1 The Axioms of Set Theory
1 The Axioms of Set Theory
In the late nineteenth century, Georg Cantor created mathematical theories, first of sets or aggregates of real
numbers (or linear points), and later of sets or aggregates of arbitrary elements. The relationship of element a to
set A is written a 2 A; it is to be distinguished from the relationship of subset B to set A, which holds if every
element of B is also an element of A, and which is written B A. Cantor is most famous for his theory of
transfinite cardinals, or numbers of elements in infinite sets. A subset of an infinite set may have the same number
of elements as the set itself, and Cantor proved that the sets of natural and rational numbers have the same
number of elements, which he called @0 ; also that the sets of real and complex numbers have the same number of
elements, which he called c. Cantor proved @0 to be less than c. He conjectured that no set has a number of
elements strictly between these two.
In the early twentieth century, in response to criticism of set theory, Ernst Zermelo undertook its axiomatization;
and, with amendments by Abraham Fraenkel, his have been the accepted axioms ever since. These axioms help
distinguish the notion of a set, which is too basic to admit of informative definition, from other notions of a one
made up of many that have been considered in logic and philosophy. Properties having exactly the same
particulars as instances need not be identical, whereas sets having exactly the same elements are identical by the
axiom of extensionality. Hence for any condition there is at most one set fxj(x)g whose elements are all and
only those x such that (x) holds, and fxj(x)g = fxj(x)g if and only if conditions and hold of exactly
the same x. It cannot consistently be assumed that fxj(x)g exists for every condition . Inversely, the existence
of a set is not assumed to depend on the possibility of defining it by some condition as fxj(x)g .
One set x0 may be an element of another set x1 which is an element of x2 and so on, x0 2 x1 2 x2 2 : : :, but the
reverse situation, : : : 2 y2 2 y1 2 y0 , may not occur, by the axiom of foundation. It follows that no set is an
element of itself and that there can be no universal set y = fxjx = xg. Whereas a part of a part of a whole is a
part of that whole, an element of an element of a set need not be an element of that set.
Modern mathematics has been greatly influenced by set theory, and philosophies rejecting the latter must
therefore reject much of the former. Many set-theoretic notations and terminologies are encountered even outside
mathematics, as in parts of philosophy:
pair fa; bg fxjx = a or x = bg
singleton fag fxjx = ag
empty set ; fxjx 6= xg
union [X faja 2 A for some A 2 Xg
binary union A [ B faja 2 A or a 2 Bg
intersection \X faja 2 A for all A 2 Xg
binary intersection A \ B faja 2 A and a 2 Bg
dif ference A B faja 2 A and not a 2 Bg
complement AB
power set }(A) fBjB Ag
(In contexts where only subsets of A are being considered, A B may be written B and called the complement
of B.)
While the accepted axioms suffice as a basis for the development not only of set theory itself, but of modern
mathematics generally, they leave some questions about transfinite cardinals unanswered. The status of such
questions remains a topic of logical research and philosophical controversy.
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
The accepted axioms ZFC consist of ZF plus the axiom of choice (AC): given a partition or set x of sets that are
non-empty (each has at least one element) and disjoint (no two have a common element), there exists a choice set y
having exactly one element in common with each element of x.
The intuition behind the axioms is that sets lie in a hierarchy of levels (see Set theory, different systems of 5), and
when x 2 y, then x lies below and y, above. At the lowest level there is just ; (though a variant of the standard notion
allows at this level or items that have no elements and are not sets but may be elements of sets of any
level). There is no highest level (though a variant allows a top level of or items that may have sets of any
level as elements but are not themselves sets nor elements of anything). If set x lies at one level then the of
x, x# = x [ fxg lies at the next higher level. The levels continue into the infinite, so that above ;, ;# = f;g,
;## = f;; f;gg; : : : lies the set f;; f;g; f;; f;gg; : : :g , whose existence is asserted by the axiom of infinity. The
hierarchy is supposed to be as and as as possible, an intuition (only) partly expressed by the remaining
existence axioms. These consist of axioms asserting the existence of the pair, union and power sets, along with
infinitely many further axioms of each of two kinds. First, for any condition (x) there is an axiom asserting that, for
any set u, separation from u of those elements x for which (x) holds forms a set:
fx 2 uj(x)g = fxjx 2 u and (x)g
Second, for any binary condition there is an axiom asserting that if for every x there is a unique y = (x) such
that (x; y) holds, then for any set u, replacement of each element x of u by (x) forms a set:
f(x)jx 2 ug = fyj(x; y) for some x 2 ug
(In a rigorous formalization, the vague, intuitive notion of would be explained in terms of a precise, logical
notion of a of a symbolic language.)
