The Banach–Tarski paradox is a theorem stating that a solid ball in three-dimensional Euclidean space can be partitioned into finitely many disjoint sets that can be moved by rotations and translations to form two disjoint solid balls, each identical in size to the original. Proved in 1924 by Stefan Banach and Alfred Tarski, it is a consequence of the axiom of choice. Its “pieces” are arbitrary sets of points, not ordinary solid fragments. The result is called a paradox because it contradicts familiar intuitions about volume, not because it produces a logical contradiction.
Precise statement
Let be a closed solid ball of positive radius. There exist a finite partition
and rigid motions such that
where and are disjoint balls congruent to . The symbol denotes a union of pairwise disjoint sets.
A rigid motion is an isometry: it preserves distances within each moved piece. For this theorem, rotations and translations suffice. No stretching, shrinking, or addition of points is permitted. Every point of the original ball belongs to exactly one piece, and the moved pieces fill the two target balls exactly.
This is a statement about the final positions of sets. It does not require the pieces to move along collision-free paths, nor does it describe a cutting process realizable with physical tools.
Two sets related by such a finite partition and rearrangement are called equidecomposable. In this terminology, the theorem says that a ball is equidecomposable with two disjoint copies of itself.
Why volume is not contradicted
For ordinary solids, rigid motions preserve volume, and the volume of a finite disjoint union equals the sum of the component volumes. If every piece in a Banach–Tarski decomposition had a volume satisfying those rules, then
which is impossible because has finite, positive volume.
The resolution comes from measure theory. Ordinary volume is formalized by Lebesgue measure, which is defined on a large collection of sets but not on every subset of Euclidean space. At least one piece in any Banach–Tarski decomposition must be non-Lebesgue-measurable.
A non-measurable set does not have volume zero, infinite volume, or an unknown ordinary volume. Rather, it lies outside the domain on which Lebesgue measure assigns volume consistently.
The issue is stronger than a failure of countable additivity. The decomposition uses only finitely many pieces. It therefore shows that, in three dimensions, no nonnegative volume function defined on all subsets can simultaneously:
- agree with ordinary volume on balls;
- remain unchanged under rigid motions;
- be finitely additive on disjoint sets.
Lebesgue measure avoids this contradiction by restricting which sets are measurable.
Structure of the proof
The proof combines an algebraic decomposition with a geometric construction. Its central ingredients are a free group of rotations, a choice of representatives from orbits, and methods for handling exceptional points.
Free groups and paradoxical decompositions
A free group on two generators and consists of reduced finite words built from
“Reduced” means that adjacent inverse pairs have been canceled. Apart from cancellations forced by inverses, there are no relations between the generators.
Such a group admits a paradoxical decomposition: it can be partitioned into finitely many subsets that group translations rearrange into two copies of the entire group. The construction exploits the way reduced words can be classified by their initial letters.
This is more than the familiar fact that an infinite set can have the same cardinality as two copies of itself. A cardinality argument allows arbitrary bijections. A paradoxical decomposition restricts the rearrangements to finitely many transformations supplied by the group.
Realizing the group by rotations
The rotation group of three-dimensional space contains subgroups isomorphic to the free group on two generators. Thus suitably chosen rotations and satisfy no nontrivial reduced-word relation.
These rotations define a group action on the unit sphere. The orbit of a point is the set of its images under all group elements:
There is an exceptional set of points fixed by nonidentity group elements. Every nonidentity rotation fixes exactly two points on the sphere, and the free group is a countable set, so the exceptional set is countable.
Outside this set, the action is free: no nonidentity group element fixes a point. Consequently, each orbit can be identified with the group itself.
Choosing representatives
The axiom of choice supplies one representative from each orbit outside the exceptional set. Once those representatives have been selected, the algebraic partition of the free group can be transferred to each orbit.
Combining the corresponding parts across all orbits produces a finite paradoxical decomposition of the sphere with the exceptional points removed.
