-- an introduction and links collection by

Eric Schechter, Vanderbilt University

The Axiom of Choice (**AC**) was formulated about a century ago,
and it was controversial for a few of decades
after that; it might be considered **the last
great controversy of mathematics**. It is now a basic
assumption used in many parts of mathematics. In fact,
assuming AC is equivalent to assuming
any of these principles (and many others):

- Given any two sets, one set has cardinality less than or equal
to that of the other set -- i.e., one set is in one-to-one
correspondence with some subset of the other. (
*Historical remark:*It was questions like this that led to Zermelo's formulation of AC.) - Any vector space over a field F has a basis --
i.e., a maximal linearly independent subset -- over that field.
(
*Remark:*If we only consider the case where F is the real line, we obtain a slightly weaker statement; it is not yet known whether this statement is also equivalent to AC.) - Any product of compact topological spaces is compact. (This is
now known as
*Tychonoff's Theorem*, though Tychonoff himself only had in mind a much more specialized result that is*not*equivalent to the Axiom of Choice.)

AC has many forms; here is one of the simplest:

Axiom of Choice.
Let C be a collection of nonempty sets. Then we can choose
a member from each set in that collection. In other words, there
exists a function f defined on C with the property that, for each
set S in the collection, f(S) is a member of S. |

The function *f* is then called a *choice function*.

To understand this axiom better, let's consider a few examples.

- If
*C*is the collection of all nonempty subsets of {1,2,3,...}, then we can define*f*quite easily: just let*f*(*S*) be the smallest member of*S*. - If
*C*is the collection of all intervals of real numbers with positive, finite lengths, then we can define*f*(*S*) to be the midpoint of the interval*S*. - If
*C*is some more general collection of subsets of the real line, we may be able to define*f*by using a more complicated rule. - However, if
*C*is the collection of*all*nonempty subsets of the real line, it is not clear how to find a suitable function*f*. In fact, no one has ever found a suitable function*f*for this collection*C*, and there are convincing model-theortic arguments that no one ever will. (Of course, to*prove*this requires a precise definition of "find," etc.)

- If we follow the constructivists, and "exist"
means "find," then the axiom is
*false*, since we cannot find a choice function for the nonempty subsets of the reals. - However, most mathematicians give "exists" a much weaker
meaning, and they consider the Axiom to be true:
To define
*f*(*S*), just arbitrarily "pick any member" of*S*.

To assert that a mathematical object "exists," even when you cannot give an example of it, is a little bit like this: Suppose that one day you go to a football game by yourself. There are thousands of other people in the stadium, but you don't know the names of any of them. (And let's suppose you're shy, so you're not about to ask anyone their name.) Then you know those people have names, but you cannot give any of those names. (Admittedly, this is only a metaphor, and not a perfect one; don't make too much of it.)

The "existence" of *f* -- or of any mathematical
object, even the number "3" -- is purely formal. It
does not have the same kind of solidity as your table and your
chair; it merely exists in the mental universe of mathematics.
Many different mathematical universes are possible. When we
accept or reject the Axiom of Choice, we are specifying
something about which mental universe we're choosing to
work in. Both possibilities
are feasible -- i.e., neither accepting nor
rejecting AC yields a contradiction; that fact follows from
models devised by
Gödel
and Cohen.
However,
most "ordinary" mathematicians -- i.e., most mathematicians
who are not logicians or set theorists --
accept the Axiom of Choice chiefly
because their work is *simpler* with the Axiom of Choice
than without it.

Bertrand
Russell was more famous for his work in philosophy and political
activism, but he was also an accomplished mathematician.
His book *Introduction to Mathematical Philosophy* includes
some discussion of AC. Here is my paraphrasing of part of what he said:

To choose one sock from each of infinitely many pairs of socks requires the Axiom of Choice, but for shoes the Axiom is not needed.The idea is that the two socks in a pair are identical in appearance, and so we must make an arbitrary choice if we wish to choose one of them. For shoes, we can use an explicit algorithm -- e.g., "always choose the left shoe." Why does Russell's statement mention

Jerry Bona once said,

The Axiom of Choice is obviously true; the Well Ordering Principle is obviously false; and who can tell about Zorn's Lemma?This is a

For another indication of the controversy that initially surrounded the
Axiom of Choice, consider this anecdote (recounted by
Jan Mycielski in *Notices of the AMS*
vol. 53 no. 2 page 209). Tarski, one of the early great researchers in set theory and
logic, proved that AC is equivalent to the statement that any infinite set X has the
same cardinality as the Cartesian product *Comptes Rendus Acad. Sci. Paris*, where it was refereed by two very famous mathematicians,
Fréchet and Lebesgue. Both wrote letters rejecting the article. Fréchet wrote that
an implication between two well known truths is not a new result. And
Lebesgue wrote that an
implication between two false statements is of no
interest. Tarski said that he never again submitted a paper to the *Comptes Rendus*.