To develop mathematics on the basis of these axioms, for each system of numbers, points or other mathematical
objects, a system of set-theoretic objects must be specified to serve as substitutes. Usually the system of set-theoretic
substitutes is easy to specify intuitively, in terms of the relationship of the substitutes to the original objects, but it
may be less easy to define formally, in terms taken only from the language of set theory. Set-theoretic counterparts to
the basic operations on the original objects must also be defined in the language of set theory, and the counterparts of
the basic laws for these operations deduced from the axioms of set theory. When this is done, the set-theoretic
substitutes are thereafter called by the name of the mathematical objects in question, which are said to have been
their set-theoretic substitutes.
No more can be done here than to describe intuitively for the natural, integral, rational, real and complex numbers the
related sets with which these numbers are identified. First, each natural number n 2 N is identified with the set of
natural numbers less than n, so that 0 = ;, 1 = f0g = f;g , 2 = f0; 1g = f;; f;gg , and so on; and N N itself with the
set whose existence is asserted by the axiom of infinity (above). For other kinds of numbers, one needs the
identification of the ordered pair (a; b) of a; b with the set ffag; fa; bgg , which allows one to define
A B = f(0; a)ja 2 Ag [ f(1; b)jb 2 Bg
A B = f(a; b)ja 2 A and b 2 Bg
B A = fF A Bj for all a 2 A there is
a unique b 2 B with (a; b) 2 F g
dom(F ) = faj(a; b) 2 F for some bg
ran(F ) = fbj(a; b) 2 F for some ag
F dC = f(a; b) 2 F ja 2 Cg
F [C] = fbj(a; b) 2 F for some a 2 Cg
1
F [C] = faj(a; b) 2 F for some a 2 Cg
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
-.)
Now an integer w can be identified with the set of pairs of naturals whose difference is w; a rational x with the set of
pairs of integers whose quotient is x; a real y with the set of rationals x < y or, alternatively, with the set of functions
f from naturals to rationals such that f (x) for larger and larger x gives a better and better approximation to y; and a
complex z with the ordered pair of its real and imaginary parts (each a real number). Ultimately any hypothesis that
can be stated in conventional mathematical language can be stated in the language of set theory, and any theorem
provable by conventional mathematical methods can be proved from the axioms ZF, with AC, or one of its many
known implications, such as (DC), being needed in certain branches of higher mathematics. (DC
asserts that if R is a relation on A such that for every a 2 A there exists at least one b 2 A with aRb, then there exists
a function F from N N to A with F (n)RF (n + 1) for all n 2 N.)
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
the ordinal of the natural order on the natural numbers, is called !0 or simply !. It follows from the definitions that if
is the ordinal of a well-order S on a set B, then there is an isomorphism F from S to the order on the set fj < g
of ordinals less than (namely, F (b) = jSb j for b 2 B). It is now customary to identify the ordinal with this set
of lesser ordinals (generalizing what was done in the case of finite ordinals - natural numbers - in 1). With this
identification, it is easy to describe explicitly # and supT : they are just [ fg and [T , respectively. Thus
!# = f0; 1; 2; : : : ; !g
!## = f0; 1; 2; : : : ; !; ! # g
and so on; and the next limit is
! y = f0; 1; 2; : : : ; !; ! # ; !## ; : : :g:
Addition and multiplication for order types can be defined as follows. Given linear orders R, S on sets A, B, one
defines orders R S, R S on A B, A B by
(a; b)(R S)(c; d) if (a = c = 0 and bRd)
or (a = 0 and c = 1)
or (a = c = 1 and bSd)
(a; b)(R S)(c; d) if bSd
and (if b = d then aRc):
Intuitively, R S is like a copy of A in the R order, followed by a copy of B in the S order, while R S is like a
copy of B but with each element replaced by an entire copy of A. Then jRj + jSj = jR Sj and jRj jSj = jR Sj .