This representative selection is the central nonconstructive step in the usual proof. It establishes that the required sets exist without giving an ordinary geometric description of their boundaries or a practical membership test for their points.
Recovering the whole sphere and solid ball
Additional rearrangements absorb the countable exceptional set. The underlying idea resembles a shift along an infinite sequence: a set can be moved so that it fills itself together with an extra point or countable collection of points.
The spherical decomposition then extends radially. For a subset of the unit sphere, consider
Rotations preserve the radial coordinate, so a decomposition of the sphere induces one of the punctured solid ball. The center is handled by a further absorption argument, and translations separate the two resulting balls.
Although intermediate stages use infinite sets and countable sequences, the final decomposition has only finitely many pieces.
Generalizations and dimensional limits
Arbitrary bounded bodies with interior
A stronger form states that any two bounded subsets of with nonempty interior are equidecomposable by finitely many rigid motions.
Their volumes and shapes need not agree. For example, a small solid ball and a much larger solid cube are equidecomposable in this set-theoretic sense. The hypotheses matter: this statement does not apply indiscriminately to every pair of subsets, such as a finite set and a ball.
Corresponding results hold in Euclidean dimensions greater than three.
Why the line and plane differ
There is no analogous finite duplication of a bounded interval or disk using one- or two-dimensional rigid motions.
In dimensions one and two, ordinary length or area can be extended to a nonnegative, finitely additive, isometry-invariant function defined on all subsets. Such extensions are not countably additive, but finite additivity already rules out doubling an interval or disk by a finite rigid-motion decomposition.
The distinction is therefore not simply that three-dimensional sets contain infinitely many points. Intervals, disks, and balls all have the same infinite cardinality. The decisive difference concerns the algebra of the allowed transformations.
Different planar equidecomposition theorems are possible when the initial and target sets have equal area. These do not amount to planar Banach–Tarski duplication.
Amenability and invariant measures
The theorem is closely connected with amenability, a concept in group theory concerning the existence of invariant averaging procedures.
For a discrete group, amenability can be expressed through a finitely additive probability measure defined on every subset and invariant under left translation. Such a measure prevents the group from admitting a paradoxical decomposition: finite additivity would force the total mass to equal twice itself.
The free group on two generators is not amenable. Its paradoxical decomposition provides the algebraic mechanism used in the Banach–Tarski theorem.
More generally, Tarski’s alternative relates paradoxical decompositions of group actions to the failure of suitable invariant finitely additive measures. This connection places Banach–Tarski within a broader theory of symmetry and invariant measure, rather than treating it as an isolated geometric curiosity.
History
In 1914, Felix Hausdorff established a precursor now called the Hausdorff paradox. He showed that a sphere, after a countable exceptional set is removed, admits a paradoxical decomposition under rotations.
Banach and Tarski’s 1924 work developed these ideas into the theorem for solid bodies and the stronger equidecomposability result for bounded sets with nonempty interior.
In 1929, John von Neumann investigated the group-theoretic structure behind such decompositions. His work introduced the ideas that developed into the theory of amenable groups, helping explain both the three-dimensional paradox and its failure for rigid motions in the plane.
Foundational status and interpretation
The Banach–Tarski theorem is provable in Zermelo–Fraenkel set theory with the axiom of choice, commonly abbreviated ZFC. It is not an inconsistency in that framework.
Its relation to choice requires care:
- The standard proof uses choice to select orbit representatives.
- The theorem is not equivalent to the full axiom of choice.
- It is not provable from Zermelo–Fraenkel set theory alone, assuming the usual consistency conditions.
The theorem concerns exact sets of mathematical points, not physical matter. Its pieces are not ordinary fragments separated by surfaces, and no physical procedure for producing them follows from their mathematical existence. It therefore supplies neither a method of duplicating material objects nor a violation of conservation of mass.
Its precise mathematical consequence is that ordinary geometric volume cannot be extended to every subset of three-dimensional space while retaining nonnegativity, rigid-motion invariance, and finite additivity.