AC permits arbitrary choices from an arbitrary collection of nonempty sets. Some mathematicians have investigated some weakened forms of AC, such as

- CC (Countable Choice), which permits arbitrary choices from a sequence of nonempty sets.
- DC (Dependent Choice), which permits the more general process of selecting arbitrarily from a sequence of nonempty sets where only the first set is specified in advance; each subsequent set of options may depend somehow on the previous choices. This is precisely what is needed for some choice processes in topology and analysis -- e.g., for the proof of the Baire Category Theorem.

The full strength of the Axiom of Choice does
not seem to be needed for *applied* mathematics.
Some weaker principle such as CC or DC
generally would suffice. To see this, consider that
any application is based on measurements, but humans
can only make *finitely many* measurements.
We can extrapolate and take limits, but usually
those limits are sequential, so even in theory we
cannot make use of more than countably many
measurements. The resulting spaces are separable.
Even if we use a nonseparable space such as
L^{∞}, this
may be merely to simplify our notation; the
relevant action may all be happening in some
separable subspace, which we could identify
with just a bit more effort. (Thus, in some
sense, nonseparable spaces exist *only*
in the imagination of mathematicians.) If we restrict
our attention to separable spaces, then much
of conventional analysis still works with
AC replaced by CC or DC. However, the
resulting exposition is then more complicated,
and so this route is only followed by a few
mathematicians who have strong philosophical
leanings against AC.

A few pure mathematicians and many applied mathematicians
(including, e.g., some mathematical physicists) are
uncomfortable with the Axiom of Choice. Although
AC simplifies some parts of mathematics, it also
yields some results that are unrelated to, or perhaps even
contrary to, everyday "ordinary" experience; it implies the
existence of some rather bizarre, counterintuitive
objects. Perhaps the most bizarre is the
**Banach-Tarski Paradox**:
It is possible to take the 3-dimensional
closed unit ball,

At first glance, the Banach-Tarski result
seems to contradict some of our intuition about physics
-- e.g., the Law of Conservation of Mass, from classical
Newtonian physics. If we assume that the ball has
a uniform density, then the Banach-Tarski Paradox
seems to say
that we can disassemble a one-kilogram ball
into pieces and rearrange them to get two
one-kilogram balls. But actually, the
contradiction can be explained away:
Only a
set with a defined *volume* can have a defined mass.
A "volume" can be defined for many subsets
of **R**^{3} ---
spheres, cubes, cones, icosahedrons, etc. --- and in fact
a "volume" can be defined for nearly any subset of
**R**^{3} that we can think of.
This leads beginners to expect that the notion
of "volume" is applicable to
*every* subset of **R**^{3}.
But it's not. In particular, the pieces in the
Banach-Tarski decomposition are sets whose
volumes cannot be defined.

More precisely,Lebesgue measureis defined on some subsets ofR^{3}, but it cannot be extended toallsubsets ofR^{3}in a fashion that preserves two of its most important properties: the measure of the union of two disjoint sets is the sum of their measures, and measure is unchanged under translation and rotation. The pieces in the Banach-Tarski decomposition are not Lebesgue measurable. Thus, the Banach-Tarski Paradox gives as a corollary the fact that there exist sets that are not Lebesgue measurable. That corollary also has a much shorter proof (not involving the Banach-Tarski Paradox) which can be found in every introductory textbook on measure theory, but it too uses the Axiom of Choice.

Here is a brief sketch of that shorter proof: Work in "the real numbers modulo 1" -- that is, the number system that you get if you cut the interval [0,1) out of the real line and loop it around into a circle, so that 0 and 1 are the same number. (Like the way that 0 and 12 are the same on a circular clock.) In that number system, multiplication and division don't really work very well any more, but addition and subtraction still work fine, and so does Lebesgue measure. Let's call that number system T; its entire measure is 1. Now, the Axiom of Choice is used to "construct" a rather peculiar subset of T -- let us call it C -- with the property that the sets C+r = {x+r : x in C} are all disjoint from each other, for different values of the rational number r. The union of these sets is all of T. Now, if C were measurable, then so would each C+r, and they would all have the same measure, and their measures would add up to the measure of T -- that is, they would add up to 1. But how many of these C+r are there? There are a countable infinity of them. If the measure of C were zero, their sum would be zero. If the measure of C were positive, their sum would be infinite. You can't get 1, either way.

