Looking Back and Looking Forward

Loosely speaking, a formal mathematical system refers to some objects (of unspecified
nature),some axioms that describe precisely how the objects behave, and the body of
definitions and theorems that grow out of the axioms. A Peano system is a formal
mathematical system.

An informal mathematical system is a not very precise, everyday term for how we view a
body of mathematical knowledge, often very well-developed, when we are not bothering
to think about its basis at the axiomatic level.

An example would be our informal system of whole numbers, =, together with all its
arithmetic. We have been taking the point of view that the system of whole numbers, =,
has not yet been carefully defined, even though it is a system that (somehow, informally)
we seem to know a lot about. We have, as yet, no answer for the question “Just what
exactly is the whole number system?”

We believe that the Peano Axioms P1-P5 are true in this informal system, =, if we
interpret the objects in c as whole numbers and interpret successor in c to mean “the
next whole number.” Our understanding  although we only have informal knowledge
about = to judge by  is that the system of whole numbers is a model for a formal Peano
system c .

Moreover we can start with a Peano system, define  , † , Ÿ in c , and prove theorems
about how these “abstract” operations behave. These theorems turn out (when
interpreted in the model =) to be in agreement with how we understand arithmetic in the
informal system, =.

For example, we proved that in c À 7  ! œ 7 œ !  7 and 7 † " œ 7 œ " † 7à that

addition and multiplication are commutative and associative; and that the distributive
law connects addition and multiplication. We also proved some other theorems about
Peano arithmetic, such as the cancellation laws for addition and multiplication. We
defined an order Ÿ in c which interacts nicely with addition and multiplicationÞ

Because mathematicians don't care (unless they are becoming philosophers) what the
whole number “really are”, and because a formal, abstract Peano system c seems to
perfectly mirror the informal system =, mathematicians are perfectly happy to accept as a
formal definition that = is some Peano system or another.

Since all Peano systems “look exactly alike,” it doesn't really matter much which
particular Peano system we define = to be. However, set theory is supposed to be the
foundation for mathematics, so it would be nice to define = to be some particular Peano
system whose objects are sets.

To do this, we would need to have:

 i) a collection of sets, and

ii) an operation (called “successor” ) that, within this collection,

assigns to each set B a successor set B and does so in such a way that
the axioms P1-P5) are true.

This would be a Peano system whose objects are actually sets, and which (like any Peano
system) would behave just like the whole number system. We could then define  once
and for all, officially and formally  this particular Peano system to be the system = of
whole numbers. We would call the sets (objects) in = whole numbers.

Suppose this can be done.

i) We are not claiming to have proved that = is this particular Peano system  it
makes no sense to prove a definition. But the official, formal definition for =
does give us a system of sets which behaves, as best we can judge, just like the
informal system = that we started with.

ii) There might also be other ways in which someone could officially define = and
get a system that behaves in the right way. For philosophical or aesthetic reasons,
someone might prefer this other approach. But mathematically, the choice really
doesn't matter: how the whole numbers behave is all the that really counts
mathematically  so we can all manage to live with whatever particular formal
definition for = was chosen.
Getting a Peano system of sets

We want to actually construct a Peano system whose objects are sets in order to complete
the official definition of = in terms of set theory.

The notion of “successor” for sets came up earlier in a Homework Set 3. Here is the
definition again.

Definition For any set B, B œ B  ÖB×Þ The set B is called the successor of BÞ

Notice that

i) We can form the successor of any set whatsoever. For example,

Ö+ß ,× œ Ö+ß ,×  Ö Ö+ß ,×× œ Ö+ß ,ß Ö+ß ,××, and

‘ œ ‘  Ö‘×

ii) Since B − B ß it is always true that B Á g.

iii) If C − Bß then C − B  ÖB× œ B . Therefore B © B is always true.

(Note: iii) tells us that “successor sets” B are rather special. If you pick
a set D “at random” and B − D , then it usually doesn't happen that B © D
is also true.

Example

g
g œ g  Ög× œ Ög×
g œ Ög× œ Ög×  ÖÖg×× œ Ögß Ög××
g 
œ Ögß Ög×× œ Ögß Ög××  ÖÖgß Ög××× œ Ögß Ög×ß Ögß Ög××
ã

and so on.

Definition Suppose M is a collection (set) of sets. M is called inductive if

 a) g − M , and
b) if B − Mß then B − MÞ

Then we ask: are there any inductive sets? Informally, it certainly looks (informally) like
there are. For example, we could form the set

Ögß g ß g ß g ß ÞÞÞ×

However, this “example” of an inductive set (although fine, informally) seems just a little
shaky if we're being very careful. After all, this set is only inductive because it is
described using a rather casual “etc.” Ðthat is, “ÞÞÞ” ÑÞ

At this point, we should take a brief look at the axioms for set theory itself. This is “as
deep as we ever dig” since set theory, we have agreed, is to be the very foundation for all
mathematics. In other words, all mathematics flows from these axioms.

Axiomatic (that is, formal) set theory starts with an abstract system of unspecified objects
(called sets) and a relation “ − ” among them. We know absolutely nothing about these
objects except that they behave in ways described by 10 axioms  almost always called
the ZFC axioms ( œ “Zermelo-Fraenkel-with Choice” axiomsÑÞ (Of course, these axioms
were chosen to create a formal system of objects that would behave, as best we can tell,
“just like” naive (informal) set theory.)

