x, then each set x to a unique set y), then if we are given a set w, we can form a
y ∪ {y} is a member of x. Consequently, this axiom guarantees the new set v as follows: collect all of the sets to which the members of w are
existence of a set of the following form: uniquely related by ϕ.
{∅, {∅}, {∅, {∅}}, {∅, {∅}, {∅, {∅}}}, …}
Note that the Replacement Schema can take you ‘out of’ the set w when
Notice that the second element, {∅}, is in this set because (1) the fact that forming the set v. The elements of v need not be elements of w. By
∅ is in the set implies that ∅ ∪ {∅} is in the set and (2) ∅ ∪ {∅} just is contrast, the Separation Schema of Zermelo only yields subsets of the
{∅}. Similarly, the third element, {∅, {∅}}, is in this set because (1) the given set w.
fact that {∅} is in the set implies that {∅} ∪ {{∅}} is in the set and (2)
The final axiom asserts that every set is ‘well-founded’:
{∅} ∪ {{∅}} just is {∅, {∅}}. And so forth.
Regularity:
The next axiom is the Separation Schema, which asserts the existence of a
∀x[x ≠ ∅ → ∃y(y ∈ x ∧ ∀z(z ∈ x → ¬(z ∈ y)))]
set that contains the elements of a given set w that satisfy a certain
condition ψ. That is, suppose that ψ(x, û ) has x free and may or may not A member y of a set x with this property is called a ‘minimal’ element.
have u1 , … , uk free. And let ψx,û [r, û ] be the result of substituting r for x This axiom rules out the existence of circular chains of sets (e.g., such as

50 Stanford Encyclopedia of Philosophy Winter 2014 Edition 51

 Set Theory

x ∈ y ∧ y ∈ z ∧ and z ∈ x) as well as infinitely descending chains of sets

(such as … x3 ∈ x2 ∈ x1 ∈ x0 ).

Copyright © 2014 by the author

Joan Bagaria

52 Stanford Encyclopedia of Philosophy