Personally, I am not surprised to find the Axiom of Choice coming into play in a subject that is so inherently complicated as unmeasurable sets. I am much more surprised to find AC coming into play in this simpler and more concrete example: I want to classify all subsets of {1, 2, 3, 4, 5, . . .} as "small" or not "small," defining the word "small" in such a way that

- any set with zero or one members is "small";
- any union of two "small" sets is "small"; and
- a set is"small" if and only if its complement isn't "small."

- Define "small" to mean "finite." This satisfies rules a and b. But it does not satisfy rule c, since the even numbers and the odd numbers are complements of each other, and neither of those sets is finite.
- Say that a set is "small" if the number 1 is not a member of that set. This definition satisfies rules b and c, but it classifies the set {1} as "not small," thus failing rule a.
- Say that a set is "small" if it contains at most one of the three numbers 1, 2, 3. That satisfies rules a and c. But it classifies the sets {1} and {2} as small and the set {1,2} as not small, thus failing rule b.

Technical details for experts:To prove the existence of such a classification scheme, just call "large" the members of somenonprincipal ultrafilteron the positive integers, and call their complements "small." Note that, with this scheme, any superset of a "large" set is also "large." The converse is slightly more complicated: If you have a "small/large" classification, the "large" sets do not necessarily necessarily form a nonprincipal ultrafilter, but the supersets of "large" sets do. An introduction to nonprincipal ultrafilters can be found in my book and in many other places in the literature.

The existence of nonprincipal ultrafilters follows easily from Zorn's Lemma, by arguments that will be obvious once you've digested all definitions involved (admittedly not a small task). But showing that the existence proof isinherently nonconstructiveis much harder, and requires some definitions that I've made up. By an "example" I mean anything whose existence can be proved using just ZF+DC --- that is, I'll allow Dependent Choice but no higher relatives of AC.

Let BP be the statement that "every subset of the reals has the Baire property." The existence of a nonpricipal ultrafilter on the integers implies not-BP (by fairly straightforward functional analysis and topology). But in 1984 Shelah proved that the consistency of ZF implies the consistency of ZF+DC+BP. Therefore,ifordinary set theory is free of contradictions, then ZF+DC cannot be used to prove not-BP. I say "if" because we don't know that for sure, and Gödel's Incompleteness Theorem assures us that we never will know the consistency of ZF for sure. However, I would say that ZF isempiricallyconsistent: In a century of study, mathematicians have not yet found any contradictions in ZF, despite the incentive that any mathematician finding such an important proof would instantly be promoted to full professor at any university in the world.

My example with positive integers might appear to be simpler than the Banach-Tarski Paradox, but it does not really get us completely away from measure theory. A nonprincipal ultrafilter can be reformulated as a two-valued probability measure that is finitely additive but not countably additive.

In the preceding paragraphs I have attempted to
introduce the Axiom of Choice in the language of
*informal set theory* (also known as "naive set theory"), in
which one assumes that sets are
"collections of objects," with the meanings of the
words "collection" and "object"
based on our everyday nonmathematical experience.
However, this web page is merely an introduction to the subject,
and gives no indication of the proofs. A
mathematically precise study of AC would require *formal set
theory* (also known as "axiomatic set theory"); that
is the language spoken by the real experts in this subject.
In formal set theory, we put aside any nonmathematical,
notions of "set". Instead of working with imagined sets,
we work with *sentences* about some objects that we
*call* "sets." We begin with a list of axioms,
i.e., properties that these abstract objects are assumed
to satisfy. Generally we assume axioms that seem intuitively
reasonable, but we assume as few axioms as possible --
we try to choose axioms that enable us to prove the
other properties that we want.
We assume nothing else beyond what is
contained in those axioms; we may not use a property
of sets just because it is "familiar" or "obvious."
Then we study the nature of the proofs, to determine
what kinds of things can or cannot be proved.
The axiomatic approach is much drier, and less appealing
to all but a few specialists who are interested in it;
but its conclusions are much more reliable than
any mere, informal discussion. In this web page I have
attempted to summarize, in informal
language, some of the conclusions that are reached
by that formal theory.

