You are on page 1of 4

Leonard F.

Richardson, Measure and Integration

APPENDIX:
THE BANACH-TARSKI THEOREM

A.1 THE LIMITS TO COUNTABLE ADDITIVITY

In Section 1.1, we considered the pivotal role of the discovery of incommensurable


line segments in the development of Euclidean geometry. We identified commensu-
rability problems as belonging to the early development of measure theory. In the
pages that followed, we have learned much about Lebesgue’s theory of measure and
integration. We have seen how the Lebesgue theory yields complete normed linear
spaces such as LP(X.8.p). We have learned about the dual spaces of the latter
spaces, and about the dual of the space C ( X ) ,if X is a compact Hausdorff space. It
would be difficult to overstate the importance of Lebesgue measure and integration
throughout modern pure and applied analysis.
The reader has seen that much effort must be made to ensure that we deal only
with measurable sets and measurable functions in Lebesgue’s theory. We showed
in Example 3.1, using the Axiom of Choice, that no translation-invariant. countably
additive measure can be defined on all the subsets E E IR if

0 < p[O, 1) < cc


Measure arid Integration: A Corzcise Introduction to Real Aiialwis. By Leonard F. Richardson 225
Copyright @ 2009 John Wiley & Sons, Inc.
226 THE BANACH-TARSKI THEOREM

In this appendix, we will discuss a theorem that is as challenging for modem


mathematicians as incommensurable line segments were for their ancient Greek
forerunners. It is a theorem of Banach and Tarski, and it is often called the Banach-
Tarski Paradox. It is not a paradox in fact, but a theorem. This theorem is described
as a paradox because of the extraordinarily counterintuitive nature of its conclusions.
We begin with a definition.

Definition A . l . l Two sets A and B in a metric space ( X ,p ) are said to be congruent


by finite decomposition provided that there is a natural number n such that it is
possible to decompose A and B into disjoint unions

in such a way that Ak and Bk are congruent for each k < n. This is denoted by
f
A 2 B.
Congruence means that there exists a bijection Tk : Ak + Bk such that

p ( x . y) = p ( T k . E . T k y )

for all z and y in Ak. Such a mapping Tk is also called a bijective isometp. We
denote congruence as
Ak Z Bk.
We call A and B congruent by coiintable decomposition provided that the decom-
position of A and B into mutually congruent pieces can be accomplished by using
countably many pieces. This is denoted by

A 2 B.
One may prove readily as an exercise that a linear isometry of Euclidean space
must preserve the measure of any Lebesgue measurable set.
The reader will note that Steinhaus's' theorem (7.4.3) shows that two sets of the
same Lebesgue measure in the real line must be congruent by countable decomposi-
tion. The Banach-Tarski theorem is much more startling.

Theorem A.l.l (Banach-Tarski) Let A and B be two subsets of Rn, each having
f
nonempty interior, and with n 2 3. Then A 2 B. (For example, any two spherical
balls are congruent by3nite decomposition regardless of the difference in their radii.)
For dimensions 1 and 2 , A & B.

We remark that this theorem implies that for R", with n 2 3, there cannot exist
even afinitely additive measure defined on all subsets that is invariant under Euclidean

' A s an historical sidelight, we remark that Steinhaus played a pivotal role in Banach's decision to become
a professional mathematician. Further information is available in [ 171.
THE LIMITS TO COUNTABLE ADDITIVITY 227

motions. For n = 1 or n = 2 , there is no countably additive measure that is defined


on all subsets that is invariant under Euclidean motions. That is, in each of these two
sets of circumstances, nonmeasurable sets must exist.
The proof given by Banach and Tarski can be found in their original paper [ 2 ] .
There is also a modern treatise on the subject [25]. The reasoning comes primarily
from abstract set theory and from the study of groups of linear transformations acting
on vector spaces.
We give a simple example below that illustrates the Banach-Tarski theorem.

EXAMPLEA.l

We will show that there exists a subset S c [O. 2 ) , in the real line, for which
S IR. That is, S c [O,2) will be congruent by countable decomposition to
the entire real line. The congruence mappings will be translations.
The example begins with the proof in Example 3.1 that there exists a non-
measurable subset of the line. There we defined an equivalence relation:

r - y - y E $.

Note that each real number is equivalent modulo rational translation to numbers
in the interval [ O , l ) . We use the Axiom of Choice to select an uncountable
set C in [ O , 1 ) having the property that C consists of one element from each
equivalence class in IR/ -. The set C is called a cross section of IR/ -. If we
were still in Example 3.1, we would proceed to explain why the set C must be
nonmeasurable. Instead, the example takes a surprising turn. '
Let U = Q n [O. 11, and let C, = C + q. Define

the disjoint union of the countably many translates of C by q E U . Thus


S c [O, 2 ) . Let
r:U+$
be a bijection between the two countable sets, U and 0. Let
xq = T(4) - g

for each q E U . Observe that

as claimed.

'The author learned the following example from the website of Professor Terence Tao at UCLA. This
surprise ending for the famous example of a nonmeasurable set deserves to be better known. because it is
so simple, compelling, and delightful.
228 THE BANACH-TARSKI THEOREM

This example shows with convincing simplicity an instance of a congruence,


by countable decomposition, between two seemingly incongruous sets. It pro-
vides a compelling explanation of why it is necessary to check for measurability
in analysis. It gives also a startling sense of the geometrical possibilities of
nonmeasurable sets.
The theorem of Banach and Tarski demonstrates that the tools of Lebesgue measure
and integration, with all their strength, cannot measure all sets, even with firzite
additivity. More striking is the capacity of nonmeasurable sets to be reconfigured in
astounding ways, with geometrical perfection, losing nary a point.
Thanks to Lebesgue and many other mathematicians, the understanding of measure
and integration has been advanced in many ways that are invaluable to analysis. Yet
the mysteries that stand are more profound than those that came before. This is a
challenge and an invitation to the student to engage in the search to expand human
knowledge. It affirms for us the grandeur of truth, the finiteness of ourselves, and the
good fortune to be allowed a glimpse.

You might also like