proof of Banach-Tarski paradox
We deal with some technicalities first, mainly concerning theproperties of equi-decomposability. We can then prove the paradox in aclear and unencumbered line of argument: we show that, given two unitballs
and with arbitrary origin, and areequi-decomposable, regardless whether and are disjoint ornot. The original proof can be found in [BT].
Technicalities
Theorem 1.
Equi-decomposability gives rise to an equivalence relation on thesubsets of Euclidean space.
Proof.
The only non-obvious part is transitivity. So let , , besets such that and as well as and areequi-decomposable. Then there exist disjoint decompositions of and of such that and are congruent for . Furthermore, there existdisjoint decompositions of and of such that and are congruent for . Define
Now if and are congruent via some isometry , we obtain a disjoint decomposition of by setting. Likewise, if and arecongruent via some isometry , we obtain adisjoint decomposition of by setting. Clearly, the and define a disjoint decomposition of and , respectively, into parts. By transitivity of congruence
, and arecongruent for and . Therefore, and are equi-decomposable.∎
Theorem 2.
Given disjoint sets , such that and are equi-decomposable for , theirunions and areequi-decomposable as well.
Proof.
By definition, there exists for every , an integer such that there are disjoint decompositions
such that and are congruent for . Rewriting and in the form
gives the result.∎
Theorem 3.
Let , , be sets such that and are equi-decomposableand , then there exists such that and are equi-decomposable.
Proof.
Let and be disjoint decompositions of and , respectively, such that and are congruent viaan isometry for all . Let a map such that if . Since the are disjoint, is well-definedeverywhere. Furthermore, is obviously bijective. Now set and define , so that and are congruent for , so the disjoint union
and are equi-decomposable. Byconstruction, . Since is a proper subset
of and is bijective, is a proper subset of .∎
Theorem 4.
Let , and be sets such that and areequi-decomposable and . Then and areequi-decomposable.
Proof.
Let and disjoint decompositions of and , respectively, such that and are congruent viaan isometry for . Like in the proof oftheorem 3, let be the well-defined,bijective map such that if . Now,for every , let be the intersection of all sets satisfying
- •
,
- •
for all , the preimage
lies in ,
- •
for all , the image lies in .
Let such that and are notdisjoint. Then there is a such that for suitable integers and. Given , we have for a suitable integer ,that is , so that. The reverse inclusion followslikewise, and we see that for arbitrary either or and aredisjoint. Now set
then obviously . If for , consistsof the sequence of elements which is infinite
in bothdirections, then . If the sequence is infinite in only onedirection, but the final element lies in , then as well.Let and , then clearly .
Now let . If , then , so consists of a sequence which is infinite inonly one direction and the final element does not lie in . Now, but since does lie in , it is not the final element. Therefore thesubsequent element lies in , in particular and , so. It follows that , and furthermore and are disjoint.
It now follows similarly as in the preceding proofs that and are equi-decomposable. By theorem 2, and are equi-decomposable.∎
The proof
We may assume that the unit ball is centered at the origin, thatis , while the other unit ball has an arbitraryorigin.Let be the surface of , that is, the unit sphere. By theHausdorff paradox
, there exists a disjoint decomposition
such that , , and are congruent, and iscountable. For and , let be the set ofall vectors of multiplied by . Set
These sets give a disjoint decomposition of the unit ball with theorigin deleted, and obviously , , and are congruent (but is of course not countable). Set
where is the origin. and are triviallyequi-decomposable. Since and as well as and arecongruent, and are equi-decomposable. Finally, and as well as and are congruent, so and are equi-decomposable. In total, and are equi-decomposable by theorem 1, so and areequi-decomposable by theorem 2. Similarly, we conclude that, and are equi-decomposable.
Since is only countable but there are uncountably many rotationsof , there exists a rotation such that, so is a propersubset of . Since and areequi-decomposable, there exists by theorem 3 a proper subset such that and (and thus ) areequi-decomposable. Finally, let an arbitrary pointand set
a disjoint union by construction. Since and , and as well as and are equi-decomposable, and are equi-decomposable by theorem 2. and are disjoint (but !).
Now and are congruent, so and areequi-decomposable by theorem 1. If and aredisjoint, set . Otherwise we may use theorem 3 andchoose such that and areequi-decomposable. By theorem 2, and are equi-decomposable. Now we have
so by theorem 4, and are equi-decomposable.
References
- BT St. Banach, A. Tarski, Sur la décompositiondes ensembles de points en parties respectivement congruentes,Fund. math. 6, 244–277, (1924).