Please write to me if you have suggestions for additions or alterations to this web page. However, I will warn you that I am

- Another online introduction currently available for AC is the Math FAQ. It also includes an introduction to the Banach-Tarski paradox.
- For more extensive information about the Banach-Tarski Paradox, see Stan Wagon's book.
- Handbook of
Analysis and its Foundations, by Eric Schechter.
The website is an advertisement,
but it does include a few interesting excerpts from the book
-- e.g., a
**list of 27 forms of the Axiom of Choice**and a few dozen weak forms of Choice, as well as a chart showing how some of the weak forms are related. (The book is intended for beginning graduate students; only a small portion of the book is actually concerned with Choice.) - The Math History Archive contains some mentions of AC. See especially: the beginnings of set theory, Cantor, Hausdorff, Zermelo, Sierpinski, Kuratowski, Zorn. Gödel, Cohen
*Zermelo's axiom of choice : its origins, development, and influence*, by Gregory Moore. A fascinating history of AC. Originally published by Springer, now available as an inexpensive reprint from Dover. Here is a web page giving the table of contents of that book.- Constructivism is Difficult -- a brief introduction to constructivism. Constructivism and AC are two different but overlapping topics. AC is a nonconstructive axiom; that's what made it so controversial.
- Topological Equivalents of the Axiom of Choice and of Weak Forms of Choice by Eric Schechter -- a brief introduction to this subject, which I wrote for the ``Topology Atlas''
- One of AC's most important applications in analysis is the Hahn-Banach Theorem. It may be viewed as a weak form of Choice. Here is an on-line survey article, The Hahn-Banach Theorem: The Life and Times, by Lawrence Narici and Edward Beckenstein.
- Axiome du Choix, by David Madore. This page includes a list of several equivalents and weaker consequences of AC, and a list of some of the implications among them. If your French is weak, you might try an automatic translation service, such as AV's translator. However, such translation programs don't know mathematics, and so some of the results may be a bit odd. For instance, what we call "well ordering" is what the French call "bon ordre," but the AV translator turns that into "good command."
- An introduction to the generalized Riemann integral, by me (Eric Schechter). This integral -- actually more general than the Lebesgue integral, but much simpler to define -- is a good topic to include in the "introduction to real analysis" course that many universities teach to advanced undergraduates majoring in math. Near the end of the web page, the Axiom of Choice arises naturally in a discussion about the measurability of sets.
- The ProvenMath website includes a web page on AC. That page includes formal statements of several of the equivalents of AC, and formal proofs of equivalence. To read the statements and proofs, you'll need to familiarize yourself with the notation that ProvenMath uses for representing mathematical symbols on web pages. Also on the same server is a page about how the Axiom of Choice affects the relation between sequential continuity and epsilon-delta continuity.

- Consequences of the Axiom of Choice is a book by Paul Howard and Jean E. Rubin that was published by the American Mathematical Society in 1998. It is a vast survey of Choice and its weaker relatives. It is a reference book, not intended for beginners. The authors are continuing their research project, which now goes a bit beyond the book. Their web page contains a list of the errata and addenda to the book, and a form for downloading copies of the project's main tables. You may also want to look at some related papers.
- Home page of Thomas Jech.
Jech is the author of the book titled
*The Axiom of Choice*, which is not recent but is still excellent. He has worked in set theory, logic, and other areas since then. Some of his papers are available online. - Saharon Shelah's papers. Shelah is one of the leading logicians of our century; he has made great contributions to the theory of forcing. My favorite among his results is the fact that Con(ZF) implies Con(ZF + DC + BP); this result was shown by J.D.M.Wright to be important to functional analysis. (It's explained further in my book.) Among Shelah's subpages is a list of his coauthors, many of whom have web pages of their own.
- Andreas Blass's home page. It is a very old result that the Axiom of Choice implies the existence of bases for vector spaces; Blass can be credited with proving the converse. Blass also did some of the early work on proving the unconstructability of nonprincipal ultrafilters. (Those are results of his that I've understood; he has probably done some other much more important things that I don't understand.)
- Edward Nelson's home page. Nelson is the father of Internal Set Theory (IST), a variant of Nonstandard Analysis. IST has acquired a large following; some analysts are of the opinion that IST is the most intuitive approach to limits. (Personally, I suspect that most of those analysts were first trained as logicians, but that may be a reflection of my own ignorance.) Some of Nelson's writings are available online.
- Home page of Fred Richman, a leading constructivist. The page includes some amusing subpages, e.g., Confessions of a formalist, Platonist intuitionist -- (Hey, I thought those terms contradicted each other!) -- and Math without countable choice.
- Home page of John L. Bell, another leading researcher. Bell's page includes many interesting links and downloadable papers. Among those papers, Axiom of Choice and Zorn.pdf may be of particular interest to beginning and intermediate readers about AC. It includes (among other things) a chronology of choice principles, a chronology of maximal principles, and a list of many choice principles.