A careful study of the ZFC Axioms, and the theorems that can be proved from them, is a
whole field of study in itself, usually called “Axiomatic Set Theory.” In order to get the
flavor, here is a partial list of the axioms, omitting some of the more technical ones that
we don't need to think about at all for this course. (The quantifiers in these axioms apply
to the “universe” of sets  so ÐaBÑ means “for all sets B”, etc. Convince yourself that
the English translations given below are correct.)
Zermelo-Fraenkel-with-Choice (ZFC): Axioms for Set Theory

ZFC1 aB aC Ð aD (D − B Í D − CÑ Í B œ CÑÑ
“ Two sets are equal iff they have the same members. ”

ZFC2 bB aC ( C Â B)
“ There is a set with no members (an empty set). ”

If “two” sets both have no members, then they certainly have the same
members (none at all !) so, by ZFC1, they are the same set. In other
words, ZFC1 implies that there is only one empty set. For convenience
we can give it a name: g

ZFC3 aB aC ÐbD (a? Ð? − D Í Ð? œ B ” ? œ CÑÑÑ

“ If B and C are sets, then there is a set D œ ÖBß C×Þ”

ZFC4 aB bC ÐaD ÐD − C Í b, ÐD − , • , − B ÑÑÑ

“ For any set B, there is a set D consisting of the members of the
members of B.”

We can agree to give this set z a name: -B

ZFC& aB bC (aD (D − C Í ÐaA ÐA − D Ê A − BÑÑÑÑ

If we agree to define “D © B” to be shorthand for aA ÐA − D Ê A − BÑ

Then ZFC5 could be written aB bC ÐaD ÐD − C Í D © BÑÑ. That is,
axiom ZFC5 says “every set B has a power set.”

ZFC6 bM Ðg − M • (aC (C − M Ê C  − MÑ Ñ
There exists an inductive set.

plus 4 other more technical axioms (omitted ) À ZFC7, ZFC8, ZFC9,

ZFC10 (AC)

The axiom ZFC10 is called the Axiom of Choice (AC). It arouses some
philosophical controversy among those mathematicians who worry about
foundations of mathematics, so a few mathematicians omit it from the
list. The 10-axiom system ZFC is the axiom system most
mathematicians would use for set theory (and therefore for all mathematics).
If AC is omitted, then ZFC1-ZFC9 are referred to as the “ZF” axioms.
We may say a little about the Axiom of Choice later in the course.
Although we have taken a naive (informal) approach to set theory, everything that we
have done (or will do) with sets can be justified by theorems provable from the ZFC
axioms.

Returning to the question we were asking: we are convinced (informally) that set theory
should contain an inductive set  for example, the one described informally by

Ögß g ß g ß g ß ÞÞÞ ×

In formal set theory, the existence of an inductive set (at least one, maybe more)
is guaranteed by axiom ZF6  it's built right in so that the formal ZFC system will
contain an object that our intuition expects to be there.

Definition I‡ Choose an inductive set M and.

define = œ ÖB − M À B − N for every inductive set N ×Þ

This set, =, together with the successor operation for sets, is going to turn out to
be a Peano system. And it is going to be the particular Peano system that we use
as the official definition for the system of whole numbers. For now, choosing to
call this set “=” is in anticipation of what's going to happen later. But for the
moment, it is nothing but called =, which just happens to have the same name as
the system of whole numbers.

By definition, = © M , but the definition tells us that, in fact, = is a subset of every

= œ +ÖN À N is an inductive set×Þ

inductive set N (including, for example, N œ MÑÞ Therefore you could think of = as

As a matter of fact, axiomatic set theory contains many inductive sets. But

i) The definition of = would make sense even if there were only one
inductive set, M .
ii) Convince yourself that if a different inductive set M w were used instead
of M in the definition, then the resulting set = w would the same would be the
same: = w œ =Þ

Although = is an intersection of inductive sets, we can't assume just assume that makes =
itself an induction set. But the next theorem confirms that, in fact, it is. .

Theorem 17 = is an inductive set.

Proof a) g is in every inductive set N ß so g − =Þ

b) Suppose B − =. Then (by definition of =) B is a member of every inductive

set N Þ Therefore B is also a member of every inductive set N (by definition of
“inductive set”). Hence B − =Þ
Therefore = is inductive. ñ

Since = © N for every inductive set N , and because we now know that = itself is an
inductive set, we can now say that = is the smallest inductive set.

Since g − =; therefore g − =à therefore g − =à therefore g − =à and so on.

Caution: We are using the same notation B for “successor of a set B” as we

used for “successor” in a Peano system. But don't let the notation deceive: we
have no right to assume that the successor operation for sets obeys the rules
axioms P1-P5. We need to check whether that is true.

(It does turn out to be true; when we have proved that, then we will know that the
set =, with the set successor operation, is a Peano system.)

Definition ! œ g
(We are simply agreeing that ! will be another name for g.)

Since g − =, we now have

P1: There is a special object in = named !.

( ! is the set gÞ )

Because = is inductive, the successor set B for each set B in = is also in =. Therefore

P2: For every object (set) B − =, there is a successor B in =.

We remarked earlier that B Á g for every set B. Therefore

P3: For all B − =, B Á ! Ð œ gÑ

Suppose E © =, and suppose that

i) ! Ð œ gÑ − E, and
ii) ÐaB − =Ñ ÐB − E Ê B − EÑ

then E, by definition, is an inductive set. But = is the smallest inductive set, so = © EÞ

Therefore E œ =Þ In other words:

P5: Suppose E © = À
if ! Ð œ gÑ − E, and
if ÐaB − =Ñ ÐB − E Ê B − EÑ,
then E œ =Þ