For ordinals the methods of proof by transfinite induction and definition by transfinite recursion, analogous to the
ordinary methods of induction and recursion in higher arithmetic or number theory, are available. To prove by
transfinite induction that () holds for all ordinals , one must prove, as in finite induction, that (0) holds and if
() holds then (0 ) holds, and additionally that if is a limit and () holds for all < , then ( ) holds.
Similarly, to define by recursion () for all ordinals , one must define (1) (0), (2) (0 ) in terms of (), and
(3) ( ) in terms of () for < at limits . Equivalent, more convenient, definitions of addition and
multiplication, plus a definition of ordinal exponentiation, can be given for ordinals by transfinite recursion:
+0 = + (# ) = ( + )#
0 = 0 (# ) = ( ) +
0 = 1 (
#
)
= ( )
and at the limits:
+ = supf + j < g
= supf j < g
= supf j < g:
If ; both have cardinality , then so do + , and even (in contrast to cardinal exponentiation) . Note
# = + 1. Also ! # , ! ## and ! y are ! + 1, ! + 2, ! + !.
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
is neither zero nor a successor.
As with ordinals, the cardinals 0, 1 = 0# , 2 = 1# ; : : : less than the least limit are called finite, and the least limit
supf0; 1; 2; : : :g , the cardinal of the natural numbers, is called @0 (- or -). Sets of cardinality
@0 are called . @+ + ++
0 is called @1 , and the least ordinal of cardinality @1 is called !1 or . @1 , @1 ; : : :
are called @2 , @3 ; : : :, and the next limit cardinal supf@0 ; @1 ; @2 ; : : :g is called @! , while the least ordinals of these
cardinals are called !2 , !3 ; : : : ; !! . There is only one ordinal of each finite cardinality (since there is an
isomorphism between any two orders on a finite set). Not so for infinite cardinals, since the ordinals ! # ,
! ## ; : : : ; ! y can also be described as the ordinals of certain unnatural well-orders on the natural numbers, namely
1; 2; 3; : : : ; 0
2; 3; 4; : : : ; 0; 1
1; 3; 5; : : : ; 2; 4; 6; : : : ; : : :
and hence all have cardinality @0 . Indeed, Cantor gave an explicit description of + by proving that it is the cardinal
of the set of ordinals of cardinality . It is now customary to identify a cardinal with the least ordinal of that cardinal,
so that @0 = !0 , and so on.
Cardinal addition, multiplication and exponentiation are defined as follows: jjAjj + jjBjj = jjA Bjj ,
jjAjj jjBjj = jjA Bjj and jjAjj . These agree with the usual arithmetic notions on natural numbers, and some of the
usual arithmetic laws generalize from the finite to the infinite (the associative, commutative and distributive laws of
addition and multiplication, and the laws of exponents). But there are some differences. Bijections from N N to NNN N
and to N NN N are suggested by the following.
(0; 0) (0; 0)
(1; 0) (0; 1)
(0; 1) (1; 1)
(1; 1) (1; 0)
(0; 2) (0; 2)
(1; 2) (1; 2)
(0; 3) (2; 2)
(1; 3) (2; 1)
(0; 4) (2; 0)
(1; 4) (0; 3)
.. ..
. .
This shows that @0 + @0 = @0 @0 = @0 . This result can be generalized in two directions to show (1) that a union of
countably many countable sets is countable and (2) that
(@0 )n = @0 : : : @0 = jjN
N : : : Njj
N = @0
N<! for the set of all finite ordered sequences (pairs, triples,) of natural
for all n 2 N, and hence (3) writing N
numbers and @<! 0 for supf(@0 )n jn = 0; 1; 2; : : :g , one has
@<!