- Realism in mathematics -- book review by Morris Hirsch (in AmsTex format) of book by Penelope Maddy.
- The Dehn Invariant, explained by Douglas Zare. Hilbert's Third Problem, solved. One step of the solution uses AC. (David Joyce has prepared a nice website about Hilbert's 23 problems.)
- A theorem about polyomino tilings proved with AC by Michael Reid
- How to Fill n-Dimensional Space with Hoops, by Evelyn Sander [intermittent server?]
- This Week's Finds in Mathematical Physics (Week 68), by John Baez; includes some discussion of topoi
- New
Foundations
home page, by Randall Holmes. NF is a refinement of
Russell's theory of types, introduced by Quine in 1937. Thus, it is an
alternate form of set theory or higher-order logic, a little different
from conventional set theory, but still capable of doing approximately
the same things. Of course, it differs from the usual set theory in a
few respects; an obvious difference is that there is a universal set
in NF. A surprising difference is that
*the Axiom of Choice is false in NF*; this was established in a 1953 paper by E.P.Specker. If Holmes's website interests you, you might continue with T.E.Forster's 1995 book. (For bibliographic details see Holmes's website.) - Issues in commonsense set theory, an online article by Mujdat Pakkan and Varol Akman. From the abstract: "In this survey, we briefly review classical set theory from an AI perspective, and then consider alternative set theories." Includes an overview of ZF set theory, which makes it relevant to this home page.

- Metamath Solitaire, an elementary game implemented as a Java applet that lets you prove simple theorems in logic and set theory. Includes introductory explanation.
- Mathematical Logic around the world -- linklist provided by the Mathematical Logic Group in Bonn
- Isabelle
is a generic theorem prover which can support a wide variety of logics;
it is available for free and will run on most Unix systems.
You may be interested in some of
Larry
Paulson's papers on Isabelle.
Especially, you may be interested in "Mechanising
Set Theory: Cardinal Arithmetic and the Axiom of Choice", coauthored by
Krzysztof Grabczewski. This paper mechanizes
the proof of numerous equivalents of the Axiom of Choice, covering
most of Chapter 1 of Kunen's
*Set Theory*and most of Chapters 1 and 2 of Rubin and Rubin's*Equivalents of the Axiom of Choice*. - Merging HOL with Set Theory (Talk) -- (I don't really know what this is)
- Logic Eprints -- there is probably some stuff on AC in here, but I haven't explored it yet.
- A Survey of the QED Project and Related Topics

- Newsgroups: sci.logic, sci.philosophy.tech, sci.math, sci.math.research sometimes contain something about AC.
- Kurt Godel in Blue Hill, a description of Godel and some of his work, by Peter Suber
- The Axiom of Choice was used for
a tongue-in-cheek "proof" of the existence of God, by
Robert K. Meyer in
"God exists!",
*Nous***21**(1987), 345-361. The basic idea is to put a suitable partial ordering on the universe, and then use Zorn's Lemma to prove the existence of a maximal element, which is therefore God. A web page reprinting the proof has been made available by Alexander Pruss. - A musical band named Axiom of Choice. Their music is a fusion of Persian Traditional with modern. Okay, it's not math, but I couldn't resist posting this here. Iranian born guitarist Ramin Torkian and singer Mamek Khadem were both trained as mathematicians, and that's where the group gets its name. Their first album's liner notes say "There is an exciting and profound artistic value in the mathematical principle, Axiom of Choice. The mathematician has the right to choose elements without explanation. In a world where everything must be explained, these choices are voluntary and do not need explanation."
- A 16mm film named Zorn's Lemma was presented at the 1970 New York Film Festival. I have not seen it. It is an "experimental" or "art" film -- i.e., it has not much plot -- but at least one viewer said it was compelling, beautiful and unforgettable.

All links tested 7 Nov 2005. Latest alterations 11 Nov 2009. My thanks to Andreas Blass, who assisted me with part of this page.