0 N<! jj = @0 :
= jjN
These results have counterparts for the sets Z , Q
j j and A of integers, rational and algebraic
numbers (these last being the real numbers that are roots of polynomial equations with rational coefficients); namely,
jjZ
Z jj = jjQ
jj
jj = jjA
Ajj = @0 :
Finally, there is the generalization that + = = <! = for all infinite , and
+ = = supf; g for all infinite , , so cardinal addition and multiplication trivialize.
Exponentiation is not trivial. A bijection between subsets of A and functions from A to f0; 1g is obtainable by
relating a subset B with its , B .
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
(
1 if a 2 B
B (a) =
0 if a2
=B
Thus jj}(A)jj = 2jjAjj . Cantor proved that jjAjj < 2jjAjj . (Indeed, no function F : A ! }(A) can be a bijection,
since F (a) 6= fx 2 Ajx2 = F (x)g for any a.)
The cardinal c (for ) is defined to be jjR Ijj, where IRis the set of real numbers (or points in the line). By the
general result jjAjj jjAjj = jjAjj , c is equal to Ijj
jjR
I R or jjCj jj,
where Cj or IR IRis the set of complex numbers (or
@0
points in the plane). It can also be shown to be equal to jj}(N N)jj or 2 , and by the general result jjAjj < 2 one has
@0 < c. A consequence of this result is that there exist transcendental (that is, non-algebraic real)
numbers.
The continuum is the conjecture that 2 = + for all cardinals . The
(CH) is the special case for = @0 . Equivalently, CH asserts there is no intermediate cardinal with
@0 < < c. The most interesting alternatives to CH include the hypotheses CH0 : that there is just one such
intermediate; and CH# : that there are c intermediates. Where CH asserts that c = @1 and CH0 asserts that c = @2 ,
CH# implies that c > @1 ; @2 ; : : : ; @! , and more. Assuming the generalized continuum hypothesis, exponentiation
simplifies.
8
<
> if < cf
+
= if cf <
>
: +
if cf <
Otherwise, one must distinguish between a cardinal being a limit, meaning that + < whenever < , and
being a , meaning that 2 < whenever < .
The cf of a cardinal is the least such that there is a set B of sets with = jjBjj and jjbjj < for
all sets b 2 B, but jj [ Bjj = . is or according as cf is equal to or less than . @0 is
regular (otherwise cf@0 would be finite, but a union of finitely many finite sets is finite), @1 is regular (otherwise
cf@1 would be @0 , but a union of countably many countable sets is countable), and generally + is regular. @! is
singular, with cf@! = @0 (a union of countably many sets with cardinality @0 , @1 , @2 and so on has cardinality
@! ). It can at least be proved that c 6= @! , and more generally that cf2 > cf.
of the form ]a; b[ = fxja < x < bg for a < b and a interval is a set [a; b] = fxja x bg , a b.
B A is if it is disjoint from no open interval and if it is a subset of some closed interval. is
if there exists a countable dense subset B, and satisfies the chain (or ) if
there exists no uncountable set of disjoint open intervals. (Separability implies CCC since each of a set of disjoint
open intervals must contain a distinct element of the countable dense set.) is if every chain of closed
intervals has a non-empty intersection. Q j j , IRare each unbounded and dense in themselves, and Q j j is dense in IR(and
countable), so IRis separable and CCC. IRis complete, but Q j j is not (for example, the intervals
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
p
have intersection f 2g). Cantor proved that any unbounded, dense linear order on a countable set has order type .
Similarly, any unbounded, dense, separable, complete linear order has order type . The is that the
latter result holds with CCC in place of separability.
5 Point sets
The oldest part of set theory and the part most closely connected with the rest of mathematics (analysis, topology) is
the theory of sets of points in the line, plane and so on (identified with I,RIR I;R: : :). Modern analysis and topology
consider more complicated functions and sets than traditional calculus and geometry. A function F : IR! IRis
if F 1 [U ] is open for every open U I,Rwhere an set is a union of open intervals and a
set is the complement of an open set. The Cantor , the complement in the unit interval [0; 1] of
the union of the open intervals ] 13 ; 23 [, ] 19 ; 29 [, ] 79 ; 89 [, ] 271 ; 272 [, ] 277 ; 278 [, ] 19
27
; 20
27
[ and so on, is a famous example of a
closed set (see Continuum hypothesis 2) more complicated than any considered in traditional mathematics. Sets
obtainable from open sets by iterated application of complementation and countable union are called sets.
They occur ubiquitously in modern mathematics (as in the rigorous formulation of probability theory), while sets
beyond Borel occur sporadically.
An set is the image of a Borel set under a continuous function, a - set is the complement of an
analytic set, a -co- (PCA) set is the image of any set under a continuous function, and a
-pro-co- (CPCA) set is the complement of a pro-co-analytic set. Sets obtainable from Borel sets by
iterated application of continuous image and complementation are called . Projective sets are rather special
compared with arbitrary sets. It can be shown that there are only c projective sets, whereas there are 2c subsets of I.R
The study of such special classifications of sets is called set theory. In one notation, open sets are
denoted 01 , closed sets, 01 , unions of countably many closed sets, 02 and their complements, 02 ; : : : . Analytic,
co-analytic, PCA and CPCA sets are denoted, respectively, 11 , 11 , 12 and 12 ; : : : sets.
Central to attempts to define integration for complicated functions, and length, area and so on for complicated linear,
planar, sets, is the notion of a (complete probability) on a set or X. Such a measure is a function
from the set of ( -)measurable subsets of X to the unit interval J = [0; 1], (A) being called the ( -)measure
of A X . has the following properties: (1) - the whole space has measure one; (2) - -
any one-point set has measure zero; (3) - any subset of a set of measure zero has measure zero; (4)
- the complement of a measurable set is measurable; and (5) - the union of a countable
set of measurable sets is measurable and, if the sets are disjoint, the measure of their union is the sum of their
measures. The analogue of additivity with replaced by of cardinality less than that of the whole
may be called ( 50 ) . If we restrict ourselves to the unit interval, the of A J by
t 2 J is defined to be fa # tja 2 Ag, where x # y is defined to be whichever of x + y, x + y 1 is an element of
J, and one may consider the further property (6) - a translation of a measurable set is measurable, and its
measure is the same. The key positive result is the existence of a unique measure on J with properties (1)-(6), called
. (Analogues can also be defined on I,RIR IRand so on.) For Lebesgue measure, (A) intuitively
represents the probability that a point randomly chosen from J will be an element of A. One respect in which the
discontinuum is is that it is large in the sense of having cardinality c but small in the sense of having
Lebesgue measure 0. (Like some other positive results, the existence and uniqueness of Lebesgue measure require for
their proof only DC (see 1 above), not full AC.)
To mention a negative result, the hypothesis that all sets are Lebesgue measurable (AM) fails. Let S be a choice set for
fy 2 Jj(y x) 2 Q jj g
. If S were measurable, then using invariance it would follow that the disjoint sets of form
fy 2 Jjy # p 2 Sg for p 2 Q jj
, whose union is all of J, would all have the same measure ". But then using additivity,
(J) would have to be " + " + " + : : : , which is 0 or 1 according as " = 0 or > 0. The failure of AM leads to
consideration of some alternative hypotheses, notably the hypothesis (known to imply CH and CH0 ) of the
existence of a measure with properties (1)-(5) for which all sets are measurable, or the hypothesis (known to imply
CH# ) of the existence of a measure with properties (1)-(4) and superadditivity for which all sets are measurable. A
more positive result using AC may also be mentioned. Given Q I,Rand a continuous F from IRto I,Ra uniformization
for Q, F is a choice set P for ffy 2 QjF (y) = F (x)gjx 2 Qg . AC immediately implies the existence of a
uniformization for any Q, F.
The failure of AM also leads to consideration of some restricted hypotheses: for one of the special classifications
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
considered above, - is the hypothesis that all sets in are measurable. Similarly,
- is the hypothesis that for any Q 2 and any continuous F there is a uniformization P 2 . It can
be shown that while 1p -measurability and 1p -measurability imply each other for the same p, 1p -uniformization
and 1p -uniformization cannot both hold for the same p. Positive results of descriptive set theory obtained by the end
of the 1930s include 11 - and 11 -measurability, an example of a theorem or result of the form set
in has some , and also 11 - and 11 -uniformization, an example of a theorem or result
of the form every set in there is another set in related to it in some . All attempts to prove 12 -
and 12 -measurability, and 13 - or 13 -uniformization have failed.
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
according as x is or is not an element of A. A for a player is, intuitively, a rule telling the player how to play
as a function of the previous moves, and is winning if by so playing the player will win the game, no
matter what the moves. A is if one or other player has a winning strategy. 11 -AD, PD and
AD (to add to the notation set up in 5) are, respectively, the hypotheses that all 11 , projective or arbitrary sets are
determinate. AD is refutable using AC, among other reasons because AD implies AM, while, by essentially the same
proof, PD implies projective measurability, and various other regularity theorems. Similarly, PD implies projective
uniformization and various other structural theorems. 11 -AD yields, correspondingly, 12 - and 12 -measurability
and 13 -uniformization.
Despite superficial differences, there are deep connections between large cardinals and determinacy. By the 1970s it
was established that a large cardinal axiom somewhat weaker than MC implies a conclusion somewhat stronger than
11 -AD, with the partial converse that such a determinacy hypothesis implies the consistency of such a large cardinal
hypothesis. In the 1980s an analogous relation between SC and PD was established. So the cases for large cardinal
and for determinacy axioms reinforce each other. The search for further axioms to decide CH continues.
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
and so w appears to X to be the pair fu; vg of u; v if and only if it really is that pair, and so the axiom of pairs,
every u; v there exists a w such that w = fu; vg, holds inside X if and only if fu; vg 2 X whenever u; v 2 X , or,
as is said, if and only if X is closed under the pairing operation f ; g. Similarly it can be shown that the axiom of
union holds in X if and only if X is closed under the union operation [ that the axiom of infinity holds in X if and
only if ! 2 X ; and that the axiom of choice holds in X if and only if every partition in X has a choice set in X.
As for power, u appears to X to be a subset of v if and only if it really is, but w appears to X to be the set }(v) of all
subsets u of v if and only if w really is the set }(v) \ X of all subsets u of v with u 2 X . Thus the axiom of power
appears to X to hold if and only if X is closed under the operation of forming }(v) \ X from v, which may be the case
even if X does not contain all subsets of v and therefore is not closed under the operation of forming }(v) from v.
Similarly, a necessary and sufficient condition for separation to hold inside X is that for each condition , X is closed
under the operation that applies to any v forms fu 2 vjX (u)g . Similarly, a necessary and sufficient condition for
replacement to hold inside X is that for each condition such that for every x 2 X there is a unique
y = X (x) 2 X such that X (x; y) holds, X is closed under the operation that applied to v forms f X (u)ju 2 vg .
X is if whenever u v 2 X , then u 2 X . For supertransitive X, }(v) \ X = }(v) for any
v 2 X , so the necessary and sufficient condition for power to hold is just that X is closed under the power operation
}. Supertransitivity is a more than sufficient condition for separation to hold. Similarly, a more than sufficient
condition for replacement to hold is that whenever u X , jjujj = jjvjj and v 2 X , then u 2 X .
One kind of model of a large part of ZFC is obtained as follows. For any x let the t(x) of x be
[fx; [x; [ [ x; : : :g , the set of all elements, elements of elements and so on of x. For any uncountable cardinal , let
H() be fxj > jjt(x)jjg . It follows from the definition that H() is supertransitive, and it can be checked, using
the necessary and sufficient conditions indicated above, that H() is a model of all axioms of ZFC except perhaps
power and replacement. Moreover, if is a strong limit, then H() will be closed under } and hence a model of
power; while if is regular, then one will have u 2 H() whenever u H(), jjujj = jjvjj , v 2 H() and
H() will be a model of replacement. (Thus if is inaccessible, H() is a model of the whole of ZFC.)
Another kind of model of a large part of ZFC is obtained as follows. The hierarchy of levels of sets, intuitions about
which underlie the axioms (as in 1), can be defined by transfinite recursion.
V (0) = ;
V ( + 1) = }(V ())
at limits V ( ) = [fV ()j < g
That V () is supertransitive can be proved by transfinite induction. Given the other axioms of ZFC, the principle of
replacement can be proved equivalent to the following principle of : for any formula there are arbitrarily
large ordinals such that is absolute for V (). In particular, for any finite conjunction of axioms of ZFC there
are arbitrarily large ordinals such that holds in V ().
While both the V () (for > !) and the H() (for > @0 ) are uncountable, it can be shown that if a ZFC or
some variant set theory has a transitive model, then it has a countable transitive model, using a celebrated result of
mathematical logic (see -Skolem theorems and nonstandard models). Since the uncountability of the power
set of the set of natural numbers is a theorem of set theory, the existence of a countable model for set theory is
sometimes considered paradoxical, and called the . There is, however, no genuine contradiction. (For
the conditions x = }(!) and jjxjj > @0 are not absolute: the former holds inside a model X if x = }(!) \ X ,
which may well be countable even though }(!) is uncountable; the latter holds inside the model X if there is no
surjection F 2 X from ! to x, which there may well not be even if there is such an F 2 = X .)
See also: theorem; Logical and mathematical terms, glossary of
JOHN P. BURGESS
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)
H. Putnam (eds) Philosophy of Mathematics: Selected Readings, Cambridge: Cambridge University Press, 2nd
edn, 1983.(A philosophical puzzle about as in 1.)
Cantor, G. (1895, 1897) zur der transfiniten , Mathematische Annalen 46:
481-512, 49: 207-46; trans. P.E.B. Jourdain (1915), Contributions to the Founding of the Theory of Transfinite
Numbers, New York: Dover, 1955.(The original source for 2-3, with an introduction setting the work in its
historical context.)
Halmos, P. (1960) Naive Set Theory, New York: Van Nostrand.(A perennially popular elementary textbook,
expounding for students of mathematics the development of mathematics on the basis of ZFC mentioned in 1.)
Jech, T. (1977) the Axiom of , in J. Barwise (ed.) Handbook of Mathematical Logic, Amsterdam:
North Holland, 345-70.(Surveys all aspects of AC mentioned here, and many more besides.)
Kunen, K. (1977) , in J. Barwise (ed.) Handbook of Mathematical Logic, Amsterdam: North
Holland, 370-402.(Surveys at length combinatorial set theory as mentioned in 4; and, briefly, large cardinals as in
6.)
Maddy, P. (1988) the , Journal of Symbolic Logic 53: 481-511, 736-64.(Surveys axioms old and
new as in 1, 6; especially advances made in the 1980s.)
Martin, D.A. (1977) Set , in J. Barwise (ed.) Handbook of Mathematical Logic, Amsterdam:
North Holland, 783-818.(Surveys theory of projective sets (as in 5-6) through the 1970s.)
Shoenfield, J.R. (1977) of Set , in J. Barwise (ed.) Handbook of Mathematical Logic, Amsterdam:
North Holland, 321-44.(Surveys material mentioned in 1-4, 6.)
Zermelo, E. (1908) die Grundlagen der Mengenlehre, , Mathematische Annalen 65: 261-81;
trans. S. Bauer-Mengelberg, in the Foundations of Set , in J. van Heijenoort (ed.) From
Frege to : A Sourcebook in Mathematical Logic, Cambridge, MA: Harvard University Press, 1967,
200-15.(The original source for the axiomatization of set theory, with an introduction setting the work in its
historical context.)
Routledge Encyclopedia of Philosophy, Version 1.0, London and New York: Routledge (1998)